<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Carly K., M.S. Ed. on Medium]]></title>
        <description><![CDATA[Stories by Carly K., M.S. Ed. on Medium]]></description>
        <link>https://medium.com/@carlykayinnj?source=rss-64783a362389------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*2dGwfSn-fdEL68n8</url>
            <title>Stories by Carly K., M.S. Ed. on Medium</title>
            <link>https://medium.com/@carlykayinnj?source=rss-64783a362389------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 01:23:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@carlykayinnj/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Beyond the Test: Why Project-Based Learning Creates More Inclusive and Meaningful Learning]]></title>
            <link>https://carlykayinnj.medium.com/beyond-the-test-why-project-based-learning-creates-more-inclusive-and-meaningful-learning-1c9b0e9986c5?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c9b0e9986c5</guid>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[teaching]]></category>
            <category><![CDATA[project-based-learning]]></category>
            <category><![CDATA[instructional-design]]></category>
            <category><![CDATA[education]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Sat, 26 Sep 2026 17:38:25 GMT</pubDate>
            <atom:updated>2026-09-26T17:38:25.869Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A diverse group of high school students participate in project-based learning activities, including performing an ancient Egypt play, filming and editing video, photographing a product in a lightbox, collaborating on a computer, and sharing culturally significant foods. Students take on different creative and technical roles while a teacher supports the activities." src="https://cdn-images-1.medium.com/max/1024/1*LCcEeuFznMupG0ehvsbt6A.png" /><figcaption>This image was generated with artificial intelligence to visually represent concepts discussed in the article. It does not depict an actual classroom, students, or specific educational event.</figcaption></figure><h3>As an instructional designer and former classroom teacher, I have never believed that the best way to determine whether someone has learned something is to continuously give them tests, quizzes, and written assignments. Do those assessments have a place in education? Absolutely. Should they be the primary way we determine whether a student understands something? I do not believe they should.</h3><p>Some of the best learning experiences I have had, both as a student and as an educator, involved project-based learning (PBL). I remember those experiences because I was not simply memorizing information long enough to repeat it on a test. I was doing something with what I learned. That distinction matters.</p><p>Research supports what many educators have seen in their own classrooms. Chen and Yang’s (2019) meta-analysis examined 30 studies comparing project-based learning with more traditional instruction and found an overall positive effect on academic achievement, although the effects varied across subject areas and contexts. More recently, Farshad and Fortin (2026) reviewed 15 PBL meta-analyses and found a consistently positive direction across academic achievement, higher-order thinking, and 21st-century skills. At the same time, they cautioned that the methodological quality of the existing meta-analytic literature was limited, which is an important distinction when discussing what the research actually tells us.</p><p>For me, though, the value of project-based learning goes beyond test scores. It is about creating learning experiences where more students have an opportunity to participate, apply what they know, develop their strengths, and demonstrate what they can actually do.</p><h3>I Still Remember the Project. I Don’t Remember the Test.</h3><p>One of the best examples comes from my own experience as a student. Back in <strong>1999</strong>, when I was a sophomore in high school, my world history teacher gave us a choice for our unit on ancient Egypt. We could take a traditional test, write a paper, create a poster, write and perform a play in period-appropriate costumes, or select from several other options. We were put into groups, and my group chose to write and perform a play.</p><p>What stands out to me now, especially as an instructional designer, is how forward-thinking that approach was. At the time, I certainly was not thinking about Universal Design for Learning, autonomy, authentic assessment, or multiple means of action and expression. I was a high school student who was excited that I did not have to take another traditional test. Looking back more than 25 years later, however, I can see that my teacher gave us different ways to demonstrate the same learning while still maintaining expectations for what we needed to know.</p><p>I could not tell you what questions would have appeared on the test. I could not tell you what vocabulary words I would have been asked to define. What I <em>can</em> remember is researching ancient Egypt, determining what information mattered, writing a script, deciding how to communicate that information to an audience, putting together costumes, rehearsing, working with my classmates, and eventually performing what we created.</p><p>We had to know the content to make the project work. Giving us a choice did not remove academic rigor. If anything, it required us to understand the material deeply enough to transform what we learned into something new. We were not simply remembering information about Egypt. We were understanding it, applying it, analyzing it, making decisions about it, and ultimately creating something with it.</p><h3>Bloom’s Taxonomy Should Not Stop at “Remember”</h3><p>This is where Bloom’s Taxonomy becomes particularly important. The revised taxonomy moves from remembering and understanding toward applying, analyzing, evaluating, and creating (Anderson &amp; Krathwohl, 2001). The framework is also intended to help educators align learning objectives, instruction, and assessment rather than treating those pieces as separate from one another.</p><p>Tests and quizzes <em>can</em> assess higher-order thinking when they are designed well. I am not arguing otherwise. The problem is that assessment can very easily become a cycle of reading, memorizing, recalling, testing, and moving on. Project-based learning gives students opportunities to move further up Bloom’s Taxonomy because they have to take what they know and actually use it.</p><p>When my photography students learned about product photography, for example, I did not want them to simply identify lighting terminology on a quiz. They had to <strong>make a lightbox</strong>, determine how to light the product, position it, photograph it, evaluate the image, make adjustments, edit the photograph, place the finished image into an advertisement, and explain <em>why</em> they made their creative and technical decisions.</p><p>That one project incorporated technical knowledge, photography, lighting, composition, problem-solving, graphic design, editing, communication, and critical thinking. I could have asked, “What is a lightbox?” Instead, they built one and showed me they knew how to use it.</p><p>To me, the second tells us far more about whether meaningful learning actually occurred.</p><h3>My Students Learned the Most When They Were Creating</h3><p>I saw the same thing repeatedly during my years teaching television production. Some of my students’ strongest learning happened when they were working together to create something. They wrote, produced, filmed, and edited stop-motion animated films, commercials, broadcasts, and other video projects. Students had to develop an idea, research when necessary, write scripts, storyboard, plan shots, operate cameras, work with lighting and audio, direct, perform, edit, create graphics, meet deadlines, and troubleshoot the inevitable problems that come with production.</p><p>And there were always problems. A camera stopped cooperating. Someone forgot a line. Lighting looked great to the eye and terrible on camera. A shot that sounded amazing when students planned it did not work once they filmed it. Audio was bad. An edit did not work. Group members disagreed.</p><p>Students had to figure it out.</p><p><strong>That problem-solving was not an interruption to the learning. It was part of the learning.</strong></p><p>One of my favorite examples was our <strong>Ethnic Foods of the Holidays</strong> project. Students researched foods and traditions that were important to their families and then presented them as part of a television production. In many cases, parents or guardians became involved in preparing the dishes, bringing another layer of family and cultural connection into the assignment. Students appeared on camera to explain the tradition behind the food, how it was prepared, and why it was meaningful to their families. There were even conversations about those closely guarded “secret ingredients” that someone might acknowledge existed without actually giving them away.</p><p>Because the course ran for two marking periods with different groups of students, we produced the project twice during the school year. One version centered on the December holiday season, while the spring version provided another opportunity to explore the cultural and religious celebrations represented throughout our school community.</p><p>What made the project so valuable was that students were learning far more than television production skills. They were researching their own traditions, learning about the traditions of their classmates, determining what they wanted to communicate, presenting information to an audience, collaborating with a production team, and applying technical skills to produce an actual program.</p><p>Students had already learned the different roles involved in television production, so the project also gave them opportunities to take on responsibilities that aligned with their interests and strengths. Some wanted to be on camera. Others were much happier directing, producing, editing, operating cameras, working with audio, or taking responsibility for another technical role behind the scenes.</p><p>There was no expectation that every student had to contribute in exactly the same way in order to be valuable to the project.</p><p>My principal supported the project enough to give us a full school day for recording and collaboration. Students brought in their dishes, sampled one another’s foods, and talked about the similarities and differences among their families’ cultures and traditions. In a diverse public high school, the experience became about much more than producing a television segment. Students were learning from each other, asking questions, sharing pieces of their family traditions, and using the communication and production skills they had learned in class to create something meaningful together.</p><p>That type of learning is very difficult to reproduce with a traditional test or quiz.</p><h3>Project-Based Learning Does Not Have to Mean a Giant Project</h3><p>I also think it is important to clarify that authentic, application-based learning does not always have to mean a giant, multiweek production.</p><p>When I taught <strong>computer applications</strong>, I used a smaller version of the same approach. For each new task or skill, students first completed an introductory assignment so they could learn the basic process and become familiar with the software. Once they understood the mechanics, they were given a second assignment using the <strong>same skill</strong>, but the follow-up task was based on a real-world situation they had already encountered or were likely to encounter in the future.</p><p>That second assignment mattered.</p><p>Students were no longer completing steps simply because I told them which buttons to click. They had to take the skill they had just learned and determine how to use it appropriately within a realistic context. There is a significant difference between following directions to demonstrate that you know how a tool works and being placed in a situation where you have to determine how to use that tool to accomplish something.</p><p>The first develops foundational knowledge.</p><p>The second asks students to <strong>apply</strong> it.</p><p>Students often need direct instruction first. They may need demonstrations, guided practice, examples, and an introductory assignment before they are ready to independently apply a skill. Project-based learning does not mean eliminating those supports. It means <strong>not stopping there</strong>.</p><p>Once students know <em>how</em> to do something, the next question should be: <strong>Can they determine when, why, and how to use it on their own?</strong></p><h3>This Is Where Universal Design for Learning Comes In</h3><p>Project-based learning also aligns naturally with Universal Design for Learning when it is designed intentionally. CAST’s UDL Guidelines 3.0 emphasize multiple means of engagement, representation, and action and expression while placing particular emphasis on learner agency. The current guidelines include considerations related to choice and autonomy, relevance and authenticity, collaboration and interdependence, multiple media for communication, tools for creativity, graduated support, planning, and monitoring progress (CAST, 2024).</p><p>That is one reason I see such a natural connection between UDL and well-designed PBL.</p><p>We should not assume there is only one appropriate way for every student to access learning or demonstrate what they know. If the actual learning objective says students need to <strong>analyze</strong> something, why does every student necessarily need to write a five-page paper to prove they can analyze it? If the objective is to <strong>design</strong>, let them design. If it is to <strong>evaluate</strong>, let them evaluate. If it is to <strong>communicate</strong>, there may be several legitimate ways to communicate that understanding.</p><p>CAST’s Guidelines specifically include the use of multiple media for communication and multiple tools for construction, composition, and creativity, while also emphasizing graduated support for practice and performance (CAST, 2024).</p><p>That does <strong>not</strong> mean students should always be allowed to choose anything they want. Sometimes the format <em>is</em> the skill being assessed. If I am teaching students how to write an academic research paper, producing a video instead does not demonstrate mastery of academic writing.</p><p>But if the learning objective is to analyze the causes of the Civil War, demonstrate understanding of a scientific process, explain a mathematical concept, or propose a solution to an environmental problem, then we should at least ask ourselves whether there is an educational reason every student must demonstrate that knowledge through exactly the same format.</p><p>That is where inclusive instructional design begins.</p><h3>Inclusive Does Not Mean Easier</h3><p>This is something I believe is misunderstood far too often.</p><p>Providing options does not mean lowering expectations. Accessibility does not mean lowering expectations. UDL does not mean lowering expectations. Project-based learning does not mean students get to make something pretty instead of learning the material.</p><p>The learning objectives and expectations remain clear. <strong>What changes is the path students can take to demonstrate that they have met them.</strong></p><p>Going back to my high school Egypt project in 1999, my teacher did not tell us, “You don’t have to learn about Egypt because you’re doing a play.” The exact opposite happened. We had to understand the content well enough to transform it into another format and communicate it to someone else. That required research, comprehension, application, creativity, communication, and collaboration.</p><p>The academic expectation remained. The method changed.</p><p>The same thing happened with my own students. Producing <em>Ethnic Foods of the Holidays</em> did not allow students to avoid the skills and objectives of television production. They had to <strong>use</strong> those skills to produce something authentic. Product photography did not allow students to avoid learning lighting, composition, or editing. They had to apply those concepts successfully. The real-world computer applications assignments did not replace foundational instruction. They required students to demonstrate that they could <strong>transfer</strong> those skills into another situation.</p><p>That transfer is where learning becomes much more meaningful.</p><h3>Choice and Autonomy Matter</h3><p>One of the things my world history teacher understood back in 1999, long before I ever learned the terminology behind it, was the importance of autonomy. He gave us a choice, and that one decision changed how I approached the assignment.</p><p>The current UDL Guidelines explicitly include optimizing choice and autonomy as part of engagement and learner agency (CAST, 2024). Research on student motivation supports the broader importance of autonomy as well. Bureau et al. (2022), in a meta-analysis drawing from 144 studies, examined pathways leading to autonomous and controlled forms of student motivation and found strong connections among autonomy-supportive environments, students’ psychological needs, and autonomous motivation.</p><p>Again, autonomy does not mean a free-for-all. Students still need boundaries, learning objectives, rubrics, deadlines, and clear expectations. The difference is that we can create structured opportunities for them to make meaningful decisions within those boundaries.</p><p>Maybe they choose the topic. Maybe they choose the format. Maybe they choose their role within a team. Maybe they select a tool. Maybe they determine how they are going to solve the problem.</p><p>In <em>Ethnic Foods of the Holidays</em>, students were working toward the same larger production goals, but they did not all have to perform the same role. One student’s strength might have been presenting on camera while another student’s strength was editing, directing, organizing, writing, audio, or camera operation.</p><p><strong>That is autonomy with structure.</strong></p><h3>Multiple Intelligences: Looking at What Students Bring to the Table</h3><p>Howard Gardner’s Theory of Multiple Intelligences has also influenced the way I look at students and the strengths they bring into a learning environment. Gardner (1983) challenged the idea that intelligence should be viewed only as a single general capacity and proposed several relatively distinct intellectual capacities (Gardner, 1983).</p><p>I do not use Multiple Intelligences to put students into boxes or decide that one student can only learn in one particular way. Gardner himself has specifically cautioned against collapsing Multiple Intelligences into the concept of “learning styles” (Gardner, 2013).</p><p>For me, the value is much simpler: <strong>students bring different strengths into the classroom.</strong></p><p>One student may be an exceptional writer. Another may understand visual composition almost immediately. One might be technically gifted and figure out editing software before anyone else. Another may be the person who can organize five completely different personalities and somehow keep everyone moving toward a deadline. Someone may thrive on camera while another never wants to appear on camera but can write an incredible script.</p><p>A well-designed project gives more of those strengths an opportunity to become visible.</p><p>I saw that constantly in television production. A student did not need to be the strongest writer, performer, editor, camera operator, director, and producer all at once. Production itself depends upon people bringing different strengths together to create a final product.</p><p>That is also one of the reasons projects such as <em>Ethnic Foods of the Holidays</em> were so inclusive. There were multiple legitimate ways for a student to make a meaningful contribution while still working toward shared learning objectives.</p><p>A traditional test often does not provide that same opportunity.</p><h3>Scaffolding Is What Keeps PBL From Becoming Chaos</h3><p>There is an important caveat to all of this. You cannot simply give students a giant project on Monday, tell them it is due six weeks later, and call that project-based learning.</p><p><strong>That is not instructional design.</strong></p><p>Large projects can require substantial planning, organization, time management, self-monitoring, collaboration, and task initiation. Those demands can become barriers when we assume students already know how to manage all of them. That is where scaffolding becomes critical.</p><p>A large project can be divided into manageable stages: research, a proposal, an outline, script, storyboard, or project plan, followed by a draft, prototype, rough cut, or initial design, then feedback, revision, and eventually the finished product. Students should know what is expected before they begin. They should see the rubric. They should have models or examples when appropriate. Checklists, templates, milestones, mini-lessons, instructor feedback, peer feedback, and progress checks can all help students manage a complex learning experience.</p><p>Bardack and Obradović (2019), for example, examined teachers’ executive-function-related behaviors and scaffolding in elementary classrooms and found that teachers’ executive-function scaffolding positively predicted students’ later executive-function performance. Although that study focused on elementary classrooms rather than PBL specifically, it reinforces an important point: we should intentionally support planning and self-management rather than simply assuming those abilities are already in place.</p><p>The goal is not to hold students’ hands forever. The goal is to give them enough structure to be successful and then gradually allow them to take more responsibility for the process.</p><p>That is how autonomy develops.</p><p>My computer applications courses followed the same principle on a smaller scale. The introductory activity provided the scaffold. Students learned the skill with guidance first. Then the authentic follow-up assignment removed some of that support and required them to determine how to apply the same skill within a realistic situation.</p><p>That gradual shift from <strong>show me you can follow the process</strong> to <strong>show me you can use the process independently</strong> is an important part of learning.</p><h3>Group Projects Are Not the Problem. Poorly Designed Group Projects Are.</h3><p>Whenever I talk about project-based learning, group work inevitably comes up. Everyone has been in <em>that</em> group: one person does everything, one person disappears, and someone shows up five minutes before the presentation asking, “What am I supposed to say?”</p><p>We have all been there.</p><p>But that is not a reason to eliminate collaborative learning. It is a reason to <strong>design it better</strong>.</p><p>Collaboration is itself a skill. Students need opportunities to learn how to divide responsibilities, communicate expectations, manage conflict, meet deadlines, give and receive feedback, and depend on other people. UDL Guidelines 3.0 explicitly includes fostering collaboration, interdependence, and collective learning as part of sustaining effort and persistence (CAST, 2024).</p><p>Television production made this especially visible because successful production genuinely depended on collaboration. A director could not do every job. An editor needed usable footage from the camera operators. On-camera talent depended on the script. Audio and lighting affected the quality of everything being recorded. Everyone’s work connected to someone else’s.</p><p>Projects like <em>Ethnic Foods of the Holidays</em> gave students an authentic reason to collaborate because they were producing something that could not realistically be completed well by one person working in isolation.</p><p>As educators, however, we still need accountability. Defined roles, individual reflections, production logs, peer evaluations, progress meetings, drafts, individual components, or having each student explain their contribution can all help prevent collaborative learning from becoming “one student does everything.”</p><p>The final product matters.</p><p><strong>So does the process that created it.</strong></p><h3>Students Need to Understand the “Why”</h3><p>One of the most important questions students ask is also one of the simplest:</p><p><strong>Why do I need to know this?</strong></p><p>Sometimes they ask it out loud. Sometimes you can see it written all over their face.</p><p>Project-based learning and authentic assessment can help answer that question because students are using what they are learning for a purpose. Photography concepts matter because the student needs to create the photograph. Lighting matters because the shot looks terrible without it. Research matters because the script needs accurate information. Writing matters because someone actually has to perform the words. Computer application skills matter because students recognize how the same tools can be used in situations outside of school.</p><p>In <em>Ethnic Foods of the Holidays</em>, research mattered because students were communicating something real about their families and cultures. Writing mattered because they had an audience. Production skills mattered because they were creating an actual program. Collaboration mattered because the final product depended upon the group.</p><p>Suddenly, the content has context.</p><p>That does not mean every lesson needs to become a giant semester-long project. Sometimes direct instruction is exactly what is needed. Sometimes students need practice. Sometimes a quiz is useful. Sometimes writing is the most appropriate assessment.</p><p>Inclusive instructional design is not about replacing one rigid model with another.</p><p>It is about asking whether the learning experience and the assessment actually match what we want students to <strong>know and be able to do</strong>.</p><h3>What Are We Really Measuring?</h3><p>Ultimately, that is the question I keep coming back to as an instructional designer:</p><p><strong>What are we actually trying to measure?</strong></p><p>If I want to know whether a student can remember terminology, a quiz may be completely appropriate. If I want to know whether that student can apply the terminology, solve a problem, evaluate a situation, design something, communicate an idea, collaborate with other people, or create something new, I need to give them an opportunity to actually do those things.</p><p>Students are going to leave our classrooms and enter environments where they will be expected to solve problems that do not come with four answer choices. They will have to communicate, collaborate, make decisions, and create. They will make mistakes. They will have to adjust when their first solution does not work. They will have to explain why they made a particular decision.</p><p>Those are exactly the kinds of skills that well-designed project-based learning can help develop.</p><p>I still believe tests, quizzes, papers, and traditional written assignments have a place in education. I use them. I design them. I just do not believe they should be the <strong>only</strong> ways students are allowed to show us what they know.</p><p>Some of the best learning I experienced as a student happened because, back in 1999, a teacher trusted me enough to give me a choice. Some of the best learning I saw as a teacher happened when my students were researching, building, writing, filming, editing, designing, troubleshooting, sharing their cultures, learning from one another, arguing over ideas, starting over, and eventually looking at something they had created and saying, <strong>“We made that.”</strong></p><p>Even in a computer applications classroom, some of the most meaningful learning happened when students moved beyond the introductory assignment and had to take the same skill into a situation that actually looked like something they might encounter outside of school.</p><p>Those are the learning experiences students remember.</p><p>More importantly, those are often the experiences where students learn how to actually <strong>use</strong> what we taught them.</p><p>And to me, that is the point.</p><h3>References</h3><p>Anderson, L. W., &amp; Krathwohl, D. R. (Eds.). (2001). <em>A taxonomy for learning, teaching, and assessing: A revision of Bloom’s taxonomy of educational objectives</em>. Longman. <a href="https://books.google.com/books/about/A_Taxonomy_for_Learning_Teaching_and_Ass.html?id=bcQlAQAAIAAJ&amp;utm_source=chatgpt.com">Book record and publication information</a></p><p>Bardack, S., &amp; Obradović, J. (2019). Observing teachers’ displays and scaffolding of executive functioning in the classroom context. <em>Journal of Applied Developmental Psychology, 62</em>, 205–219. <a href="https://doi.org/10.1016/j.appdev.2018.12.004">https://doi.org/10.1016/j.appdev.2018.12.004</a>. <a href="https://www.sciencedirect.com/science/article/abs/pii/S0193397317303696?utm_source=chatgpt.com">Article and DOI</a></p><p>Bureau, J. S., Howard, J. L., Chong, J. X. Y., &amp; Guay, F. (2022). Pathways to student motivation: A meta-analysis of antecedents of autonomous and controlled motivations. <em>Review of Educational Research, 92</em>(1), 46–72. <a href="https://doi.org/10.3102/00346543211042426">https://doi.org/10.3102/00346543211042426</a>. <a href="https://pubmed.ncbi.nlm.nih.gov/35330866/?utm_source=chatgpt.com">Article and DOI</a></p><p>CAST. (2024). <em>CAST Universal Design for Learning Guidelines version 3.0</em>. <a href="https://udlguidelines.cast.org/?utm_source=chatgpt.com">CAST UDL Guidelines 3.0</a></p><p>Chen, C.-H., &amp; Yang, Y.-C. (2019). Revisiting the effects of project-based learning on students’ academic achievement: A meta-analysis investigating moderators. <em>Educational Research Review, 26</em>, 71–81. <a href="https://www.sciencedirect.com/science/article/pii/S1747938X19300211?utm_source=chatgpt.com">Article on ScienceDirect</a></p><p>Farshad, S., &amp; Fortin, C. (2026). An umbrella review of meta-analyses on project-based learning: Effects on academic achievement, higher-order thinking, and 21st-century skills. <em>Educational Research Review, 52</em>, 100809. <a href="https://doi.org/10.1016/j.edurev.2026.100809">https://doi.org/10.1016/j.edurev.2026.100809</a>. <a href="https://www.sciencedirect.com/science/article/abs/pii/S1747938X26000485?utm_source=chatgpt.com">Article and DOI</a></p><p>Gardner, H. (1983). <em>Frames of mind: The theory of multiple intelligences</em>. Basic Books. <a href="https://books.google.com/books/about/Frames_Of_Mind.html?id=_hWdAAAAMAAJ&amp;utm_source=chatgpt.com">Book record</a></p><p>Gardner, H. (2013, October 16). Multiple intelligences are not learning styles. <em>Harvard Graduate School of Education</em>. <a href="https://www.gse.harvard.edu/ideas/news/13/10/multiple-intelligences-are-not-learning-styles?utm_source=chatgpt.com">Harvard Graduate School of Education article</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1c9b0e9986c5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Inside Garfield Public Schools’ AI Pilot: New Records Reveal the Scope of Student Data Being Sent…]]></title>
            <link>https://carlykayinnj.medium.com/inside-garfield-public-schools-ai-pilot-new-records-reveal-the-scope-of-student-data-being-sent-dc8c0ff1006b?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/dc8c0ff1006b</guid>
            <category><![CDATA[data-privacy]]></category>
            <category><![CDATA[special-education]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Fri, 18 Sep 2026 18:08:12 GMT</pubDate>
            <atom:updated>2026-09-18T18:08:12.073Z</atom:updated>
            <content:encoded><![CDATA[<h3>Inside Garfield Public Schools’ AI Pilot: New Records Reveal the Scope of Student Data Being Sent to Parthion</h3><h4>The integration includes PowerSchool records, assessment data, IEPs, evaluations, disciplinary information, and even records from graduated students. The newest public records provide a clearer picture of what is going into Parthion, while raising new questions about storage, cybersecurity, retention, and access.</h4><figure><img alt="Editorial illustration showing student data flowing from a public school into a glowing cloud-based AI system. Digital records representing attendance, grades, assessments, IEPs, disability accommodations, discipline, and graduated students move toward centralized servers protected by a lock and shield, while a shadowed hacker figure and red warning symbols represent cybersecurity risks." src="https://cdn-images-1.medium.com/max/1024/1*sAwJMlUZrk5W1JIkHFdXUQ.png" /><figcaption>This image was created with generative AI for illustrative purposes. It is a conceptual representation of student data, cloud-based technology, and cybersecurity concerns discussed in this article and does not depict actual Garfield Public Schools systems, facilities, students, employees, Parthion infrastructure, or a real cybersecurity incident.</figcaption></figure><p>When I first began examining Garfield Public Schools’ pilot of Parthion, one of the most basic questions was also one of the most important: <strong>What student information will this system actually receive?</strong></p><p>Earlier public records provided part of the answer. Garfield’s Pilot Agreement identified an Early Warning System and an IEP Sandbox Module, while later records offered additional information about the platform, vendor due diligence, cybersecurity, and the district’s implementation process. The newest OPRA production provides something different. It shows, in considerably greater detail, how information from several of Garfield’s existing systems is being mapped, exported, uploaded, and integrated into Parthion.</p><p>On September 18, Garfield provided additional records consisting of emails involving Parthion and district technology personnel. The district also stated that it was continuing to gather additional emails responsive to my request.</p><p>Those emails move the conversation beyond general descriptions of what Parthion <em>could</em> access. They show what district and vendor personnel are actually working to connect.</p><p>And it is substantially more than a student’s name, grade level, and class schedule.</p><h3>From PowerSchool to Parthion</h3><p>One series of emails documents Parthion’s work mapping data from Garfield’s PowerSchool student information system.</p><p>The information discussed includes:</p><ul><li>student identification information</li><li>grade level</li><li>school and building information</li><li>state-reporting school codes</li><li>courses and sections</li><li>internal class codes</li><li>school calendars and in-session dates</li><li>period-level attendance</li><li>daily attendance status</li><li>behavioral information</li><li>disciplinary incidents</li><li>suspensions</li><li>detentions</li><li>response and consequence codes</li></ul><p>Attendance is not limited to a simple number of absences. The correspondence discusses period-level information and how that data can be interpreted to determine whether a student was present or absent for the day.</p><p>Behavioral information is also part of the mapping process. The records reference both PowerSchool’s Behavior Log and Incident Management functions, meaning Parthion is not being designed solely around academic performance.</p><p>Taken together, the emails describe a system capable of bringing information that currently exists in different areas of PowerSchool into one student profile.</p><p>There can be a legitimate educational benefit to doing that. Educators routinely have to work across systems to identify patterns and determine whether a student needs support. A centralized platform may make that work more efficient.</p><p>Centralization, however, also changes the privacy and cybersecurity implications.</p><p>An isolated attendance entry reveals relatively little. The same can be said for one assessment score or one disciplinary incident. Combine years of attendance, assessment, behavior, course performance, interventions, and special-education information under one student profile, and the resulting record becomes considerably more revealing.</p><h3>Assessment Data Is Also Being Integrated</h3><p>The records show work underway to connect Renaissance Star assessment information with Parthion.</p><p>The correspondence discusses recurring data transfers and potential technical methods including APIs and SFTP. Parthion indicated that it would provide Renaissance with the required fields, secure-transfer information, and technical contacts needed for the integration.</p><p>That means assessment information is not simply something a teacher may choose to type into Parthion manually. The platform is being designed to receive that information directly or through recurring transfers.</p><p>Parthion also asked district personnel where Garfield’s NJSLA scores were stored after apparently being unable to locate them through the PowerSchool integration.</p><p>That correspondence shows an effort to locate the data. It does <strong>not</strong>, based on the records I reviewed, establish that NJSLA scores were ultimately imported into Parthion.</p><p>That distinction matters. Public records should be used to document what actually occurred, not to fill gaps with assumptions.</p><h3>The Special-Education Data Is Far More Extensive</h3><p>Perhaps the most significant new information involves Garfield’s Realtime special-education system.</p><p>Earlier discussion of an “IEP Sandbox Module” did not fully convey the volume or sensitivity of the information involved. The newer records show planning to move existing special-education information into Parthion, including actual IEP documents.</p><p>Parthion sought PDF copies of existing IEPs so those documents could be uploaded into the system. Realtime separately described providing a ZIP file containing <strong>“all student IEPs and other special education documents,”</strong> excluding IEPs that had not yet been finalized.</p><p>The migration material identifies numerous categories of special-education information, including:</p><ul><li>IEP documents and structured IEP information</li><li>special-education rosters</li><li>eligibility and program information</li><li>case-manager information</li><li>meeting and agreement information</li><li>evaluations</li><li>evaluator reports</li><li>related services</li><li>supplementary aids</li><li>modifications</li><li>accommodations</li><li>testing accommodations</li><li>assistive technology</li><li>special considerations</li><li>transition information</li><li>assessment participation</li><li>assessment exemptions</li><li>removal from general-education settings</li><li>plan status and duration</li><li>other special-education process-tracking information</li></ul><p>Evaluator reports may also be uploaded into Parthion and combined through what the vendor describes as a collaborative report engine.</p><p>That is considerably different from simply saying the district is testing software related to IEP development.</p><p>The records describe a platform that may hold actual special-education documents alongside structured student information.</p><p>For students receiving special-education services, these files can contain some of the most personal information maintained by a school district. That makes questions about who can access those records, where they are stored, how long they remain there, and how they are secured especially important.</p><h3>Graduated Students Are Being Included</h3><p>One of the most significant exchanges in the new records involves students who no longer attend Garfield Public Schools.</p><p>During the PowerSchool mapping process, Parthion encountered students assigned <strong>Grade Level 99</strong> and asked the district what that designation meant and whether those students should be included.</p><p>The district’s response was explicit:</p><blockquote>“Grade level 99 is our graduated students. I would include them in case there is data we need.”</blockquote><p>That exchange removes much of the uncertainty surrounding whether former students might be part of the Parthion integration.</p><p>Parthion identified the records.</p><p>Parthion asked whether they should be included.</p><p>Garfield instructed the vendor to include them.</p><p>The unanswered question is now not <strong>whether</strong> graduates are included, but <strong>which graduates, how much of their historical information is included, and how long that information will remain in the platform</strong>.</p><p>The records reviewed so far do not explain whether Grade Level 99 includes students who graduated one year ago, five years ago, ten years ago, or longer. They also do not establish the complete set of records Parthion will receive for those former students.</p><p>If a graduate had an IEP, evaluations, accommodations, disciplinary incidents, attendance history, assessment results, or other educational records, which of those records will accompany that student into Parthion?</p><p>That question becomes especially important because those individuals are no longer receiving educational services from Garfield.</p><p>The stated reason for including them was that there may be data the district needs. There may be a legitimate administrative reason for historical access, but the records raise a broader data-governance question:</p><p><strong>How much historical student information does a new technology platform need, and for how long?</strong></p><h3>How Is All of This Information Getting Into Parthion?</h3><p>The new records provide a fairly clear picture of the methods being used or considered to transfer information.</p><p>They include:</p><ul><li>API integrations</li><li>SFTP transfers</li><li>CSV exports</li><li>Excel files</li><li>recurring data feeds</li><li>direct document uploads</li><li>ZIP files containing documents</li><li>data-field mapping between existing systems and Parthion</li></ul><p>Those methods tell us a great deal about <strong>ingestion</strong>.</p><p>They tell us considerably less about <strong>storage</strong>.</p><p>Once a student record leaves PowerSchool, Realtime, Renaissance, or another district system and becomes part of Parthion, a different set of questions applies.</p><h3>Where Will the Data Be Stored?</h3><p>The records show that Parthion is being configured to receive both structured student data and actual documents.</p><p>What remains much less clear is the full lifecycle of that information once Parthion receives it.</p><p>The records reviewed so far do not clearly establish:</p><ul><li>exactly where Garfield student data is physically hosted</li><li>where uploaded IEPs and evaluations are stored</li><li>the geographic location of the servers</li><li>whether all student information remains within the United States</li><li>how Garfield’s records are separated from other Parthion customers</li><li>how long current-student records are retained</li><li>how long graduate records are retained</li><li>whether graduation triggers any archival or deletion process</li><li>whether deleted records remain in system backups</li><li>how long backups are retained</li><li>which vendor employees can access Garfield’s production data</li><li>whether vendor developers or technical-support personnel can access actual student records</li><li>what happens to all district data when Garfield stops using Parthion</li><li>how data is securely destroyed at the end of the relationship</li></ul><p>These are not obscure technical questions.</p><p>They determine what happens to a student’s information after it leaves the district’s existing systems.</p><h3>What Protects the Information From Hackers?</h3><p>That leads to the other major question raised by the expanded data integration: <strong>What happens if someone attempts to gain unauthorized access?</strong></p><p>The new emails reference secure-transfer information and technologies such as SFTP. Protecting information while it moves between systems is important.</p><p>But encryption or secure transmission during transfer addresses only one part of the cybersecurity problem.</p><p>Once student information reaches Parthion, the protections surrounding the stored data become equally important.</p><p>Earlier public records provided a SOC 2 Type I report associated with Lantern BRP and Parthion. That examination provided independent evidence concerning the design of certain security controls at a particular point in time. It did not test whether those controls operated effectively over an extended period.</p><p>The auditor expressly stated that operating effectiveness was outside the scope of that Type I examination. The records I previously received did not establish completion of a SOC 2 Type II examination, which would evaluate operating effectiveness over a defined period.</p><p>That does <strong>not</strong> mean Parthion’s security controls failed or that student information has been compromised.</p><p>It means the level of assurance provided by the released documentation has limits.</p><p>That distinction matters even more now that the data environment is clearer.</p><p>We are no longer discussing a hypothetical system with an undefined collection of student information. The integration records describe current and former students, attendance, discipline, assessments, IEPs, evaluations, accommodations, related services, and other sensitive educational information.</p><p>The original OPRA request specifically sought records concerning cybersecurity assessments, privacy protections, vendor evaluations, due diligence, and safeguards used to protect sensitive student information.</p><p>Important cybersecurity questions therefore remain, including:</p><ul><li>Is student information encrypted while stored, not just while being transferred?</li><li>Are uploaded IEPs and evaluations encrypted at rest?</li><li>Are system backups encrypted?</li><li>Is multi-factor authentication required?</li><li>Is MFA required for privileged and administrative accounts?</li><li>How are administrator privileges assigned?</li><li>Can vendor employees access student records?</li><li>Is access based on least privilege and legitimate need?</li><li>Is every access to sensitive student information logged?</li><li>Can Garfield audit who viewed or downloaded a student’s record?</li><li>Are suspicious login attempts automatically detected?</li><li>What intrusion-detection or prevention systems are used?</li><li>How frequently are vulnerabilities scanned?</li><li>How frequently is penetration testing conducted?</li><li>Who performs that testing?</li><li>What happens when a vulnerability is discovered?</li><li>How are software patches managed?</li><li>How is one school district’s data isolated from another’s?</li><li>How are compromised credentials handled?</li><li>What protections exist against ransomware?</li><li>Are backups protected from alteration or deletion?</li><li>How quickly must Parthion notify Garfield of unauthorized access?</li><li>Who determines whether families must be notified?</li><li>What third-party subprocessors can access Garfield’s information?</li><li>What cybersecurity requirements apply to those companies?</li></ul><p>Those questions are not evidence that a breach has occurred.</p><p>They are questions about preparedness for a risk that exists for virtually every system holding sensitive information.</p><h3>Centralization Can Help Educators. It Also Concentrates Risk.</h3><p>There is an educational argument for bringing information together.</p><p>Teachers, counselors, administrators, intervention teams, and child study teams often have to navigate multiple systems to understand a student’s complete situation. A platform that consolidates relevant information may help professionals recognize patterns sooner, coordinate support more efficiently, and spend less time moving between disconnected databases.</p><p>That potential deserves to be acknowledged.</p><p>So does the other side of centralization.</p><p>When student information is distributed among several systems, compromising one system may expose one category of information. When those systems continuously feed information into a central platform, one compromised account, application vulnerability, configuration error, or successful intrusion can potentially expose much more.</p><p>That does not mean Parthion will experience a breach.</p><p>It means the <strong>consequences</strong> of a breach should be considered alongside the likelihood of one occurring.</p><p>The more information a platform contains, the more important access controls, monitoring, encryption, data minimization, retention policies, backup security, incident response, and vendor oversight become.</p><h3>There Is Still Another Question: What Information Is Actually Being Processed by AI?</h3><p>The records also leave an important distinction unresolved.</p><p>Which information is merely <strong>stored or displayed</strong> in Parthion, and which information is actually <strong>processed by artificial intelligence</strong>?</p><p>Those are not necessarily the same thing.</p><p>The records reviewed so far do not clearly establish whether entire IEPs, evaluator reports, behavioral records, or other sensitive documents are submitted to an external AI model.</p><p>They also do not clearly establish whether student information is:</p><ul><li>sent to a third-party AI provider</li><li>included in prompts</li><li>retained in prompt logs</li><li>converted into embeddings</li><li>placed into a vector database</li><li>stored separately for AI processing</li><li>used to improve algorithms</li><li>used to fine-tune models</li><li>retained by an external AI provider</li><li>incorporated into AI-generated student recommendations</li></ul><p>The absence of an answer does not mean any of those practices are occurring.</p><p>It means the public records produced so far do not provide enough information to determine whether they are.</p><p>The appropriate response to that uncertainty is not speculation.</p><p>It is disclosure.</p><h3>The New Records Change the Conversation</h3><p>When I began this series, many questions about Parthion’s data environment remained largely theoretical.</p><p>The newest records change that.</p><p>We now know that:</p><ul><li>PowerSchool student information is being mapped into the system.</li><li>Attendance information is part of that process.</li><li>Behavior and discipline information are involved.</li><li>Renaissance Star assessment data is being integrated.</li><li>Parthion sought information about where NJSLA scores were stored.</li><li>Realtime special-education information is being prepared for migration.</li><li>Actual IEP documents are involved.</li><li>Evaluations and accommodations are part of the available special-education data.</li><li>Evaluator reports may be uploaded.</li><li>Graduated students coded as Grade Level 99 were specifically identified.</li><li>Garfield explicitly instructed Parthion to include those graduates.</li></ul><p>Those facts make the unanswered questions more consequential.</p><p><strong>Where is all of this information going?</strong></p><p><strong>Who will be able to see it?</strong></p><p><strong>How long will it remain there?</strong></p><p><strong>How far back do the graduate records extend?</strong></p><p><strong>What gets deleted, and when?</strong></p><p><strong>What happens to information stored in backups?</strong></p><p><strong>What prevents an unauthorized user from gaining access?</strong></p><p><strong>How would Garfield know if someone did?</strong></p><p><strong>And which parts of these student records are actually being analyzed by artificial intelligence?</strong></p><h3>Transparency Is Not Opposition to Technology</h3><p>As I have emphasized throughout this series, asking these questions is not an argument against artificial intelligence.</p><p>AI may ultimately provide valuable tools for educators. It may help identify students who need additional support, reduce administrative work, organize information that currently exists across multiple systems, and give professionals a more complete picture when making educational decisions.</p><p>Responsible innovation, however, requires more than functionality.</p><p>It requires knowing what information is being collected, why it is needed, who has access to it, where it is stored, how long it is retained, how it is protected, what role artificial intelligence plays in processing it, and what happens if those safeguards fail.</p><p>Those expectations become particularly important when the information involves children, students with disabilities, and former students who may have left the district years ago.</p><p>The September OPRA production provides the clearest picture yet of how much student information is being prepared for Parthion.</p><p>It also makes the next question difficult to avoid:</p><p><strong>Now that we have a much clearer picture of what is going into the system, do the public records provide an equally clear picture of how all of that information will be protected once it gets there?</strong></p><p>So far, that part of the story remains incomplete.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dc8c0ff1006b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[25 Years Later: My 9/11/01 Story - The Car Accident, My Father’s Sixth Sense, and the Day That…]]></title>
            <link>https://carlykayinnj.medium.com/25-years-later-my-9-11-01-story-the-car-accident-my-fathers-sixth-sense-and-the-day-that-f33598abb164?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/f33598abb164</guid>
            <category><![CDATA[new-york-city]]></category>
            <category><![CDATA[world-trade-center]]></category>
            <category><![CDATA[new-york]]></category>
            <category><![CDATA[9-11-attacks]]></category>
            <category><![CDATA[new-jersey]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Fri, 11 Sep 2026 07:10:29 GMT</pubDate>
            <atom:updated>2026-09-11T15:37:38.932Z</atom:updated>
            <content:encoded><![CDATA[<h2>25 Years Later: My 9/11/01 Story - The Car Accident, My Father’s Sixth Sense, and the Day That Changed Everything</h2><figure><img alt="" src="https://cdn-images-1.medium.com/max/724/1*Zazx_VYIDByDCYsTsvJFkA@2x.jpeg" /><figcaption>Needlepoint of the New York City skyline, including the World Trade Center, that I made in 2001.</figcaption></figure><p>The day before September 11, 2001, I was involved in a car accident that completely totaled my parents’ car, a Ford Tempo. My original plan had been to go into New York City with a friend the following morning to do some sightseeing, and we had even talked about visiting the World Trade Center’s Observation Deck, as she had never been. After the accident, however, my father insisted that I go to class instead. He kept saying he did not have a good feeling about me going into the city the next day… and was adamant about that. When my father gets that sixth sense, nine times out of ten, he is right. At the time, I did not think much about it, and thought he was once again just being overprotective. Twenty-five years later, I do.</p><h3><strong>A Perfect Fall Morning</strong></h3><p>I was 18 years old, and September 11, 2001, was only my second day as a freshman at William Paterson University in Wayne, New Jersey. The sky was clear, the air was crisp, and it seemed like a perfect fall morning.</p><p>The same friend I had planned to go into the city with drove me to campus that morning since my car had been totaled the day before. As we traveled north on Route 208 toward my 9:00 AM Music Appreciation class, we were listening to Elvis Duran and the Z Morning Zoo on Z100 when reports came in that a plane had struck the World Trade Center. At first, the hosts believed it might have been a helicopter or small prop plane accident. Earlier that summer, a French stuntman had become stranded on the Statue of Liberty, so the initial reaction was confusion rather than alarm. In fact, my friend and I looked at each other and jokingly said, “Who the hell crashes a plane into the World Trade Center?” Nothing about that moment prepared me for what was actually happening.</p><p>I continued to class as if it were any other day. It was not until I left that first course and checked my cellphone that I realized something far more serious was happening. I had a frantic voicemail from my mother telling me that I was absolutely not going into the city that night because New York was under attack by terrorists. I had tickets to see O-Town at the Hammerstein Ballroom that evening, something that only hours earlier had been just another normal plan on my calendar. When I tried calling my mother back, all I could get was a busy signal.</p><h3><strong>Watching the Skyline I Grew Up With Disappear</strong></h3><p>As I walked toward the Communications Building for my next class, I stopped in the foyer where two large television screens were broadcasting live coverage. I stood there and watched the second tower collapse. Just two months earlier, I had celebrated the Fourth of July on a pier in Brooklyn with a breathtaking view of the World Trade Center. During my senior year of high school, I had even created a needlepoint of the Twin Towers in art class.</p><p>Growing up in North Jersey, the towers were simply part of the landscape of home. I can still remember coming to the top of the hill heading east on Route 46 in Teterboro, New Jersey, and seeing the New York City skyline open up in front of me, with the Twin Towers standing prominently within it. Whenever I traveled somewhere else and then came back, seeing that skyline in the distance meant I was almost home. The towers were familiar, constant, and something I never imagined would one day be gone. Watching those buildings disappear before my eyes was surreal and almost impossible to comprehend.</p><p>When I arrived at my second class, Communication in Action, the professor announced that it had been canceled. A few classmates and I walked over to the Tower dorms on campus and went to the top-floor common room to get a better view of the city. From there, all we could see was a massive gray cloud hanging over Lower Manhattan. The skyline had disappeared.</p><p>About an hour later, the university announced that all classes were canceled for the remainder of the day. Because I had been dropped off that morning, I had no way to get home on my own. Cellphone service was overwhelmed, and I struggled to reach my parents. When I was finally able to get through to them, they came to William Paterson to pick me up. That was around 5:00 PM EST.</p><p>I remember the drive home being filled mostly with silence. There really was not much to say. We were all trying to process what had just happened in New York City, only about 15 miles from my hometown of Garfield, New Jersey. The weight of the day was almost impossible to put into words. I remember thinking that nothing would ever be the same again. The skyline I had grown up with was gone, and so was the sense of security so many of us had taken for granted. At only 18 years old, I understood that I had just witnessed something that would change our lives forever.</p><h3><strong>The Candles in the Windows</strong></h3><p>That night, on the drive home, I remember looking around not only my own neighborhood, but the surrounding neighborhoods as well, and seeing lighted candles placed in people’s windows. I knew people across the country were doing the same thing as a way to mourn, remember those who had been lost, and show unity after the attacks. Some people used traditional candles while others placed electric candles in their windows, but the message was the same. Those small lights became a visible symbol of grief, solidarity, and hope during a time when so many of us were struggling to process what had happened. Seeing them glowing from house to house across entire neighborhoods is one of those memories from September 11 that has never left me.</p><h3><strong>How Close It Really Was</strong></h3><p>What I learned afterward made the proximity of everything even more unsettling. Federal prosecutors later disclosed that two of the Flight 77 hijackers had used computers at a New Jersey college library in the weeks leading up to September 11, including to check reservations for that day. Although prosecutors did not publicly identify the college, William Paterson later confirmed that investigators had seized several public-access computers from its library shortly after the attacks.</p><p>Within days of September 11, William Paterson also established individual computer logins for current students. At the time, no real explanation was given to us for why the change happened so suddenly. I just remember that almost overnight, we could no longer simply sit down at one of the public computers and use it without identifying ourselves. Years later, when I learned about the computers that had been seized as part of the investigation, that sudden change finally made sense to me.</p><p>The Wayne connection made it hit even closer to home. Some of the Flight 77 hijackers had spent time in the same town where I was attending college. Nawaf al-Hazmi opened a mailbox on Valley Road in Wayne shortly before the attacks, and he and Hani Hanjour, who would ultimately pilot Flight 77 into the Pentagon, rented vehicles from a dealership in Wayne. Even more chilling was learning that the dealership was located less than a mile from William Paterson’s campus. These were not unfamiliar places somewhere else in the country. They were roads and businesses surrounding the campus where I had just begun college. William Paterson eventually became my alma mater, which makes those connections even more personal to me.</p><h3><strong>Returning to New York</strong></h3><p>About a month after the attacks, I returned to New York City with friends. The city I knew felt completely different. The energy and liveliness I had always associated with New York had been replaced by a quiet heaviness.</p><p>As we walked along the Hudson River waterfront in Manhattan, we passed photographs of people who were still missing, posted by families desperately hoping for answers. There were flowers, candles, handwritten notes, and makeshift memorials for those whose deaths had already been confirmed. Seeing their faces was very different from hearing numbers reported on television. They were someone’s parent, child, spouse, sibling, friend, or coworker. Some families were mourning while others were still waiting for news that would never come. The grief was visible everywhere.</p><p>Over the next several years, I frequently traveled into the city by bus. At first, I attended tapings at MTV as an audience member, and later I commuted there as an intern. Every trip reminded me how much New York had changed. As I rode the escalators down into the main concourse of the Port Authority Bus Terminal, members of the New York National Guard stood watch. Bomb-sniffing dogs and their handlers regularly patrolled the terminal. It was a frightening time, and that level of security quickly became part of the new reality of traveling into New York.</p><h3><strong>A Piece of the Skyline I Remember</strong></h3><p>Years later, I purchased Thomas Kinkade’s <em>The Spirit of New York</em>, a painting of the Brooklyn Bridge and New York City skyline. What makes my particular piece so meaningful to me is what is on the back. Kinkade added a pencil sketch of the painting and personally drew the Twin Towers into the skyline. For me, that small detail makes the piece incredibly special. It captures the New York skyline I grew up with and still remember so clearly, before September 11 changed it forever.</p><figure><img alt="Framed Thomas Kinkade painting The Spirit of New York hanging on a light blue wall, depicting the Brooklyn Bridge at sunset with the Manhattan skyline, boats on the water, and the Statue of Liberty in the distance." src="https://cdn-images-1.medium.com/max/1024/1*OnH22ZK8cYLklcB5uG1U-w.png" /><figcaption><em>My purchased copy of Thomas Kinkade’s</em> <strong>The Spirit of New York</strong>, <em>photographed in my home. The painting depicts the Brooklyn Bridge, New York Harbor, and the Manhattan skyline illuminated by the warm light of sunset. Photograph by Carly Kutsup.</em></figcaption></figure><h3><strong>Twenty-Five Years Later</strong></h3><p>There is one place, however, that I have never been able to bring myself to visit: the 9/11 Memorial. I have returned to New York countless times over the years, but going to the site itself has always felt different. This year, 25 years later, I plan to finally go.</p><p>I was 18 years old. My friend and I had talked about visiting the World Trade Center that morning, but because of the accident the day before and my father’s sixth sense that I should not go into the city, I ended up at William Paterson instead. I will never know whether we actually would have made it to the World Trade Center, what time we might have arrived, or where we would have been when the attacks began.</p><p>For most of the past 25 years, I knew those facts without really allowing myself to consider what they meant. Only recently have I started to understand how differently that morning could have unfolded.</p><p>Maybe that is part of why visiting the Memorial has always felt so difficult. It is not only a place where I will remember the people who were lost. It is also a place where I will finally have to acknowledge how unexpectedly my own path changed that morning.</p><p>Those events may very well have saved my life.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f33598abb164" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Is Not Replacing Instructional Designers. It Is Giving Us Time to Actually Design.]]></title>
            <link>https://carlykayinnj.medium.com/ai-is-not-replacing-instructional-designers-it-is-giving-us-time-to-actually-design-90fc76bf8d9f?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/90fc76bf8d9f</guid>
            <category><![CDATA[elearning]]></category>
            <category><![CDATA[educational-technology]]></category>
            <category><![CDATA[higher-education]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[instructional-design]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Sun, 30 Aug 2026 15:51:08 GMT</pubDate>
            <atom:updated>2026-08-30T15:51:08.554Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A female instructional designer works at a laptop while a glowing AI figure transforms stacks of paper-based quiz questions into organized digital course content and workflows, symbolizing AI reducing repetitive instructional design tasks while the designer maintains human oversight." src="https://cdn-images-1.medium.com/max/1024/1*AQD9SsZzEnfQjq3wubuxWQ.png" /></figure><p><em>The real value of AI in instructional design is not having it do our jobs. It is finding better ways to handle the repetitive, technical work that can consume hours of an already full workload.</em></p><p>Recently, I was given the task of adding a large number of chemistry practice questions to multiple Canvas test banks. The questions were already written and included the answer choices, correct answers, and explanations. My responsibility was to get everything into Canvas and make sure each question was configured correctly.</p><p>On the surface, it sounds simple enough.</p><p>Create the question. Copy and paste the answer choices. Select the correct answer. Add the feedback. Adjust the settings. Save it. Repeat.</p><p>Then do it again.</p><p>And again.</p><p>Across multiple test banks.</p><p>Given the amount of daily responsibilities my team and I already manage, I could not help but think there had to be a quicker, easier, and more reliable way to accomplish the same thing. Manually rebuilding hundreds of question components in Canvas simply because the content existed in a Word document did not seem like the best use of anyone’s time.</p><p>So before I started copying and pasting, I took to the Internet.</p><p>That decision eventually led me to AI, but AI was not actually where I started.</p><h3>I Had a Problem Before I Had an AI Solution</h3><p>My first question was simple: <strong>Is there a way to upload these questions directly into a Canvas test bank?</strong></p><p>Through my research, I discovered QTI.</p><p>QTI, or Question and Test Interoperability, is a standard used to package and move assessment questions between compatible applications. 1EdTech describes QTI as a way of making tests and assessment items portable so they can be exchanged between systems rather than manually recreated in each one (1EdTech Consortium, n.d.).</p><p>That was exactly what I needed.</p><p>If I had a properly structured QTI file, I could import the questions into the Canvas item bank instead of manually creating each one.</p><p>Great.</p><p>There was only one problem.</p><p>I did not have a QTI file.</p><p>I had Word documents.</p><p>That is what led me to AI.</p><h3>I Was Not Asking AI to Write the Assessment</h3><p>That distinction is important.</p><p>I was not asking AI to create chemistry questions. I was not asking it to determine what students should know, choose the correct answers, or make instructional decisions for me.</p><p>The content already existed.</p><p>What I needed was a way to transform that content into the technical format Canvas required.</p><p>So I began working with AI to create the QTI file.</p><p>It quickly became clear that the task required a lot more specificity than simply saying, “Turn this Word document into a QTI file.”</p><p>The question titles needed to remain blank. The source question numbers needed to be removed. The A, B, C, and D labels could not appear as part of the visible answer choices. The answer choices needed to shuffle. Each question needed to be worth one point. Chemistry notation, subscripts, superscripts, Greek letters, units, and special characters had to remain intact.</p><p>The explanations also needed to appear specifically under the individual correct answer’s feedback field instead of being placed into general feedback for the entire question.</p><p>Those details mattered.</p><p>And AI did not get all of them right the first time.</p><h3>It Took Several Iterations to Get It Right</h3><p>The first prompt did not produce the final solution.</p><p>Neither did the second.</p><p>I generated a file, imported it into Canvas, reviewed what happened, identified problems, changed the prompt, generated another file, and tested it again.</p><p>At one point, the QTI package appeared to be structured correctly, but Canvas rejected the import entirely.</p><p>The problem ultimately came down to the XML namespace used within the QTI file.</p><p>That sent me back to the prompt again.</p><p>I added more technical requirements, regenerated the package, validated the XML, and tested it in Canvas.</p><p>Eventually, it worked.</p><p>Then I tested it again.</p><p>Getting Canvas to accept the file was only one part of the process. I still needed to make sure the correct number of questions appeared, the question titles were blank, the source numbers were gone, the answer choices did not contain A, B, C, or D labels, the correct answers were identified properly, the choices shuffled, and the explanations appeared exactly where I wanted them.</p><p>Only after all of those checks passed did I consider the process successful.</p><p>That experience reinforced something I think is critical when we talk about AI in instructional design:</p><p><strong>The first output should rarely be treated as the final output.</strong></p><h3>Prompting Is Only Half the Skill</h3><p>There is a tremendous amount of attention right now on learning how to prompt AI.</p><p>Prompting matters. In this case, it mattered a lot.</p><p>By the time I was finished, my prompt contained detailed requirements for QTI structure, Canvas compatibility, scoring, individual answer feedback, XML namespaces, packaging, validation, and quality assurance.</p><p>But the prompt alone was not what made the process work.</p><p>I also had to know what the final product was supposed to do.</p><p>I had to recognize when the output was wrong.</p><p>I had to interpret the Canvas error well enough to troubleshoot it.</p><p>I had to revise the process and test it again.</p><p>That is why validation is such an important part of AI literacy.</p><p>NIST’s <em>Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile</em> emphasizes managing trustworthiness and risk throughout the design, use, and evaluation of generative AI rather than assuming that generated output is automatically dependable (Autio et al., 2024).</p><p>In instructional design, that means knowing enough about the work to question the output.</p><p>Did it preserve everything?</p><p>Did it change something it was not supposed to change?</p><p>Did it put information in the wrong field?</p><p>Does it work in the actual learning management system?</p><p>Can I reproduce the process?</p><p>If I do not know what “correct” looks like, a polished AI response can be more dangerous than a visibly broken one.</p><h3>The Best Part Was Not the QTI File</h3><p>Once I had the correct prompt, I could have stopped.</p><p>The assigned work was getting done. The QTI files worked. I had found a much faster way to populate the test banks.</p><p>But getting there had taken several iterations and a fair amount of troubleshooting.</p><p>There was no reason someone else should have to repeat all of that.</p><p>So I created a <strong>standard operating procedure (SOP)</strong>.</p><p>The SOP documented the process, the QTI requirements, the validation steps, troubleshooting information, and the final prompt. Someone else could now take a properly formatted source document and reproduce the workflow without having to go through the same trial-and-error process I had just completed.</p><p>What started as a solution for one assignment became something my team could reuse.</p><p>The SOP was then shared beyond my immediate team with our entire eLearning department.</p><p>To me, <strong>that</strong> is where the example becomes much more interesting than simply saying, “I used AI to make a QTI file.”</p><p>AI helped me solve one problem.</p><p>Instructional design turned the solution into a repeatable workflow.</p><p>Recent higher education research suggests that this type of use is not unusual. In EDUCAUSE’s 2026 study of AI and work in higher education, 70% of respondents identified automating repetitive processes as one of AI’s most promising opportunities. Among respondents who had recently used AI for work, 35% reported using it to write procedure guides such as SOPs, and 33% had used it to automate repetitive tasks or workflows (Robert, 2026).</p><p>That is almost exactly what happened here.</p><h3>This Is the Kind of AI Use I Want to See More Of</h3><p>A lot of conversation around generative AI in education still focuses on content generation.</p><p>Can it write a quiz?</p><p>Can it create a discussion question?</p><p>Can it build a lesson plan?</p><p>Can it summarize an article?</p><p>Those uses certainly have a place, but some of the most useful ways I use AI as an instructional designer are much less flashy.</p><p>AI can help transform content between formats, troubleshoot HTML, identify inconsistencies, reorganize information, assist with course quality assurance, generate first drafts of alt text, help with technical documentation, compare course components, and reduce repetitive production work.</p><p>Fang and Broussard (2024) describe generative AI tools as “intelligent assistants” that can increase efficiency and scalability in instructional design and course development.</p><p>That description works for me because <strong>assistant</strong> is the important word.</p><p>I do not need AI to replace my judgment.</p><p>I need it to help remove unnecessary work between the problem I am trying to solve and the work that actually requires my expertise.</p><h3>Instructional Designers Spend a Lot of Time Not Designing</h3><p>Instructional design is not just learning theory and course mapping.</p><p>There is an enormous amount of technical and operational work involved in developing and maintaining online courses.</p><p>We work inside learning management systems. We configure assessments. We troubleshoot integrations. We review accessibility. We edit multimedia. We format content. We check navigation. We fix HTML. We help faculty. We test third-party tools. We conduct quality reviews. We troubleshoot things that were working perfectly yesterday and somehow stopped working today.</p><p>All of that work matters.</p><p>But not every part of it requires an instructional designer to manually perform every keystroke.</p><p>That is where I think AI changes the conversation.</p><p>The useful question is not:</p><p><strong>Can AI do instructional design?</strong></p><p>The better question is:</p><p><strong>Which parts of my workflow actually require my expertise, and which parts can technology help me accomplish more efficiently?</strong></p><h3>Human Oversight Becomes More Important, Not Less</h3><p>Using AI effectively does not mean turning over responsibility for the final product.</p><p>The U.S. Department of Education has specifically recommended a “human in the loop” approach to AI in education, emphasizing that educators should remain central to instructional decisions and retain judgment and control over how AI is used (Office of Educational Technology, 2023).</p><p>I think that principle applies directly to instructional design.</p><p>AI might create an assessment file.</p><p>I still need to determine whether it works.</p><p>AI might draft alt text.</p><p>I still need to determine whether that description communicates what the learner actually needs from the image.</p><p>AI might generate HTML.</p><p>I still need to determine whether that HTML is semantically correct and accessible.</p><p>AI might suggest a learning objective.</p><p>I still need to determine whether it aligns with the assessment and course outcomes.</p><p>UNESCO takes a similar position in its guidance on generative AI in education, calling for a human-centered approach that includes ethical validation and pedagogical design while maintaining human agency (Miao &amp; Holmes, 2023).</p><p>AI can assist with the work.</p><p>It cannot assume responsibility for the learning experience.</p><h3>Accessibility Makes That Especially Important</h3><p>Accessibility is one of the clearest examples of why AI should support expertise rather than replace it.</p><p>AI can help draft alt text, restructure information, simplify language, assist with remediation, and identify potential problems.</p><p>That can save time.</p><p>But “AI generated it” is not an accessibility standard.</p><p>An image description can sound polished and still miss the instructional purpose of the image.</p><p>A page can look organized visually while having a poor heading structure.</p><p>An automated tool can say something is accessible without understanding how a student using assistive technology will actually experience it.</p><p>The ability to create or modify content quickly does not remove the need to know whether it was done correctly.</p><p>If anything, the more AI increases the amount of content we can produce, the more important skilled review becomes.</p><h3>AI Literacy Needs to Go Beyond Learning ChatGPT</h3><p>Higher education is already moving toward broader AI use in everyday work.</p><p>EDUCAUSE found that among institutions whose AI strategies include increasing workforce skills, 69% are primarily doing so by upskilling or reskilling existing faculty and staff. The same study found that 73% of respondents who had recently used AI for work were using it either daily or weekly (Robert, 2026).</p><p>That means AI literacy is becoming part of professional practice.</p><p>For instructional designers, though, that literacy needs to go much further than learning how to type a good prompt.</p><p>We need to understand how AI fits into actual workflows.</p><p>Can it reduce repetitive LMS production work?</p><p>Can it support accessibility without replacing accessibility review?</p><p>Can it help with course quality assurance?</p><p>Can it transform content between formats?</p><p>Can it assist with technical troubleshooting?</p><p>What information should never be entered into a particular AI system?</p><p>When does an output require subject-matter review?</p><p>How do we document an AI-assisted workflow so someone else can reproduce it?</p><p>And one question should always remain part of the process:</p><p><strong>How will I know if the AI is wrong?</strong></p><p>That, to me, is real AI literacy.</p><h3>I Want AI to Give Me More Time to Be an Instructional Designer</h3><p>I did not use AI to create those QTI files because I wanted to avoid doing my job.</p><p>I used AI because I wanted to find a better way to do my job.</p><p>There is a difference.</p><p>I do not need to manually copy and paste hundreds of answer choices into Canvas to prove my value as an instructional designer.</p><p>I need to make sure the assessment works.</p><p>I need to make sure the course is accessible.</p><p>I need to make sure students can navigate it.</p><p>I need to make sure assessments align with learning objectives.</p><p>I need to help faculty solve instructional problems.</p><p>I need to catch issues before students encounter them.</p><p>I need time to think about learning.</p><p>If AI can reduce the amount of time my team and I spend on repetitive production tasks while maintaining or improving accuracy, that does not make instructional designers less necessary.</p><p>It gives us more time to do the work for which we are actually needed.</p><p>In my case, one assignment that could have turned into hours of repetitive data entry became something much more useful: a tested AI-assisted workflow, a reusable prompt, and a standard operating procedure that could be shared across an entire eLearning department.</p><p>That is the potential I see in AI.</p><p>Not replacing instructional designers.</p><p><strong>Giving us better tools to identify problems, improve processes, share solutions, and spend more of our time actually designing.</strong></p><h3>References</h3><p>1EdTech Consortium. (n.d.). <em>Question &amp; Test Interoperability (QTI).</em> 1EdTech Consortium.<br><a href="https://www.1edtech.org/standards/qti?utm_source=chatgpt.com">https://www.1edtech.org/standards/qti</a></p><p>Autio, C., Schwartz, R., Dunietz, J., Jain, S., Stanley, M., Tabassi, E., Hall, P., &amp; Roberts, K. (2024). <em>Artificial intelligence risk management framework: Generative artificial intelligence profile</em> (NIST AI 600–1). National Institute of Standards and Technology.<br><a href="https://doi.org/10.6028/NIST.AI.600-1">https://doi.org/10.6028/NIST.AI.600-1</a></p><p>Fang, B., &amp; Broussard, K. (2024, August 7). <em>Augmented course design: Using AI to boost efficiency and expand capacity.</em> EDUCAUSE Review.<br><a href="https://er.educause.edu/articles/2024/8/augmented-course-design-using-ai-to-boost-efficiency-and-expand-capacity">https://er.educause.edu/articles/2024/8/augmented-course-design-using-ai-to-boost-efficiency-and-expand-capacity</a></p><p>Miao, F., &amp; Holmes, W. (2023). <em>Guidance for generative AI in education and research.</em> UNESCO.<br><a href="https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research">https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research</a></p><p>Office of Educational Technology. (2023). <em>Artificial intelligence and the future of teaching and learning: Insights and recommendations.</em> U.S. Department of Education.<br><a href="https://eric.ed.gov/?id=ED631097&amp;utm_source=chatgpt.com">https://eric.ed.gov/?id=ED631097</a></p><p>Robert, J. (2026, January 12). <em>The impact of AI on work in higher education.</em> EDUCAUSE.<br><a href="https://www.educause.edu/research/2026/the-impact-of-ai-on-work-in-higher-education">https://www.educause.edu/research/2026/the-impact-of-ai-on-work-in-higher-education</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=90fc76bf8d9f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Inside Garfield Public Schools’ AI Pilot: When Education Platforms Are Breached — Part 4]]></title>
            <link>https://carlykayinnj.medium.com/part-4-when-education-platforms-are-breached-5e1b82911569?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e1b82911569</guid>
            <category><![CDATA[ai-governance]]></category>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[school-accountability]]></category>
            <category><![CDATA[student-data-privacy]]></category>
            <category><![CDATA[cybersecurity-breach]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Tue, 25 Aug 2026 16:09:19 GMT</pubDate>
            <atom:updated>2026-08-25T16:18:42.026Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Two school technology professionals review a tablet in a computer lab as a red cybersecurity alert appears beside secured server racks." src="https://cdn-images-1.medium.com/max/1024/1*eMYD3AZhK2mIs8JW-PIthw.png" /></figure><p>Cybersecurity concerns surrounding education technology are not hypothetical.</p><p>Some of the largest and most established education technology companies have experienced significant breaches involving student and educator information. PowerSchool, Canvas, Illuminate Education, Pearson, and Blackbaud are among them.</p><p>That does not mean Parthion will experience a breach.</p><p>It does mean that no school district should operate under the assumption that a breach cannot happen simply because security controls are in place.</p><h3>PowerSchool: More Than 60 Million Students</h3><p>In December 2024, an attacker gained unauthorized access to PowerSchool, one of the largest student information systems in K–12 education. Information belonging to more than 60 million students and approximately 10 million teachers was compromised. Depending on the records involved, the stolen data included names, addresses, dates of birth, Social Security numbers, contact information, and certain medical information (Reuters, 2025a; Maine Office of the Attorney General, 2025).</p><p>Federal prosecutors later disclosed that the attackers demanded approximately $2.85 million in Bitcoin. PowerSchool acknowledged paying a ransom after receiving assurances that the stolen information would be deleted (Reuters, 2025a).</p><p>But that did not necessarily end the problem.</p><p>Months later, school districts began receiving additional extortion demands involving information reportedly stolen during the original breach (Reuters, 2025b).</p><p>That is an important point.</p><p>Once student information leaves a protected system, an organization may no longer have meaningful control over where it goes, how many copies exist, or whether it has truly been deleted.</p><h3>Canvas: Nearly 9,000 Schools Affected</h3><p>In April and May 2026, Instructure disclosed a major cybersecurity incident involving Canvas, one of the most widely used learning management systems in education.</p><p>Reuters reported that approximately 6.65 terabytes of information were stolen from nearly 9,000 schools. The information reportedly included student names, email addresses, and private messages involving students, teachers, and staff (Reuters, 2026).</p><p>The incident also disrupted access to Canvas while Instructure worked to contain the attack. The company brought in outside cybersecurity specialists, including CrowdStrike, strengthened security measures, and ultimately discontinued its Free-for-Teacher service after determining that accounts associated with that service played a role in the unauthorized access (Instructure, 2026).</p><p>Instructure also reported reaching an agreement with the unauthorized actor and receiving assurances and evidence that the stolen information had been deleted. Publicly available information I reviewed does not disclose whether money was paid as part of that agreement or, if it was, how much.</p><p>If the documentation does not establish a dollar amount, I am not going to assign one.</p><h3>And These Are Not Isolated Incidents</h3><p>Other education technology providers have faced similar incidents.</p><p>A breach involving Illuminate Education affected more than 10.1 million students, according to the Federal Trade Commission. The FTC later alleged significant deficiencies in the company’s security practices and required Illuminate to establish a comprehensive data-security program, limit data collection and retention, and delete information it no longer needed (Federal Trade Commission, 2025, 2026).</p><p>Pearson experienced an intrusion involving student information and administrator credentials associated with approximately 13,000 school, district, and university customer accounts. The company later agreed to pay $1 million to settle SEC charges concerning its disclosures about the incident (Securities and Exchange Commission, 2021).</p><p>Blackbaud, another technology provider used by educational institutions, experienced a ransomware attack in 2020 and acknowledged paying the attacker. The company later agreed to a $49.5 million multistate settlement related to the breach and its security practices.</p><p>Different companies. Different systems. Different circumstances.</p><p>But the lesson is the same:</p><p>Even large, established technology companies with cybersecurity teams and extensive resources can be breached.</p><h3>What Does This Mean for Garfield?</h3><p>This brings the discussion directly back to Parthion.</p><p>The records provided through my OPRA request document multiple cybersecurity controls associated with Lantern BRP and Parthion, including encryption, multi-factor authentication, vulnerability scanning, annual third-party penetration testing, incident-response planning, business-continuity planning, vendor-risk management, and cybersecurity insurance.</p><p>Those are meaningful safeguards.</p><p>However, the independent SOC 2 documentation Garfield provided was a Type I examination. It evaluated whether controls were appropriately designed at a particular point in time. It did not evaluate whether those controls operated effectively over a period of time.</p><p>That does not mean those controls are ineffective.</p><p>It means the documentation provided does not establish their operating effectiveness over time.</p><p>And that distinction matters when we look at what has happened to other education technology companies.</p><p>For Garfield parents, educators, and taxpayers, I believe several questions deserve clear answers:</p><p>· What specific student information will Parthion receive?</p><p>· Which of those data are actually necessary for the pilot?</p><p>· How long will that information be retained?</p><p>· Who else, including subprocessors, may have access to it?</p><p>· How quickly must Parthion notify Garfield if unauthorized access occurs?</p><p>· What happens if the platform becomes unavailable during a cybersecurity incident?</p><p>· Who pays for forensic investigations, legal expenses, notifications, credit monitoring, identity protection, and other remediation?</p><p>· What cybersecurity insurance exists at both the vendor and district levels?</p><p>· How will affected students and families be protected if information is compromised?</p><p>· And what financial protection does Garfield have if a serious incident occurs when the Pilot Agreement limits Parthion’s contractual liability to just $1,000?</p><p>These questions are not predictions that Parthion will be breached.</p><p>They are questions informed by what has already happened to other major education technology providers.</p><h3>The Question Is Not Whether Risk Exists</h3><p>The lesson from PowerSchool, Canvas, Illuminate, Pearson, and Blackbaud is not that schools should stop using technology.</p><p>Technology is an essential part of modern education.</p><p>The lesson is that cybersecurity planning cannot focus solely on preventing a breach. Districts also need a documented plan for what happens if prevention fails.</p><p>No vendor can guarantee that a cybersecurity incident will never occur.</p><p>So the question for Garfield should not simply be:</p><p><strong><em>“Is Parthion secure?”</em></strong></p><p>It should also be:</p><p><strong><em>What student information are we placing into the system? What evidence do we have about its protections? What happens if those protections fail? And who bears the financial and practical consequences if they do?</em></strong></p><p>Those are not arguments against technology or against Parthion.</p><p>They are questions that should be answered before a breach occurs, not after one.</p><h3>Moving From Risk to Responsibility</h3><p>The experiences of PowerSchool, Canvas, Illuminate, Pearson, and Blackbaud demonstrate why cybersecurity deserves serious consideration whenever student information is entrusted to a technology provider.</p><p>But preventing and preparing for a breach is only one part of responsible implementation.</p><p>There are other questions that cybersecurity controls cannot answer.</p><p>Who makes the final decision when an AI system flags a student as at risk? Can an educator challenge that recommendation? How do families know when AI has influenced a decision involving their child? What happens if the data or algorithm gets it wrong?</p><p>Those questions move this conversation beyond cybersecurity and into the much larger issue of how artificial intelligence should be governed inside a public school system.</p><p>And that may be the most important part of this entire discussion.</p><h3>References</h3><p>Akgun, S., &amp; Greenhow, C. (2021, September 22). Artificial intelligence in education: Addressing ethical challenges in K–12 settings. Frontiers in Artificial Intelligence. <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/">https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/</a></p><p>Baker, R. S., &amp; Hawn, A. (2022, December). Algorithmic bias in education. International Journal of Artificial Intelligence in Education, 32(4), 1052–1092. <a href="https://doi.org/10.1007/s40593-021-00285-9">https://doi.org/10.1007/s40593-021-00285-9</a></p><p>Cowbell Cyber, Inc. (2025). <em>Cowbell Cyber Risk Insurance Policy: Garfield Board of Education</em> [Cyber insurance policy]. Obtained through the New Jersey Open Public Records Act.</p><p>Federal Student Aid. (2026, May 12; updated May 29, 2026). Technology security alert — Ongoing cybersecurity incident involving the Canvas Learning Management System. U.S. Department of Education. <a href="https://fsapartners.ed.gov/knowledge-center/library/electronic-announcements/2026-05-12/technology-security-alert-ongoing-cybersecurity-incident-involving-canvas-learning-management-system-updated-may-29-2026">https://fsapartners.ed.gov/knowledge-center/library/electronic-announcements/2026-05-12/technology-security-alert-ongoing-cybersecurity-incident-involving-canvas-learning-management-system-updated-may-29-2026</a></p><p>Federal Trade Commission. (2026, June). FTC gives final approval to order against Illuminate settling allegations it failed to secure students’ personal data. <a href="https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-gives-final-approval-order-against-illuminate-settling-allegations-it-failed-secure-students">https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-gives-final-approval-order-against-illuminate-settling-allegations-it-failed-secure-students</a></p><p>Garfield Public Schools. (2026a). Pilot Agreement between Garfield Public Schools and Parthion. Obtained through the New Jersey Open Public Records Act.</p><p>Garfield Public Schools. (2026b). Response to Open Public Records Act request regarding the Parthion AI pilot. Obtained through the New Jersey Open Public Records Act.</p><p>Instructure. (2026). Security incident update &amp; FAQs. <a href="https://www.instructure.com/incident_update">https://www.instructure.com/incident_update</a></p><p>Kannan, P. (2024, October 3). How harmful are AI’s biases on diverse student populations? Stanford University. <a href="https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations">https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations</a></p><p>Maine Office of the Attorney General. (2025). Data breach notice: PowerSchool Group LLC. <a href="https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/f977d150-2550-4ddc-8715-5ee5cb11c4bd.html">https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/f977d150-2550-4ddc-8715-5ee5cb11c4bd.html</a></p><p>New Jersey Office of the Attorney General, Division of Consumer Affairs. (2023, October 17). AG Platkin announces $49.5 million multistate settlement with Blackbaud to resolve 2020 data breach. <a href="https://www.njconsumeraffairs.gov/News/Pages/10172023.aspx">https://www.njconsumeraffairs.gov/News/Pages/10172023.aspx</a></p><p>Parthion. (n.d.). Parthion: K–12 IEP software &amp; MTSS platform. <a href="https://parthion.io/">https://parthion.io/</a></p><p>Parthion. (2026, May 28). Why we built Parthion the way we did: A whole-child early warning system that encompasses MTSS and special education. <a href="https://parthion.io/resources/why-we-built-parthion">https://parthion.io/resources/why-we-built-parthion</a></p><p>Prescient Assurance LLC. (2026, June 24). SOC 2 Report for Lantern BRP, LLC: A Type 1 Independent Service Auditor’s Report on Controls Relevant to Security. Obtained through the New Jersey Open Public Records Act.</p><p>Reuters. (2025a, May 7). U.S. school districts facing extortion attempt after hack, software provider says. <a href="https://www.reuters.com/world/us/us-school-districts-facing-extortion-attempt-after-hack-software-provider-says-2025-05-07/">https://www.reuters.com/world/us/us-school-districts-facing-extortion-attempt-after-hack-software-provider-says-2025-05-07/</a></p><p>Reuters. (2025b, May 20). Massachusetts hacker to plead guilty to PowerSchool data breach. <a href="https://www.reuters.com/legal/massachusetts-hacker-plead-guilty-powerschool-data-breach-2025-05-20/">https://www.reuters.com/legal/massachusetts-hacker-plead-guilty-powerschool-data-breach-2025-05-20/</a></p><p>Reuters. (2026, May 11). Developer of education tool Canvas issues apology after hack. <a href="https://www.reuters.com/technology/developer-education-tool-canvas-issues-apology-after-hack-2026-05-11/">https://www.reuters.com/technology/developer-education-tool-canvas-issues-apology-after-hack-2026-05-11/</a></p><p>Securities and Exchange Commission. (2021, August 16). SEC charges Pearson plc for misleading investors about cyber breach. <a href="https://www.sec.gov/newsroom/press-releases/2021-154">https://www.sec.gov/newsroom/press-releases/2021-154</a></p><p>UNESCO. (2021, November). Recommendation on the ethics of artificial intelligence. <a href="https://unesdoc.unesco.org/ark:/48223/pf0000381137">https://unesdoc.unesco.org/ark:/48223/pf0000381137</a></p><p>UNICEF Switzerland and Liechtenstein. (2024, November 20). How AI can have negative impacts on children. <a href="https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children">https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children</a></p><p>U.S. Department of Education, Office of Special Education Programs. (n.d.). IDEA regulations: Confidentiality of information (§§ 300.610–300.627). Individuals with Disabilities Education Act. <a href="https://sites.ed.gov/idea/regs/b/f">https://sites.ed.gov/idea/regs/b/f</a></p><p>U.S. Department of Education, Student Privacy Policy Office. (n.d.). Who is a “school official” under FERPA? Protecting Student Privacy. <a href="https://studentprivacy.ed.gov/faq/who-school-official-under-ferpa">https://studentprivacy.ed.gov/faq/who-school-official-under-ferpa</a></p><p>Vanta. (n.d.). SOC 2 Type I vs. Type II: What’s the difference? <a href="https://www.vanta.com/collection/soc-2/soc-2-type-1-vs-type-2">https://www.vanta.com/collection/soc-2/soc-2-type-1-vs-type-2</a></p><p>Walther, C. (2026, January 20). Why AI disclosure matters at every level. Knowledge at Wharton. <a href="https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/">https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5e1b82911569" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Designing for Everyone: Why Instructional Designers Are Essential to Accessible Online Learning]]></title>
            <link>https://carlykayinnj.medium.com/designing-for-everyone-why-instructional-designers-are-essential-to-accessible-online-learning-aae5e04d13c8?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/aae5e04d13c8</guid>
            <category><![CDATA[elearning]]></category>
            <category><![CDATA[inclusive-design]]></category>
            <category><![CDATA[universal-design]]></category>
            <category><![CDATA[instructional-design]]></category>
            <category><![CDATA[distance-learning]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Mon, 17 Aug 2026 21:43:53 GMT</pubDate>
            <atom:updated>2026-08-17T21:43:53.904Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DPPqQYSLji7_zzj1PUNpIQ.png" /></figure><p><em>Accessibility should not be something we check at the end of course development. It should be part of how we design learning from the very beginning.</em></p><p>As an instructional designer, accessibility has always been an important part of how I approach online course development. My doctoral work, however, has pushed me to look at accessibility through a much broader lens. An accessible course is not simply one that passes an accessibility checker or meets a list of technical requirements. A truly accessible and inclusive course considers how students navigate, process, engage with, and demonstrate their understanding of what they are learning.</p><p>That means looking at the Web Content Accessibility Guidelines (WCAG), Universal Design for Learning (UDL), cognitive load, course navigation, consistent module design, multiple ways of presenting information, purposeful use of educational technology, and alignment among Course Learning Objectives (CLOs), Module Learning Objectives (MLOs), instructional content, learning activities, and assessments. All of these elements contribute to the overall student experience and should be considered throughout the course development process.</p><p>One of the biggest takeaways from my doctoral research has been relatively simple:</p><p><strong><em>We should not wait for students to encounter barriers before we decide to remove them.</em></strong></p><h3>Accessibility Should Start With the Design</h3><p>Accessibility is sometimes treated as one of the final stages of course development. The course is built, content is uploaded, activities are created, and then someone reviews everything for accessibility. I believe we need to reverse that thinking.</p><p>Gronseth (2018) connected the Web Content Accessibility Guidelines with Universal Design for Learning as complementary approaches to inclusive online and blended course design. Considering both during development allows accessibility to become part of instructional design rather than a separate task completed after the course has already been built.</p><p>For example, adding alternative text is only one part of making an image accessible. We should also consider whether the image has an instructional purpose, whether it supports a learning objective, whether important information is communicated through color alone, whether there is sufficient contrast, and whether the alternative text communicates the information students actually need.</p><p>The same approach applies to videos, documents, interactive activities, assessments, publisher materials, and other course resources. Addressing accessibility during development allows many barriers to be identified before students ever encounter them.</p><h3>WCAG and UDL Provide the Foundation</h3><p>The Web Content Accessibility Guidelines (WCAG) 2.2 organize digital accessibility around four principles: <strong>Perceivable, Operable, Understandable, and Robust</strong> (World Wide Web Consortium [W3C], 2024). In online courses, these principles provide a framework for making content and learning activities accessible to students with a wide range of needs. <strong><em>Perceivable </em></strong>content includes meaningful alternative text, accurate captions and transcripts, readable text, and sufficient color contrast. <strong><em>Operable </em></strong>content allows students to navigate and interact with course materials using a keyboard or assistive technology. <strong><em>Understandable </em></strong>content includes clear directions, consistent organization, and predictable navigation, while <strong><em>robust </em></strong>content supports compatibility with assistive technologies and different ways of accessing digital content.</p><p>WCAG Level AA provides an important benchmark for accessibility, but many of these practices also improve usability for all students. Clear headings make information easier to scan and process, captions provide another way to access information presented through audio, descriptive links provide context before students select them, and consistent navigation helps students understand how the course is organized. These design choices can also reduce unnecessary cognitive load by allowing students to focus more of their attention on the course content rather than figuring out how to access or navigate it.</p><p>Universal Design for Learning (UDL) expands on this approach by considering how students access, engage with, and demonstrate their understanding of course content. CAST (2024) explains that no single means of engagement will be optimal for every learner in every situation. Online course design should account for those differences by providing meaningful options for students to interact with the material.</p><p>Depending on the learning objectives, a module might incorporate short videos, readings, charts and graphs, interactive scenarios, knowledge checks, demonstrations, case studies, or simulations. Providing multiple ways to engage with content does not mean creating several versions of the same lesson. It means intentionally selecting instructional methods and resources that support the Course Learning Objectives and Module Learning Objectives while giving students different ways to access and engage with the material.</p><p>Seymour (2024) examined the application of UDL within an online learning environment and described redesign efforts focused on removing barriers to engagement by improving access to learning content. That connection between accessibility and purposeful instructional design is important. The goal is not to add technology, multimedia, or interactive elements simply to make a course more engaging. Each design decision should improve access to learning and support what students are expected to know or be able to do.</p><h3>Navigation and Cognitive Load Matter</h3><p>Students should be able to enter an online course and quickly understand where they need to go. The home page should provide a clear starting point with easy access to modules, assignments, course resources, instructor information, and other frequently used areas. Modules should follow a predictable structure, headings should clearly organize information, and links should describe where they lead.</p><p>I have experienced the importance of this from the student side during my doctoral program. I encountered a course layout that I found difficult to navigate and not necessarily accessible from a usability standpoint. Information I needed to complete assignments was not always where I expected to find it, and I struggled to understand how different parts of the course were organized and connected. I eventually reached out to my instructor for clarification about the course layout, where information was located, and what I needed to reference to successfully complete assignments.</p><p>My instructor provided the clarification I needed, but the experience reinforced something I regularly advocate for as an instructional designer. Students should not have to contact an instructor simply to figure out where information is located or how a course is organized.</p><p>I was not struggling with the academic content. I was spending mental effort trying to figure out the course itself.</p><p>That experience made the relationship between navigation, accessibility, and cognitive load even more apparent to me. When students have to search multiple locations, interpret inconsistent organization, work through large sections of text, or move unnecessarily between platforms, some of their cognitive resources are spent navigating the learning environment rather than processing the content.</p><p>A course can technically meet accessibility requirements and still be cognitively exhausting.</p><p>Reducing unnecessary cognitive load does not mean reducing academic rigor. Students should still analyze difficult concepts, solve problems, apply knowledge, and think critically. The challenge should come from the academic content, not from figuring out how the course works.</p><h3>More Content Does Not Always Mean More Learning</h3><p>The same principle applies to the amount of material placed within a course. More readings, longer videos, additional websites, and supplemental resources do not automatically result in more learning.</p><p>Instructional designers and subject matter experts need to determine what students actually need to achieve the learning objectives. Course Learning Objectives and Module Learning Objectives should help guide those decisions.</p><p>Chunking information into manageable sections, using meaningful headings, highlighting important concepts, breaking lengthy videos into shorter segments, and giving students opportunities to check their understanding can make complex material easier to process without reducing rigor.</p><p>Students do not necessarily need more content. They need content that has a clear purpose.</p><h3>Course Learning Objectives Should Guide the Design</h3><p>Alignment is one area I continually advocate for during course development. Course content, instructional materials, learning activities, educational technologies, and assessments should align with the Course Learning Objectives and the Module Learning Objectives that support them.</p><p>Backward Design is particularly useful because the development process begins with the intended learning outcomes. Assessments, activities, resources, and instructional strategies can then be selected based on those outcomes.</p><p>An interactive activity may look impressive, but I still want to know how it supports the learning objective. The same question should be asked about videos, readings, discussion boards, simulations, publisher resources, and educational technologies.</p><p>For example, if a biology learning objective requires students to analyze physiological processes, an interactive scenario could ask students to review symptoms, interpret laboratory information, evaluate evidence, and explain the biological process involved. A chemistry scenario could provide an unexpected laboratory result and ask students to apply course concepts to determine what occurred.</p><p>The technology has a purpose because the student is applying knowledge that directly supports the intended learning outcome.</p><p>Accessibility and alignment therefore need to be considered together. An activity can be completely accessible and still have little instructional value if it does not help students achieve a learning objective.</p><h3>Inclusive Design Considers the Students Who Are Actually Taking the Course</h3><p>Accessible and inclusive design should not be limited to students with documented disabilities. Lomellini et al. (2025) examined accessible and inclusive online learning in higher education and identified the importance of broader approaches to accessibility within online education.</p><p>Online students bring different experiences, responsibilities, abilities, and educational backgrounds into a course. Some work full time. Some are parents or caregivers. Students may have disabilities, temporary injuries, different levels of technological experience, or language differences. Some use assistive technology, while others may experience barriers they never formally disclose.</p><p>Inclusive design also considers representation. The images, examples, scenarios, case studies, and perspectives included within a course matter. Schirmer and Lockman (2022) emphasized intentional instructor presence and opportunities for students to connect online course content with their own experiences and cultures.</p><p>We cannot anticipate every circumstance a student may experience, but we can avoid creating barriers that do not need to exist.</p><h3>AI Can Also Support Accessibility</h3><p>Artificial intelligence has added another consideration to accessible course design. Much of the discussion in higher education has focused on academic integrity, but AI also has potential as a learning and accessibility support when used appropriately.</p><p>Students may use AI to organize lengthy information, break material into smaller sections, explain difficult concepts in another way, develop practice questions, compare concepts, or organize notes. These uses may help students who experience barriers involving information processing, executive functioning, language, or cognitive load.</p><p>AI should not replace instructors, disability accommodations, instructional materials, or student thinking. Students also need to understand that AI can produce inaccurate information. One approach is to incorporate AI into activities that require students to evaluate its output rather than simply accept it.</p><p>A biology student might evaluate an AI-generated explanation of a biological process and identify inaccuracies. A chemistry student might compare an AI-generated explanation of a reaction with course materials and determine what is accurate, incomplete, or incorrect. Students still have to understand the content, but they are also practicing evaluation, verification, and critical thinking.</p><h3>Why Instructional Designers Matter</h3><p>Bringing all of these considerations together is one reason instructional designers are so important to distance learning.</p><p>Faculty are subject matter experts. They know their disciplines and determine what students need to learn. Instructional designers bring expertise in how to organize that knowledge and develop it into an effective online learning experience.</p><p>Moving a face-to-face course into a learning management system does not automatically create an effective online course. Uploading PowerPoints, readings, recorded lectures, assignments, and exams makes materials available electronically, but availability does not automatically mean accessibility or effective instructional design.</p><p>Instructional designers consider the course as a whole. We examine learning objectives, instructional alignment, assessments, accessibility, navigation, cognitive load, UDL, multimedia, educational technology, consistency, and the student experience. We consider whether students using assistive technology can access activities, whether color is being used appropriately, whether captions and alternative text are meaningful, whether content could be better organized, and whether the technologies being used actually support the learning objectives.</p><p>This role becomes particularly important in asynchronous learning. In a face-to-face classroom, an instructor can recognize confusion and clarify something immediately. An asynchronous student may encounter that same problem while working independently. The organization, instructions, navigation, activities, and resources within the course have to communicate much of what an instructor might otherwise explain in person.</p><p>In that sense, <strong><em>the design of an asynchronous course becomes part of the instruction.</em></strong></p><p>Lomellini et al. (2025) emphasized the importance of broader approaches to accessible and inclusive online learning within higher education. Accessibility cannot depend entirely on individual faculty members to identify and correct every possible issue on their own. Institutions need consistent processes and professional support for accessible course development.</p><p>Instructional designers provide an important part of that support, but our role should be collaborative. Faculty bring expertise in the discipline, while instructional designers bring expertise in learning, accessibility, technology, and course design. Working together allows both to consider what students need to learn, how they will access and engage with the material, how they will practice and demonstrate their knowledge, and whether the entire experience supports the Course Learning Objectives.</p><h3>Designing for Better Online Learning</h3><p>My doctoral research has reinforced something I already believed from my work in instructional design. Accessibility should not be viewed only as compliance. WCAG, UDL, cognitive load, navigation, instructional alignment, inclusive design, and educational technology all contribute to the quality of an online learning experience.</p><p>Automated accessibility tools can identify many technical problems, but they cannot determine whether a course is confusing to navigate, whether a module is overwhelming, whether instructions make sense, whether an activity supports a Course Learning Objective, or whether a technology creates more difficulty than instructional value. Those decisions require professional judgment and collaboration.</p><p>Instructional designers are not simply responsible for making online courses look better. Our work affects accessibility, usability, instructional alignment, consistency, faculty support, and the student learning experience. That is why instructional design is an important component of quality distance education.</p><p>Students should be challenged by the content. They should analyze information, solve problems, apply concepts, think critically, and demonstrate that they have achieved the Course Learning Objectives.</p><p><strong><em>They should not have to struggle with the course design in order to do it.</em></strong></p><h3>References</h3><p>CAST. (2024). <em>Engagement</em>. CAST UDL Guidelines. <a href="https://udlguidelines.cast.org/engagement/">https://udlguidelines.cast.org/engagement/</a></p><p>Gronseth, S. (2018). Inclusive design for online and blended courses: Connecting Web Content Accessibility Guidelines and Universal Design for Learning. <em>Educational Renaissance, 7</em>, 14–22. <a href="https://journals.calstate.edu/er/article/view/2845/2488">https://journals.calstate.edu/er/article/view/2845/2488</a></p><p>Lomellini, A., Lowenthal, P. R., Snelson, C., &amp; Trespalacios, J. H. (2025). Accessible and inclusive online learning in higher education: A review of the literature. <em>Journal of Computing in Higher Education, 37</em>(3), 1306–1329. <a href="https://doi.org/10.1007/s12528-024-09424-2">https://doi.org/10.1007/s12528-024-09424-2</a></p><p>Schirmer, B., &amp; Lockman, A. (2022). Culturally responsive teaching in an undergraduate online general education course. <em>Online Learning, 26</em>(3), 132–148. <a href="https://doi.org/10.24059/olj.v26i3.2805">https://doi.org/10.24059/olj.v26i3.2805</a></p><p>Seymour, M. (2024). Enhancing the online student experience through the application of Universal Design for Learning (UDL) to research methods learning and teaching. <em>Education and Information Technologies, 29</em>, 2767–2785. <a href="https://doi.org/10.1007/s10639-023-11948-6">https://doi.org/10.1007/s10639-023-11948-6</a></p><p>World Wide Web Consortium. (2024). <em>Web Content Accessibility Guidelines (WCAG) 2.2</em>. <a href="https://www.w3.org/TR/WCAG22/">https://www.w3.org/TR/WCAG22/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aae5e04d13c8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Inside Garfield Public Schools’ AI Pilot: What the Public Records Reveal — Part 3]]></title>
            <link>https://carlykayinnj.medium.com/part-3-inside-the-pilot-agreement-64a46bd5fd96?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/64a46bd5fd96</guid>
            <category><![CDATA[school-technology]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[vendor-liability]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <category><![CDATA[student-data-privacy]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Mon, 10 Aug 2026 20:53:51 GMT</pubDate>
            <atom:updated>2026-08-10T20:57:34.262Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Conceptual image of a public school behind a desk displaying a laptop with an AI network, a contract folder, a security lock, and student silhouettes connected to digital education and data-security symbols, representing the intersection of AI, student data, cybersecurity, and contractual responsibility in a school technology pilot." src="https://cdn-images-1.medium.com/max/1024/1*jJYbue90REo4RIAq9Jno9g.png" /></figure><p>The Pilot Agreement answers one of the biggest questions left by the OPRA response.</p><p>It shows what Garfield Public Schools agreed to when it began piloting Parthion.</p><p>Like many software agreements, the document contains legal language that most people never read. Yet several of its provisions deserve closer attention because they help explain how the pilot operates, what it may cost taxpayers, and where responsibility ultimately rests if problems arise.</p><h3>A Pilot That Doesn’t Automatically End</h3><p>Many people hear the word <em>pilot</em> and assume it is simply a trial period that ends unless the parties decide to move forward.</p><p>That is not how this agreement is structured.</p><p>According to the Pilot Agreement, unless either Garfield Public Schools or Parthion provides written notice terminating the agreement before the pilot ends, the pilot automatically converts into a paid annual subscription.</p><p>In other words, continuing the service does not require signing a new agreement.</p><p>Instead, the agreement states that it continues unless one of the parties actively chooses to end it.</p><p>That distinction may seem minor, but it changes the default outcome.</p><p>Rather than requiring an affirmative decision to purchase the platform, the agreement requires an affirmative decision to discontinue it.</p><p>For taxpayers, that makes the district’s evaluation process especially important. If a pilot is designed to transition into a paid service by default, it becomes even more important to understand how success will be measured before additional public funds are committed.</p><h3>What Will It Cost?</h3><p>The agreement states that once the pilot converts to a paid subscription, Garfield Public Schools will pay <strong>$5 per enrolled student each year, for the first year</strong>. <strong>After the first year, the pricing will then become $10 per enrolled student each year.</strong></p><p>The total annual cost will therefore depend on district enrollment.</p><p>Although $5 per student may not initially sound significant, districtwide software subscriptions can represent substantial recurring expenditures over time. Like any ongoing investment of public funds, taxpayers reasonably expect that measurable benefits will justify the continuing cost.</p><p>One question naturally follows.</p><p><strong><em>How will the district determine whether the pilot has been successful enough to justify that investment?</em></strong></p><p>The Pilot Agreement establishes the subscription price.</p><p>It does not identify the criteria Garfield Public Schools will use to determine whether the pilot achieved its intended goals before transitioning into a paid service.</p><p>That information may exist elsewhere.</p><p>It simply was not included in the records released in response to my OPRA request.</p><h3>Privacy, Security, and Vendor Due Diligence</h3><p>Artificial intelligence platforms used in education can process sensitive student information, making cybersecurity and data privacy important considerations alongside the platform’s functionality.</p><p>The Pilot Agreement states that Parthion will maintain “commercially reasonable” administrative, technical, and physical safeguards to protect District Data, comply with the Family Educational Rights and Privacy Act (FERPA) and other applicable privacy laws, and notify Garfield if unauthorized access to District Data is confirmed. The agreement itself, however, provides limited technical detail about how those safeguards are implemented.</p><p>FERPA is the federal law that protects the privacy of student education records, including personally identifiable information contained within those records. This becomes particularly important when student information is shared with a third-party provider. Under FERPA’s “school official” exception, districts may disclose certain education records to outside contractors without individual parental consent, but that access is subject to specific requirements and limitations (U.S. Department of Education, n.d.).</p><p>According to the U.S. Department of Education, a third-party provider must perform a service or function for which the district would otherwise use its own employees, meet the district’s criteria for a legitimate educational interest, and remain under the district’s direct control regarding the use and maintenance of education records. The provider is also subject to restrictions on the use and redisclosure of personally identifiable information (U.S. Department of Education, n.d.).</p><p>FERPA is not the only federal law relevant to this discussion. The Individuals with Disabilities Education Act (IDEA) adds another layer of protection when information involves students receiving special education services.</p><p>IDEA includes confidentiality protections for personally identifiable information collected, maintained, or used under the law. It also provides parents with important rights to participate in decisions involving their child’s identification, evaluation, educational placement, and provision of a free appropriate public education (FAPE) (U.S. Department of Education, n.d.).</p><p>That is particularly relevant to Parthion because the Pilot Agreement specifically references interventions, special education determinations, IEP decisions, and eligibility decisions. If Parthion is used to support these processes, the issue extends beyond how student information is protected. It also raises questions about how AI-generated information will be used within decision-making processes that carry specific legal protections for students with disabilities and their families.</p><p>For Garfield, that raises practical questions that go beyond simply stating that Parthion will comply with FERPA. These questions should include:</p><p>· What specific student records will Parthion receive?</p><p>· Who will be permitted to access them?</p><p>· What constitutes a legitimate educational interest for that access?</p><p>· How will Garfield maintain direct control over the use and maintenance of those records?</p><p>· What third-party companies, if any, may process or have access to student information through Parthion, and what restrictions govern that access?</p><p>· And what happens to student information when the pilot or the district’s relationship with Parthion ends?</p><p>For students receiving special education services, there are additional questions. Those questions include:</p><p>· How will AI-generated recommendations be used in decisions involving interventions, evaluations, eligibility, or IEPs?</p><p>· What safeguards will ensure that those recommendations support rather than replace individualized evaluation and professional judgment?</p><p>· And how will Garfield ensure that parents continue to have the meaningful participation in special education decision-making that IDEA requires?</p><p>Those questions are closely connected to the broader issue of vendor due diligence.</p><p>Parthion’s public security information provided additional details about its security practices and identified documentation that districts could request. This included a security whitepaper, information concerning its SOC 2 progress, a subprocessor list, and a draft Data Processing Agreement.</p><p>That raised an important question when I initially reviewed the records: <strong><em>What security and privacy documentation did Garfield request and review before approving the pilot?</em></strong></p><p>My OPRA request specifically sought records concerning the district’s cybersecurity and privacy due diligence, including Parthion’s SOC 2 Type II status, documentation concerning its reported “in progress” status, risks associated with implementation before completion of the Type II audit, and alternative security documentation reviewed by the district.</p><p>The records initially provided did not answer all those questions. In particular, the documentation available at that point did not establish completion of a SOC 2 Type II examination or provide enough information to determine the full extent of the security documentation Garfield had reviewed.</p><p>That has now changed.</p><p>On August 4, Garfield provided additional records in response to my OPRA request, including independent security documentation that was not part of the records I initially received. Those records provide important new information about Lantern BRP’s cybersecurity controls and the level of independent assurance available for those controls.</p><p>They also require an important distinction between SOC 2 Type I and SOC 2 Type II.</p><h3>New Records Provide More Information</h3><p>The district’s August 4, 2026 response provides additional information that was not included in the records I initially received. In response to my request for records concerning Parthion’s SOC 2 Type II status and alternative security documentation, Garfield Public Schools provided a redacted SOC 2 report for Lantern BRP, LLC, whose system description includes the LanternBRP platform and Parthion.</p><p>Prepared by Prescient Assurance LLC and dated June 24, 2026, the report examines Lantern BRP’s security controls as of May 25, 2026. The independent auditor concluded that Lantern BRP’s system description was fairly presented and that the controls described were suitably designed to provide reasonable assurance that the company’s stated security commitments and system requirements would be achieved.</p><p>The documentation describes a substantial cybersecurity control framework, including encryption of service data at rest and in transit, multi factor authentication for non-console access to production infrastructure, least privilege access, periodic access reviews, termination controls, security monitoring and logging, vulnerability scanning, annual third party penetration testing, secure software development practices, incident response procedures, business continuity and disaster recovery planning, vendor risk management, and cybersecurity insurance.</p><p>Lantern management states that it was not aware of significant security incidents as of May 25, 2026, that would materially affect the system covered by the examination. Management was likewise not aware of subsequent incidents through June 24 that would significantly affect its assertions or system description. This should be interpreted narrowly. It reflects management’s disclosure concerning incidents covered by the examination, not an independent finding that the company has never experienced a cybersecurity incident.</p><h3>Why Type I and Type II Are Not the Same</h3><p>Before looking more closely at the report Garfield provided, it is important to understand the difference between a SOC 2 Type I and Type II examination.</p><p>In simple terms, a Type I examination looks at whether an organization has appropriately designed security controls in place at a specific point in time. A Type II examination goes a step further by evaluating whether those controls actually operated effectively over a period of time (Vanta, n.d.).</p><p>Think of it as the difference between confirming that a security system has the right safeguards in place and evaluating whether those safeguards consistently worked as intended over time.</p><p>That distinction matters when student information is involved. A Type I report can provide valuable independent assurance about how security controls are designed, but it does not provide the same evidence about how those controls operated over time that a Type II examination does.</p><p>The most important distinction in the records provided to me is that Garfield supplied a SOC 2 Type I report, not a SOC 2 Type II report.</p><p>For Lantern BRP, the Type I examination evaluated its system description and the suitability of the design of its controls as of May 25, 2026. Operating effectiveness over a period of time was not part of this examination (Prescient Assurance LLC, 2026).</p><p>Prescient Assurance states that limitation directly:</p><p><strong><em>“We did not perform any procedures regarding the operating effectiveness of controls stated in the description and, accordingly, do not express an opinion thereon” (Prescient Assurance LLC, 2026).</em></strong></p><p>That limitation is important. The Type I examination provides independent evidence concerning Lantern BRP’s system description and the suitability of its control design as of May 25, 2026. It does not establish that those controls operated effectively over time.</p><p>This is particularly relevant because my OPRA request specifically sought records concerning Parthion’s SOC 2 Type II certification status, including records concerning its reported “in progress” status, risks associated with implementation before completion of the Type II audit, and alternative security documentation reviewed in place of a completed Type II report. Garfield supplied this Type I report in response to that request (Garfield Public Schools, 2026b).</p><p>Based on the records provided to me, Garfield has provided evidence of SOC 2 Type I completion. The records do not establish SOC 2 Type II completion.</p><p>This does not establish that Lantern or Parthion has never completed a Type II examination. It means the records Garfield provided to me do not show that it has.</p><h3>Why the Distinction Matters for Students and the Community</h3><p>For students, the distinction matters because the assurance provided by the Type I examination has a clear limit. Prescient Assurance evaluated the design of the controls as of May 25, 2026, but did not evaluate their operating effectiveness.</p><p>As a result, this document alone cannot establish that the controls described in it, including access controls, security monitoring, vulnerability management, incident response, and other safeguards, operated effectively over time.</p><p>This does not mean those controls failed, nor does it mean student information was compromised. Neither conclusion appears in the documentation. It means operating effectiveness was outside the scope of this particular examination.</p><p>For the Garfield community, understanding that limitation is important when considering the level of cybersecurity assurance available for the system. The Type I examination provides meaningful information about Lantern BRP’s security controls, but it does not answer the question of how effectively those controls operated over a defined period.</p><p>A Type II examination would address that additional question by evaluating operating effectiveness over a period of time. It would not eliminate cybersecurity risk, but it would provide a level of evidence that this Type I examination does not.</p><h3>Additional Limitations in the Report</h3><p>Some of the security responsibilities described in the documentation extend beyond Lantern BRP.</p><p>Lantern uses a subservice organization for cloud hosting, and the examination uses the carve out method. As a result, the auditor did not evaluate the suitability of the design of the controls operated by that subservice organization as part of this examination.</p><p>Customers using Lantern’s services have responsibilities as well. The documentation identifies Complementary User Entity Controls, and the auditor’s opinion assumes that applicable controls at user entities are suitably designed and implemented.</p><p>In other words, the SOC 2 examination does not cover every control involved in the overall security picture. Certain responsibilities remain with other organizations and with the entities using Lantern’s services.</p><p>The version released through OPRA contains extensive redactions in portions of the technical system description. Garfield states that information was withheld to protect confidential administrative or technical information and proprietary commercial or financial information.</p><p>Whatever the basis for those redactions, the practical result is that the public version does not contain all of the technical information found in the underlying document.</p><h3>The $1,000 Liability Provision</h3><p>Separate from the SOC 2 documentation is the agreement between Garfield Public Schools and Parthion. One provision stands out: Parthion’s contractual liability is limited to $1,000.</p><p>These documents address different issues. The SOC 2 examination concerns Lantern BRP’s security controls, while the Parthion agreement governs the contractual relationship between the vendor and Garfield.</p><p>Under the agreement, Parthion disclaims responsibility for certain damages, including lost data, lost revenue, business disruption, and other indirect or consequential damages. It further states that Parthion is not responsible for educational decisions made by Garfield personnel, including interventions, special education determinations, IEP decisions, eligibility decisions, disciplinary actions, or other decisions made in reliance on the software.</p><p>The $1,000 limitation raises an important question about the protections available to Garfield if a significant problem occurs. The provision itself does not establish who would ultimately be responsible for every possible cost or legal claim.</p><p>For parents and taxpayers, however, the question is worth asking: <strong><em>If a significant cybersecurity or system-related incident occurred, what protections are in place for the district, students, families, and taxpayers beyond those provided in the contract?</em></strong></p><p>Nothing in the records establishes that such an incident has occurred or will occur. The question concerns how the district has prepared for that possibility and how risk is allocated if it does.</p><h3>“AS IS” and “AS AVAILABLE”</h3><p>Another provision states that the platform is provided <strong><em>“AS IS”</em></strong> and <strong><em>“AS AVAILABLE.”</em></strong></p><p>These phrases appear in many software agreements.</p><p>They generally mean that the software is provided without broad warranties guaranteeing uninterrupted service or guaranteeing that the software will meet every expectation of the customer.</p><p>Again, that language is not unique to Parthion.</p><p>It is common throughout the software industry.</p><p>For school districts, however, those provisions reinforce the importance of conducting careful evaluations before relying on any technology that supports educational operations.</p><p>No software platform is perfect.</p><p>Artificial intelligence systems are no exception.</p><p>Researchers have emphasized that AI should be viewed as a tool that assists human professionals rather than one that replaces professional judgment (Akgun &amp; Greenhow, 2021). That recommendation becomes even more important when software agreements themselves acknowledge that technology has limitations.</p><p>Taken together, the $1,000 liability limitation and the “AS IS” and “AS AVAILABLE” provisions make another part of the records particularly important: <strong><em>What financial protection does Garfield have if a serious cybersecurity or system-related incident results in substantial costs?</em></strong></p><h3>Garfield’s Own Cyber Insurance</h3><p>The records Garfield provided through OPRA offer at least part of that picture.</p><p>When I requested records concerning cyber insurance applicable to Parthion, vendor indemnification, liability limitations, financial responsibility following a breach, and related contractual obligations, the district provided its own cyber-risk insurance policy. The Garfield Board of Education is the named insured under that policy, which carries a <strong><em>$2 million aggregate limit of liability </em></strong>(Cowbell Cyber, Inc., 2025).</p><p>The policy includes several categories of cyber coverage, including data restoration, extortion costs, business interruption, liability costs, and regulatory costs, although the specific limits, deductibles, and sublimits vary by coverage.</p><p>There is an important distinction here: <strong><em>this is Garfield’s insurance policy, not Parthion’s.</em></strong></p><p>In response to my request for records concerning cyber insurance applicable to Parthion and financial responsibility following a breach, Garfield provided the Board’s own policy and stated that there were <strong><em>“no other records responsive to this request.”</em></strong></p><p>Based on the records provided to me, Garfield therefore has documented cyber-risk insurance with a $2 million aggregate limit. The records provided do not establish comparable cyber-insurance coverage for Parthion.</p><p>That makes the relationship between these two numbers difficult to overlook: <strong><em>Parthion’s contractual liability to Garfield is capped at $1,000, while Garfield itself carries a cyber-risk policy with a $2 million aggregate limit.</em></strong></p><p>That does not mean Garfield would automatically be responsible for every cost arising from a cybersecurity incident, nor does it mean the full $2 million would necessarily be available for any particular claim. Insurance coverage depends on the circumstances of an incident and the terms, exclusions, deductibles, and sublimits of the policy.</p><p>But it does raise an important question for the community: <strong><em>If a serious incident involving Parthion resulted in substantial financial losses, how much of that risk would ultimately remain with the district?</em></strong></p><p>That question, along with the limits of the SOC 2 Type I examination, brings the larger picture into focus.</p><h3>What the Records Establish</h3><p>Taken together, the new records provide a clearer picture of both the cybersecurity documentation supplied to Garfield and the contractual protections contained in its agreement with Parthion. They also provide important information about how financial risk may be allocated, including Parthion’s $1,000 contractual liability limitation and Garfield’s own $2 million aggregate cyber-risk insurance coverage.</p><p>Garfield provided a SOC 2 Type I report for Lantern BRP. The examination addresses Lantern BRP’s system description and the suitability of the design of its controls as of May 25, 2026. Prescient Assurance expressly states that it did not evaluate the operating effectiveness of those controls. The records provided to me do not establish completion of a SOC 2 Type II examination.</p><p>At the same time, Lantern BRP has documented numerous cybersecurity controls, including encryption, multi-factor authentication, vulnerability scanning, annual third-party penetration testing, incident response planning, business continuity and disaster recovery planning, vendor risk management, and cybersecurity insurance.</p><p>The issue raised by these records is not whether security controls are documented. The SOC 2 report documents numerous controls. The remaining distinction is the level of assurance provided by a Type I examination compared with the Type II documentation requested through OPRA.</p><p>A Type I examination addresses the design of controls at a specific point in time. This examination does not establish their operating effectiveness over time. That distinction is especially relevant because the records I requested concerned SOC 2 Type II status, while the SOC 2 documentation Garfield provided is Type I.</p><h3>Technology Is Only Part of the Story</h3><p>After reviewing the Pilot Agreement, one conclusion became clear.</p><p>The agreement explains <strong><em>how the software is licensed.</em></strong></p><p>It does not explain <strong><em>how Garfield Public Schools evaluated the educational, ethical, or operational implications of using the technology.</em></strong></p><p>Contracts define legal relationships.</p><p>They do not typically describe governance, implementation strategies, staff training, evaluation metrics, or oversight procedures.</p><p>Those questions are usually addressed through district policies, administrative procedures, implementation plans, or other supporting documentation.</p><p>Whether those records exist is a separate question.</p><p>The Pilot Agreement simply does not answer it.</p><h3>Looking Beyond the Contract</h3><p>The agreement provides valuable insight into the business relationship between Garfield Public Schools and Parthion.</p><p>It explains the pilot structure, establishes pricing, allocates contractual risk, and outlines how the subscription continues after the pilot period.</p><p>Those are important facts.</p><p>But cybersecurity safeguards are only meaningful if we also consider what happens when they fail.</p><p>No security framework can eliminate risk entirely. And when student information is involved, understanding the consequences of a breach matters just as much as understanding the protections designed to prevent one.</p><p>That is especially important here because the Pilot Agreement limits Parthion’s contractual liability to $1,000.</p><p><strong><em>So what can happen when an education technology platform is breached despite having cybersecurity protections in place?</em></strong></p><p>We do not have to speculate.</p><p>It has already happened, including to some of the largest education technology platforms in the country.</p><p>What happened next provides important context for the questions Garfield parents, educators, and taxpayers should be asking now.</p><h3>References</h3><p>Akgun, S., &amp; Greenhow, C. (2021, September 22). Artificial intelligence in education: Addressing ethical challenges in K–12 settings. Frontiers in Artificial Intelligence. <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/">https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/</a></p><p>Baker, R. S., &amp; Hawn, A. (2022, December). Algorithmic bias in education. International Journal of Artificial Intelligence in Education, 32(4), 1052–1092. <a href="https://doi.org/10.1007/s40593-021-00285-9">https://doi.org/10.1007/s40593-021-00285-9</a></p><p>Cowbell Cyber, Inc. (2025). <em>Cowbell Cyber Risk Insurance Policy: Garfield Board of Education</em> [Cyber insurance policy]. Obtained through the New Jersey Open Public Records Act.</p><p>Federal Student Aid. (2026, May 12; updated May 29, 2026). Technology security alert — Ongoing cybersecurity incident involving the Canvas Learning Management System. U.S. Department of Education. <a href="https://fsapartners.ed.gov/knowledge-center/library/electronic-announcements/2026-05-12/technology-security-alert-ongoing-cybersecurity-incident-involving-canvas-learning-management-system-updated-may-29-2026">https://fsapartners.ed.gov/knowledge-center/library/electronic-announcements/2026-05-12/technology-security-alert-ongoing-cybersecurity-incident-involving-canvas-learning-management-system-updated-may-29-2026</a></p><p>Federal Trade Commission. (2026, June). FTC gives final approval to order against Illuminate settling allegations it failed to secure students’ personal data. <a href="https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-gives-final-approval-order-against-illuminate-settling-allegations-it-failed-secure-students">https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-gives-final-approval-order-against-illuminate-settling-allegations-it-failed-secure-students</a></p><p>Garfield Public Schools. (2026a). Pilot Agreement between Garfield Public Schools and Parthion. Obtained through the New Jersey Open Public Records Act.</p><p>Garfield Public Schools. (2026b). Response to Open Public Records Act request regarding the Parthion AI pilot. Obtained through the New Jersey Open Public Records Act.</p><p>Instructure. (2026). Security incident update &amp; FAQs. <a href="https://www.instructure.com/incident_update">https://www.instructure.com/incident_update</a></p><p>Kannan, P. (2024, October 3). How harmful are AI’s biases on diverse student populations? Stanford University. <a href="https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations">https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations</a></p><p>Maine Office of the Attorney General. (2025). Data breach notice: PowerSchool Group LLC. <a href="https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/f977d150-2550-4ddc-8715-5ee5cb11c4bd.html">https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/f977d150-2550-4ddc-8715-5ee5cb11c4bd.html</a></p><p>New Jersey Office of the Attorney General, Division of Consumer Affairs. (2023, October 17). AG Platkin announces $49.5 million multistate settlement with Blackbaud to resolve 2020 data breach. <a href="https://www.njconsumeraffairs.gov/News/Pages/10172023.aspx">https://www.njconsumeraffairs.gov/News/Pages/10172023.aspx</a></p><p>Parthion. (n.d.). Parthion: K–12 IEP software &amp; MTSS platform. <a href="https://parthion.io/">https://parthion.io/</a></p><p>Parthion. (2026, May 28). Why we built Parthion the way we did: A whole-child early warning system that encompasses MTSS and special education. <a href="https://parthion.io/resources/why-we-built-parthion">https://parthion.io/resources/why-we-built-parthion</a></p><p>Prescient Assurance LLC. (2026, June 24). SOC 2 Report for Lantern BRP, LLC: A Type 1 Independent Service Auditor’s Report on Controls Relevant to Security. Obtained through the New Jersey Open Public Records Act.</p><p>Reuters. (2025a, May 7). U.S. school districts facing extortion attempt after hack, software provider says. <a href="https://www.reuters.com/world/us/us-school-districts-facing-extortion-attempt-after-hack-software-provider-says-2025-05-07/">https://www.reuters.com/world/us/us-school-districts-facing-extortion-attempt-after-hack-software-provider-says-2025-05-07/</a></p><p>Reuters. (2025b, May 20). Massachusetts hacker to plead guilty to PowerSchool data breach. <a href="https://www.reuters.com/legal/massachusetts-hacker-plead-guilty-powerschool-data-breach-2025-05-20/">https://www.reuters.com/legal/massachusetts-hacker-plead-guilty-powerschool-data-breach-2025-05-20/</a></p><p>Reuters. (2026, May 11). Developer of education tool Canvas issues apology after hack. <a href="https://www.reuters.com/technology/developer-education-tool-canvas-issues-apology-after-hack-2026-05-11/">https://www.reuters.com/technology/developer-education-tool-canvas-issues-apology-after-hack-2026-05-11/</a></p><p>Securities and Exchange Commission. (2021, August 16). SEC charges Pearson plc for misleading investors about cyber breach. <a href="https://www.sec.gov/newsroom/press-releases/2021-154">https://www.sec.gov/newsroom/press-releases/2021-154</a></p><p>UNESCO. (2021, November). Recommendation on the ethics of artificial intelligence. <a href="https://unesdoc.unesco.org/ark:/48223/pf0000381137">https://unesdoc.unesco.org/ark:/48223/pf0000381137</a></p><p>UNICEF Switzerland and Liechtenstein. (2024, November 20). How AI can have negative impacts on children. <a href="https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children">https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children</a></p><p>U.S. Department of Education, Office of Special Education Programs. (n.d.). IDEA regulations: Confidentiality of information (§§ 300.610–300.627). Individuals with Disabilities Education Act. <a href="https://sites.ed.gov/idea/regs/b/f">https://sites.ed.gov/idea/regs/b/f</a></p><p>U.S. Department of Education, Student Privacy Policy Office. (n.d.). Who is a “school official” under FERPA? Protecting Student Privacy. <a href="https://studentprivacy.ed.gov/faq/who-school-official-under-ferpa">https://studentprivacy.ed.gov/faq/who-school-official-under-ferpa</a></p><p>Vanta. (n.d.). SOC 2 Type I vs. Type II: What’s the difference? <a href="https://www.vanta.com/collection/soc-2/soc-2-type-1-vs-type-2">https://www.vanta.com/collection/soc-2/soc-2-type-1-vs-type-2</a></p><p>Walther, C. (2026, January 20). Why AI disclosure matters at every level. Knowledge at Wharton. <a href="https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/">https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=64a46bd5fd96" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Inside Garfield Public Schools’ AI Pilot: What the Public Records Reveal — Part 2]]></title>
            <link>https://carlykayinnj.medium.com/inside-garfield-public-schools-ai-pilot-what-the-public-records-reveal-part-2-1b79305fcb0f?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b79305fcb0f</guid>
            <category><![CDATA[public-records]]></category>
            <category><![CDATA[ai-governance]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <category><![CDATA[data-privacy]]></category>
            <category><![CDATA[public-transparency]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Mon, 10 Aug 2026 00:52:07 GMT</pubDate>
            <atom:updated>2026-08-10T00:52:07.409Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Collage illustrating public records and AI oversight in education, featuring government documents, a school building, cybersecurity and privacy symbols, a records search, families reviewing information, a public meeting, and scales representing accountability and public trust." src="https://cdn-images-1.medium.com/max/1024/1*bhPPV5u7DkTVLV5ICi8LHA.png" /></figure><h3>Part 2: What Public Records Actually Show</h3><p>When I submitted my OPRA request, I wasn’t asking Garfield Public Schools to answer questions or defend its decision to pilot Parthion.</p><p>I was asking for records.</p><p>Specifically, I requested existing documents that could help explain how the district evaluated the platform before introducing it into its schools. My requests covered topics including procurement, cybersecurity, student privacy, accessibility, governance, human oversight, and how the district intended to evaluate the success of the pilot before it automatically converted into a paid subscription.</p><p>The district did produce some records.</p><p>Among them was the Pilot Agreement, which provides important details about the scope of the pilot, the software being evaluated, the subscription model, and several contractual provisions that will be discussed later in this series.</p><p>For several other requests, however, the district responded that there were <strong>“no responsive records.”</strong> Three requests were also denied as <strong>overly broad</strong> under New Jersey’s Open Public Records Act.</p><p>Those responses became just as important as the records themselves.</p><h3>What Does “No Responsive Records” Mean?</h3><p>The phrase <strong>“no responsive records”</strong> is often misunderstood.</p><p>It does <strong>not</strong> necessarily mean that Garfield Public Schools failed to evaluate cybersecurity, privacy, accessibility, or any of the other topics included in my request.</p><p>It also does <strong>not</strong> confirm that those evaluations occurred.</p><p>It simply means that the district did not produce records responsive to those requests.</p><p>That distinction is important.</p><p>Public records exist to help citizens understand how government decisions are made. When residents ask for records, the documents produced become the evidence available for public review.</p><p>In this case, the records released answered some questions.</p><p>They left others unanswered.</p><p>For example, the Pilot Agreement explains <strong>what</strong> Garfield Public Schools is piloting.</p><p>The records released do not explain <strong>why</strong> Parthion was selected over other platforms, what evaluation criteria were used before the pilot began, or how the district intends to measure success before the agreement automatically converts into a paid subscription.</p><p>Those questions may ultimately have answers.</p><p>The point is that they were <strong><em>not answered by the records released in response to this OPRA request.</em></strong></p><p>For parents and taxpayers, that distinction matters.</p><p>Transparency is not simply about announcing that new technology is being used.</p><p>It is about providing enough information for the public to understand how important decisions were made.</p><p>As AI becomes more common in schools, researchers have increasingly emphasized that transparency and accountability are essential for building public trust in these technologies (Akgun &amp; Greenhow, 2021; Walther, 2026).</p><h3>What Does “Overly Broad” Mean?</h3><p>Three portions of my OPRA request were denied because they were considered <strong>overly broad</strong>.</p><p>That phrase also deserves some explanation.</p><p>Under New Jersey’s Open Public Records Act, agencies are required to provide access to identifiable government records that already exist.</p><p>They are not required to answer questions, conduct research, or create documents in response to a request.</p><p>Sometimes a request is considered overly broad because it is not specific enough for the records custodian to determine exactly which records are being requested.</p><p>That does <strong>not</strong> necessarily mean the requested records are exempt from disclosure.</p><p>It often means the request needs to be narrowed or rewritten.</p><p>That is exactly what I did.</p><p>Rather than assuming the records did not exist, I revised those requests using more specific language and submitted them again.</p><p>Whether additional records exist remains to be determined.</p><h3>Why These Records Matter</h3><p>Some readers may wonder why I requested records related to privacy, cybersecurity, governance, accessibility, human oversight, and algorithmic bias in the first place.</p><p>The answer is simple.</p><p>Those topics have become central to the national conversation surrounding artificial intelligence in education.</p><p>Researchers have cautioned that AI systems should not be evaluated solely on whether they function as intended. They should also be evaluated on whether they operate fairly, whether appropriate human oversight exists, and whether safeguards are in place to reduce unintended bias (Akgun &amp; Greenhow, 2021; Baker &amp; Hawn, 2022).</p><p>UNESCO similarly recommends that organizations implementing AI establish governance structures that promote transparency, accountability, privacy, and meaningful human oversight (UNESCO, 2021).</p><p>More recently, Stanford researchers have highlighted concerns that algorithmic bias may disproportionately affect students from historically underserved populations if AI systems are implemented without careful evaluation and ongoing monitoring (Kannan, 2024).</p><p>These concerns are not unique to Garfield.</p><p>They are part of a much broader discussion taking place in education across the United States and around the world.</p><p>That is precisely why documentation matters.</p><h3>The Records Answer Some Questions</h3><p>The Pilot Agreement provides valuable insight into what Garfield Public Schools is piloting.</p><p>It identifies the vendor.</p><p>It describes the Early Warning System.</p><p>It references the IEP Sandbox Module.</p><p>It explains how the pilot converts into a paid subscription.</p><p>Those are important facts.</p><p>At the same time, the records released leave other questions unanswered.</p><p>The documents do not explain why Parthion was selected over competing platforms.</p><p>They do not identify the criteria the district plans to use when deciding whether the pilot has been successful.</p><p>They do not describe how the district intends to evaluate the effectiveness of the platform after implementation.</p><p>Nor do they explain what governance framework will guide the district’s ongoing oversight of the system.</p><p>Again, that does <strong><em>not</em></strong> mean those discussions never occurred.</p><p>It means the records released in response to this request do not provide that information.</p><p>For a technology that may influence how student information is analyzed and how educational support is identified, those are reasonable topics for parents and taxpayers to want to understand.</p><h3>Transparency Strengthens Public Trust</h3><p>Questioning how artificial intelligence is implemented should not be mistaken for opposing artificial intelligence itself.</p><p>In fact, transparency often strengthens public confidence in innovation.</p><p>When school districts explain how technology was evaluated, what safeguards were considered, how success will be measured, and how educators remain responsible for decisions involving students, they build trust with the communities they serve.</p><p>As Walther (2026) argues, meaningful disclosure is not simply informing people that AI is being used. It is helping them understand how it is used, why it is being used, and how accountability is maintained.</p><p>The records released by Garfield Public Schools begin that conversation.</p><p>The Pilot Agreement, in particular, offers important insight into how the district structured this pilot. It also raises several questions about the agreement itself, including provisions related to cost, liability, and what happens after the pilot concludes.</p><p>Those provisions deserve a closer look.</p><p>Because while this may be described as a pilot, the agreement makes clear that what happens after the pilot matters just as much as what happens during it.</p><p>Part 3 takes a closer look at what Garfield agreed to and what those terms could mean for the district, taxpayers, and ultimately the students whose information may be placed into the system.</p><h3>References</h3><p>Akgun, S., &amp; Greenhow, C. (2021, September 22). Artificial intelligence in education: Addressing ethical challenges in K–12 settings. Frontiers in Artificial Intelligence. <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/">https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/</a></p><p>Baker, R. S., &amp; Hawn, A. (2022, December). Algorithmic bias in education. International Journal of Artificial Intelligence in Education, 32(4), 1052–1092. <a href="https://doi.org/10.1007/s40593-021-00285-9">https://doi.org/10.1007/s40593-021-00285-9</a></p><p>Federal Student Aid. (2026, May 12; updated May 29, 2026). Technology security alert — Ongoing cybersecurity incident involving the Canvas Learning Management System. U.S. Department of Education. <a href="https://fsapartners.ed.gov/knowledge-center/library/electronic-announcements/2026-05-12/technology-security-alert-ongoing-cybersecurity-incident-involving-canvas-learning-management-system-updated-may-29-2026">https://fsapartners.ed.gov/knowledge-center/library/electronic-announcements/2026-05-12/technology-security-alert-ongoing-cybersecurity-incident-involving-canvas-learning-management-system-updated-may-29-2026</a></p><p>Federal Trade Commission. (2026, June). FTC gives final approval to order against Illuminate settling allegations it failed to secure students’ personal data. <a href="https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-gives-final-approval-order-against-illuminate-settling-allegations-it-failed-secure-students">https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-gives-final-approval-order-against-illuminate-settling-allegations-it-failed-secure-students</a></p><p>Garfield Public Schools. (2026a). Pilot Agreement between Garfield Public Schools and Parthion. Obtained through the New Jersey Open Public Records Act.</p><p>Garfield Public Schools. (2026b). Response to Open Public Records Act request regarding the Parthion AI pilot. Obtained through the New Jersey Open Public Records Act.</p><p>Instructure. (2026). Security incident update &amp; FAQs. <a href="https://www.instructure.com/incident_update">https://www.instructure.com/incident_update</a></p><p>Kannan, P. (2024, October 3). How harmful are AI’s biases on diverse student populations? Stanford University. <a href="https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations">https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations</a></p><p>Maine Office of the Attorney General. (2025). Data breach notice: PowerSchool Group LLC. <a href="https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/f977d150-2550-4ddc-8715-5ee5cb11c4bd.html">https://www.maine.gov/agviewer/content/ag/985235c7-cb95-4be2-8792-a1252b4f8318/f977d150-2550-4ddc-8715-5ee5cb11c4bd.html</a></p><p>New Jersey Office of the Attorney General, Division of Consumer Affairs. (2023, October 17). AG Platkin announces $49.5 million multistate settlement with Blackbaud to resolve 2020 data breach. <a href="https://www.njconsumeraffairs.gov/News/Pages/10172023.aspx">https://www.njconsumeraffairs.gov/News/Pages/10172023.aspx</a></p><p>Parthion. (n.d.). Parthion: K–12 IEP software &amp; MTSS platform. <a href="https://parthion.io/">https://parthion.io/</a></p><p>Parthion. (2026, May 28). Why we built Parthion the way we did: A whole-child early warning system that encompasses MTSS and special education. <a href="https://parthion.io/resources/why-we-built-parthion">https://parthion.io/resources/why-we-built-parthion</a></p><p>Prescient Assurance LLC. (2026, June 24). SOC 2 Report for Lantern BRP, LLC: A Type 1 Independent Service Auditor’s Report on Controls Relevant to Security. Obtained through the New Jersey Open Public Records Act.</p><p>Reuters. (2025a, May 7). U.S. school districts facing extortion attempt after hack, software provider says. <a href="https://www.reuters.com/world/us/us-school-districts-facing-extortion-attempt-after-hack-software-provider-says-2025-05-07/">https://www.reuters.com/world/us/us-school-districts-facing-extortion-attempt-after-hack-software-provider-says-2025-05-07/</a></p><p>Reuters. (2025b, May 20). Massachusetts hacker to plead guilty to PowerSchool data breach. <a href="https://www.reuters.com/legal/massachusetts-hacker-plead-guilty-powerschool-data-breach-2025-05-20/">https://www.reuters.com/legal/massachusetts-hacker-plead-guilty-powerschool-data-breach-2025-05-20/</a></p><p>Reuters. (2026, May 11). Developer of education tool Canvas issues apology after hack. <a href="https://www.reuters.com/technology/developer-education-tool-canvas-issues-apology-after-hack-2026-05-11/">https://www.reuters.com/technology/developer-education-tool-canvas-issues-apology-after-hack-2026-05-11/</a></p><p>Securities and Exchange Commission. (2021, August 16). SEC charges Pearson plc for misleading investors about cyber breach. <a href="https://www.sec.gov/newsroom/press-releases/2021-154">https://www.sec.gov/newsroom/press-releases/2021-154</a></p><p>UNESCO. (2021, November). Recommendation on the ethics of artificial intelligence. <a href="https://unesdoc.unesco.org/ark:/48223/pf0000381137">https://unesdoc.unesco.org/ark:/48223/pf0000381137</a></p><p>UNICEF Switzerland and Liechtenstein. (2024, November 20). How AI can have negative impacts on children. <a href="https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children">https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children</a></p><p>U.S. Department of Education, Student Privacy Policy Office. (n.d.). Who is a “school official” under FERPA? Protecting Student Privacy. <a href="https://studentprivacy.ed.gov/faq/who-school-official-under-ferpa">https://studentprivacy.ed.gov/faq/who-school-official-under-ferpa</a></p><p>Walther, C. (2026, January 20). Why AI disclosure matters at every level. Knowledge at Wharton. <a href="https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/">https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b79305fcb0f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Inside Garfield Public Schools’ AI Pilot: What the Public Records Reveal]]></title>
            <link>https://carlykayinnj.medium.com/inside-garfield-public-schools-ai-pilot-what-the-public-records-reveal-0e0059ea6c81?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/0e0059ea6c81</guid>
            <category><![CDATA[ai-in-education]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[accountability]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Fri, 07 Aug 2026 15:00:45 GMT</pubDate>
            <atom:updated>2026-08-07T15:00:45.911Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H20H_DUObGuhhIDNS1C3SQ.png" /></figure><h3>Part 1: Inside Garfield Public Schools’ AI Pilot: What the Public Records Reveal</h3><p>Artificial intelligence is no longer something reserved for Silicon Valley or science fiction. It is rapidly making its way into classrooms, school offices, and educational decision-making across the country.</p><p>Supporters believe AI has the potential to help educators identify struggling students sooner, streamline administrative tasks, and improve student outcomes. Others acknowledge those possibilities while raising equally important questions about privacy, cybersecurity, transparency, accessibility, algorithmic bias, and accountability. As artificial intelligence becomes more common in K-12 education, researchers and educational organizations have emphasized that successful implementation depends not only on the technology itself, but also on the policies, oversight, and ethical safeguards that govern its use (Akgun &amp; Greenhow, 2021; UNESCO, 2021).</p><p>Garfield Public Schools is now part of that conversation.</p><p>Public records obtained through New Jersey’s Open Public Records Act (OPRA) show that the district is piloting an artificial intelligence platform developed by Parthion. According to the Pilot Agreement, Garfield is evaluating two components of the platform: an Early Warning System and an IEP Sandbox Module. The agreement further states that unless either party terminates the pilot, it automatically converts into a paid subscription costing $5 per enrolled student for the first year after the pilot, then $10 per enrolled student each year thereafter. It also limits the vendor’s contractual liability to $1,000 and provides that the software is supplied on an</p><p>“AS IS” and “AS AVAILABLE” basis.</p><p>Those details immediately raised an important question.</p><p><strong><em>How did Garfield Public Schools evaluate this technology before introducing it into its schools?</em></strong></p><p>That question is not unique to Garfield.</p><p>As school districts across the country begin exploring artificial intelligence, parents, educators, and taxpayers are asking many of the same questions. What information does the technology analyze? What safeguards are in place to protect student privacy? How are decisions reviewed by educators? How will success be measured? And how can the public understand the process behind these decisions?</p><p>These are reasonable questions whenever a public school district adopts technology with the potential to influence how student information is analyzed and how educational support is identified.</p><h3>Why This Matters</h3><p>Artificial intelligence is fundamentally different from many of the educational technologies schools have adopted over the past two decades.</p><p>A learning management system stores assignments and grades. A digital textbook delivers instructional content. Artificial intelligence, by comparison, can analyze large amounts of information, identify patterns, and generate recommendations that may help educators recognize students who could benefit from additional academic or behavioral support.</p><p>That potential is exciting.</p><p>It also carries additional responsibility.</p><p>Unlike traditional software, AI systems introduce questions that extend beyond whether the technology functions properly. Parents naturally want to know what information is being analyzed, how student privacy is protected, and whether educators remain responsible for decisions involving their children. Taxpayers have an equally legitimate interest in understanding how public funds are being spent, how success will be measured, and whether appropriate safeguards were considered before implementation.</p><p>Those questions are not unique to Garfield.</p><p>Researchers have consistently cautioned that artificial intelligence should be implemented with transparency, meaningful human oversight, and ongoing evaluation to reduce the risk of unintended bias or unfair outcomes (Akgun &amp; Greenhow, 2021; Baker &amp; Hawn, 2022). UNESCO similarly recommends that AI systems remain subject to human oversight and be implemented in ways that respect privacy, fairness, accountability, and human rights (UNESCO, 2021).</p><p>This article is not about whether artificial intelligence is inherently good or bad.</p><p>It is about whether the public has enough information to understand how Garfield Public Schools evaluated this technology before introducing it into its schools.</p><h3>What Is Parthion?</h3><p>Before examining the public records, it helps to understand what Parthion is and what Garfield Public Schools is piloting.</p><p>Parthion is an educational technology company that develops artificial intelligence tools intended to help schools organize and analyze student information. Rather than replacing teachers or administrators, the platform is designed to consolidate information from multiple sources, identify patterns, and provide educators with additional insights that may support decision making.</p><p>According to the Pilot Agreement, Garfield Public Schools is evaluating two components of the platform.</p><p>The first is an <strong>Early Warning System</strong>, designed to help schools identify students who may benefit from additional academic or behavioral support. Depending on how the district configures the platform, it may analyze information such as attendance, grades, assessment results, behavioral trends, and other educational data to identify patterns that warrant further review by educators.</p><p>The second component is an <strong>IEP Sandbox Module</strong>, which relates to Individualized Education Programs for students receiving special education services. While the Pilot Agreement references this module, it provides little information about its specific functionality or how Garfield intends to use it during the pilot.</p><p>On the surface, these tools appear promising.</p><p>Educators have always used data to identify students who may need additional support. Artificial intelligence simply has the ability to analyze much larger amounts of information more quickly than a person can.</p><p>That capability is one reason AI has generated so much interest in education.</p><p>It is also why careful implementation matters.</p><p>Artificial intelligence can recognize patterns.</p><p>It may not account for circumstances in a student’s home environment that are not represented in the data it analyzes.</p><p>It cannot directly account for circumstances that are not represented in the data available to it.</p><p>It cannot replace the professional judgment of teachers, counselors, child study teams, and administrators who understand the context behind a student’s experiences.</p><p>Researchers have warned that AI systems learn from existing data. If those data reflect historical inequities or incomplete information, AI systems may unintentionally reinforce those patterns unless they are carefully monitored and regularly evaluated (Baker &amp; Hawn, 2022). More recently, Stanford researchers have highlighted concerns that algorithmic bias may disproportionately affect students from diverse backgrounds if schools rely too heavily on automated recommendations without appropriate human oversight (Kannan, 2024).</p><p>That is not an argument against artificial intelligence.</p><p>It is an argument for implementing it responsibly.</p><h3>Looking Beyond the Marketing</h3><p>As I began researching Parthion, I noticed something interesting.</p><p>Most of the publicly available information described <strong>what the platform was designed to do.</strong></p><p>Very little explained <strong>how Garfield Public Schools evaluated the platform before deciding to pilot it.</strong></p><p>That distinction matters.</p><p>Technology companies naturally focus on the strengths of their products.</p><p>School districts often emphasize the opportunities that new technology may provide.</p><p>Neither perspective fully answers the questions many Garfield families are likely asking.</p><p>· Why was this particular platform selected?</p><p>· What other products were considered?</p><p>· How was student privacy evaluated?</p><p>· How will the district determine whether the pilot has been successful before committing taxpayer dollars to an ongoing subscription?</p><p>How will the district ensure that artificial intelligence remains a tool that supports educators rather than replacing professional judgment?</p><p>These are not unusual questions. They reflect the broader conversation taking place in education as schools adopt artificial intelligence. Organizations focused on children’s rights have also cautioned that AI may introduce new risks involving privacy, profiling, and automated decision making if appropriate safeguards are not established before implementation (UNICEF Switzerland and Liechtenstein, 2024).</p><p>Transparency does not slow innovation.</p><p>Transparency builds trust in innovation.</p><h3>Why I Filed an OPRA Request</h3><p>To better understand how Garfield Public Schools evaluated Parthion before beginning the pilot, I submitted an Open Public Records Act request.</p><p>I was not asking the district to answer questions or create new documents.</p><p>Instead, I requested existing records that could help explain how the district evaluated the platform before implementation.</p><p>Among other things, I requested documentation related to procurement, privacy, cybersecurity, accessibility, governance, human oversight, and how the district plans to evaluate the pilot before it automatically converts into a paid subscription.</p><p>Those requests were guided by the same principles researchers and international organizations have identified as essential for responsible AI implementation in education (Akgun &amp; Greenhow, 2021; UNESCO, 2021).</p><p>The district produced some records, including the Pilot Agreement, while indicating that other requested records had <strong>“no responsive records”</strong> or were denied as <strong>overly broad</strong> under OPRA.</p><p>As I reviewed the response, I realized the records themselves raised an equally important question.</p><p><strong>Did the documents released provide the public with enough information to understand how Garfield Public Schools evaluated this technology before introducing it into its schools?</strong></p><p>That question begins with understanding exactly what the district’s OPRA response says and, just as importantly, what it does not say. That is where we turn next.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0e0059ea6c81" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Decision-Making in K-12 Education: Ethical Implications and Educational Concerns]]></title>
            <link>https://carlykayinnj.medium.com/ai-decision-making-in-k-12-education-ethical-implications-and-educational-concerns-cb4b88f65f1e?source=rss-64783a362389------2</link>
            <guid isPermaLink="false">https://medium.com/p/cb4b88f65f1e</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[education]]></category>
            <dc:creator><![CDATA[Carly K., M.S. Ed.]]></dc:creator>
            <pubDate>Wed, 27 May 2026 22:35:57 GMT</pubDate>
            <atom:updated>2026-05-27T22:35:57.117Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A realistic classroom scene showing a female educator or analyst reviewing student data on a laptop while a robotic AI hand points toward the screen. In the background, several students sit blurred and out of focus, suggesting observation or evaluation. The laptop displays charts, graphs, and network-style analytics dashboards. Books, papers, and soft lighting create an academic atmosphere, symbolizing the growing role of artificial intelligence and data-driven decision-making in student placeme" src="https://cdn-images-1.medium.com/max/1024/1*phWvg-hiwo_799cdGM3c-Q.png" /></figure><h3>The Growing Use of AI in K-12 Education</h3><p>Artificial intelligence (AI) and data-driven systems are becoming increasingly integrated into K-12 education. School districts now use AI to assist with student placement, intervention strategies, academic monitoring, behavioral tracking, and resource allocation. While these systems are often promoted as efficient and objective, growing concerns exist regarding bias, transparency, accountability, privacy, and overreliance on quantitative data when making decisions that directly affect students’ educational experiences and futures.</p><h3>AI and Student Placement Decisions</h3><p>In K-12 settings, AI systems are increasingly being used to support decisions regarding special education evaluations, Individualized Education Programs (IEPs), Response to Intervention (RTI), Multi-Tiered Systems of Support (MTSS), basic skills placement, gifted and talented identification, academic intervention groups, and behavioral monitoring. These systems often analyze standardized test scores, attendance records, assignment completion, behavioral data, demographic information, and socioeconomic indicators to identify patterns and generate recommendations.</p><p>However, many educators and researchers warn that these systems can unintentionally reinforce existing inequities because the data being analyzed are often tied to broader systemic issues such as poverty, unequal access to resources, language barriers, disability identification disparities, and historical bias within educational systems.</p><h3>Ethical Concerns Surrounding IEPs and Special Education</h3><p>The ethical implications become especially significant when AI systems are involved in high-stakes decisions concerning IEPs, special education eligibility, accommodations, and placement into remedial or intervention-based programs. These decisions directly affect students’ legal protections, educational opportunities, classroom environments, and long-term academic trajectories.</p><p>Federal laws such as Family Educational Rights and Privacy Act (FERPA)1 and Individuals with Disabilities Education Act (IDEA)2 require individualized consideration and professional judgment when determining educational services for students with disabilities. Critics argue that overreliance on AI-generated recommendations risk reducing highly individualized student needs into simplified data categories that fail to capture the full picture of a student’s growth, potential, and lived experiences.</p><h3>The Limitations of Data-Driven Systems</h3><p>AI systems struggle to account for non-quantifiable aspects of learning and development. Characteristics such as resilience, motivation, emotional well-being, creativity, effort, family circumstances, and learning styles cannot be fully measured through algorithms alone.</p><p>A student experiencing trauma, inconsistent access to technology, language acquisition challenges, or temporary hardships may be incorrectly flagged as “at risk” by a system that lacks the contextual understanding educators gain through direct interaction and observation.</p><h3>The Impact of Educational Labeling</h3><p>Another concern involves how AI-generated labels may shape educator perceptions and expectations. Labels such as “at risk,” “below benchmark,” or “behavioral concern” can unintentionally influence how students are viewed and supported within educational settings.</p><p>Research has consistently shown that educational labeling and lowered expectations can negatively impact student confidence, participation, opportunity, and academic identity. When these labels are generated through AI systems without transparency or strong human oversight, bias can become embedded under the appearance of objectivity.</p><h3>Student Data Collection and Consent Concerns</h3><p>Without clear and informed consent, the collection and analysis of student behavioral data raises serious ethical concerns regarding privacy, transparency, trust, and student autonomy. Even when the stated goal is “student success,” students and families may feel monitored rather than supported when they are unaware of how their information is being collected, analyzed, stored, or used to make educational decisions.</p><p>Data such as login activity, attendance patterns, classroom behavior, assignment completion, online participation, disciplinary records, and engagement metrics can create detailed behavioral profiles of students without their full understanding or consent.</p><p>When schools rely heavily on these systems, students may begin to feel surveilled instead of encouraged, which can negatively affect participation, trust, and the overall learning environment. Additionally, AI-generated “risk” labels or behavioral indicators may unintentionally influence how educators, counselors, and administrators perceive students before meaningful interaction occurs.</p><p>Students who are experiencing external challenges such as poverty, trauma, disability-related needs, mental health struggles, family responsibilities, or limited access to technology may be inaccurately categorized by the system, reinforcing existing inequities rather than addressing them.</p><h3>Transparency, Privacy, and Family Trust</h3><p>The lack of transparency surrounding behavioral data collection can weaken relationships between schools and families. Parents and guardians may not fully understand what information is being gathered, how long the data is stored, who has access to them, or how those data influence placement, intervention, disciplinary, or academic decisions.</p><p>This becomes especially concerning in K-12 education, where students are minors and legal protections such as FERPA and IDEA require careful handling of educational records and individualized decision-making.</p><p>Educational leaders therefore have an ethical responsibility to ensure that informed consent, transparency, human oversight, and data privacy protections remain central whenever AI systems are used to analyze student behavior. While data can help identify students who may need additional support, educational decisions should never rely solely on algorithmic outputs without meaningful human review, contextual understanding, and opportunities for families and students to question or appeal to those determinations.</p><h3>The Importance of Human Oversight</h3><p>Across K-12 education, a recurring concern is that AI systems are not inherently neutral. Because algorithms are trained using historical and institutional data, they can replicate and reinforce existing inequities already present within educational systems.</p><p>Organizations such as UNESCO continue to emphasize that human oversight must remain central in AI-supported educational decision-making. While AI can assist with identifying patterns and organizing information, it cannot fully account for human growth, context, lived experiences, relationships, or potential.</p><p>Educational leaders therefore face increasing pressure to implement ethical safeguards, maintain accountability, protect privacy, and ensure that AI functions as a supportive educational tool rather than a replacement for human judgment and individualized decision-making.</p><h3>Side Note: Professional Perspective and Experience</h3><p>As an eLearning instructional designer, I utilize AI systems regularly throughout my workflow for tasks such as brainstorming, organizing content, supporting accessibility efforts, assisting with course development, and improving efficiency in online learning environments. My work involves collaborating with subject matter experts, designing complex college-level eLearning courses, developing instructional materials, ensuring accessibility compliance, and integrating learning technologies into online education.</p><p>I also have training and experience in instructional design methodologies, accessibility standards such as Web Content Accessibility Guidelines (WCAG) and Universal Design for Learning (UDL), Quality Matters (QM) standards, Learning Management System (LMS) platforms, educational technologies, project management, and instructional design models such as Analysis, Design, Development, Implementation, and Evaluation (ADDIE) and Backward Design, all of which directly influence how AI tools are evaluated and implemented within course design.</p><p>Because of my professional responsibilities, training, and doctoral-level research, I have seen firsthand that while AI can be an extremely valuable support tool, the output still requires substantial human insight, oversight, and auditing to ensure the information generated is accurate, accessible, pedagogically appropriate, and aligned with course outcomes and institutional standards.</p><p>In practice, AI can improve productivity and streamline certain processes, but it cannot replace professional judgment, instructional expertise, or quality assurance. AI-generated materials still need to be reviewed by instructional designers, faculty, and subject matter experts to verify accuracy, identify misinformation, ensure accessibility compliance, and determine whether the content truly meets the needs of diverse learners.</p><h3>ANNOTATIONS</h3><p><strong>1 FERPA</strong> stands for the <strong>Family Educational Rights and Privacy Act</strong>. It is a federal law that protects the privacy of student education records and gives parents and eligible students certain rights regarding access to educational information.</p><p><strong>2 IDEA</strong> stands for the <strong>Individuals with Disabilities Education Act</strong>. It is a federal law that ensures students with disabilities are provided with a free appropriate public education (FAPE) tailored to their individual needs through services such as Individualized Education Programs (IEPs).</p><h3>REFERENCES</h3><p>Akgun, S., &amp; Greenhow, C. (2021, September 22). Artificial intelligence in education: Addressing ethical challenges in K–12 settings. <em>Frontiers in Artificial Intelligence</em>. <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/">https://pmc.ncbi.nlm.nih.gov/articles/PMC8455229/</a></p><p>Alvarez, B. (2024, April 16). Disproportionality in special education fueled by implicit bias. <em>National Education Association.</em> <a href="https://www.nea.org/nea-today/all-news-articles/disproportionality-special-education-fueled-implicit-bias">https://www.nea.org/nea-today/all-news-articles/disproportionality-special-education-fueled-implicit-bias</a></p><p>Asay, A. (n.d.). The pros and cons of labeling a child with special needs. <em>Apex Social Group.</em> <a href="https://apex-social.com/the-pros-and-cons-of-labeling-a-child-with-special-needs/">https://apex-social.com/the-pros-and-cons-of-labeling-a-child-with-special-needs/</a></p><p>Baker, R. S., &amp; Hawn, A. (2022, December). Algorithmic bias in education. <em>International Journal of Artificial Intelligence in Education, 32</em>(4), 1052–1092. <a href="https://doi.org/10.1007/s40593-021-00285-9">https://doi.org/10.1007/s40593-021-00285-9</a></p><p>Gardner, K., &amp; Leary, H. (2023). Online Learning for First-Generation and Underrepresented Minoritized Students: A Literature Review Using a Model of Student Engagement. <em>Online Learning</em>, <em>27</em>(1). <a href="https://doi.org/10.24059/olj.v27i1.3392">https://doi.org/10.24059/olj.v27i1.3392</a></p><p>International Association of Privacy Professionals. (2023, August 9). Data protection issues for employers to consider when using generative AI. <em>IAPP.</em> <a href="https://iapp.org/news/a/data-protection-issues-for-employers-to-consider-when-using-generative-ai">https://iapp.org/news/a/data-protection-issues-for-employers-to-consider-when-using-generative-ai</a></p><p>Kannan, P. (2024, October 3). How harmful are AI’s biases on diverse student populations? <em>Stanford University.</em> <a href="https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations">https://hai.stanford.edu/news/how-harmful-are-ais-biases-on-diverse-student-populations</a></p><p>Master, A., Meltzoff, A. N., &amp; Cheryan, S. (2021). Gender stereotypes about interests start early and cause gender disparities in computer science and engineering. <em>Proceedings of the National Academy of Sciences, 118</em>(48), Article e2100030118. <a href="https://doi.org/10.1073/pnas.2100030118">https://doi.org/10.1073/pnas.2100030118</a></p><p>National Center for Science and Engineering Statistics (NCSES). (2023). Diversity and STEM: Women, minorities, and persons with disabilities. <em>National Science Foundation</em>. <a href="https://ncses.nsf.gov/pubs/nsf23315">https://ncses.nsf.gov/pubs/nsf23315</a></p><p>Northern Michigan University. (n.d.). Understanding FERPA in the context of generative AI: A guide for faculty. <em>Center for Teaching and Learning.</em> <a href="https://nmu.edu/ctl/understanding-ferpa-context-generative-ai-guide-faculty">https://nmu.edu/ctl/understanding-ferpa-context-generative-ai-guide-faculty</a></p><p>UNICEF Switzerland and Liechtenstein. (2024, November 20). <em>How AI can have negative impacts on children</em>. <a href="https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children">https://www.unicef.ch/en/current/news/2024-11-20/how-ai-can-have-negative-impacts-children</a></p><p>United Nations Educational, Scientific and Cultural Organization (UNESCO). (2021, November). <em>Recommendations on the ethics of artificial intelligence</em>. <a href="https://unesdoc.unesco.org/ark:/48223/pf0000381137">https://unesdoc.unesco.org/ark:/48223/pf0000381137</a></p><p>U.S. Department of Education. (n.d.). 34 CFR part 99 — family educational rights and privacy<em>. FERPA | Protecting Student Privacy.</em> <a href="https://studentprivacy.ed.gov/ferpa">https://studentprivacy.ed.gov/ferpa</a></p><p>Walther, C. (2026, January 20). Why AI disclosure matters at every level. <em>Knowledge at Wharton</em>. <a href="https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/">https://knowledge.wharton.upenn.edu/article/why-ai-disclosure-matters-at-every-level/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cb4b88f65f1e" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>