<?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 Laura Schroeder on Medium]]></title>
        <description><![CDATA[Stories by Laura Schroeder on Medium]]></description>
        <link>https://medium.com/@laura7.schroeder?source=rss-63531480f361------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*qeMDfYl7sXDVzMGpNvKatQ.jpeg</url>
            <title>Stories by Laura Schroeder on Medium</title>
            <link>https://medium.com/@laura7.schroeder?source=rss-63531480f361------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 18:26:27 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@laura7.schroeder/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[Year 1 of Rewiring Education — Here’s What I’ve Learned]]></title>
            <link>https://medium.com/@laura7.schroeder/year-1-of-rewiring-education-heres-what-i-ve-learned-8b998c086e75?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/8b998c086e75</guid>
            <category><![CDATA[holistic-education]]></category>
            <category><![CDATA[project-based-learning]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[education-reform]]></category>
            <category><![CDATA[purpose]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Sun, 11 Oct 2026 11:01:01 GMT</pubDate>
            <atom:updated>2026-10-11T11:01:01.454Z</atom:updated>
            <content:encoded><![CDATA[<h3>Year 1 of Rewiring Education — Here’s What I’ve Learned</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BvYY6eysPs1N1W7TjnX1cA.jpeg" /></figure><p><strong>There are so many problems in our world right now.</strong> And it doesn’t seem like it’s getting better anytime soon (at least that’s what a pessimist might say)</p><p>TV shows and social media are filled with bad news, and people feel overwhelmed. Climate change, inflation, unemployment, war, corruption, and so much more.</p><p>BUT</p><p>I’m an optimist. And you should be one too.</p><p>BUT</p><p>This article won’t be about optimism, self-help, or any of these things. This article is about <strong>education</strong>.</p><p>I’ve been deeply engaged in this topic for over a year now. I’ve worked on various projects, met incredible people, and gained invaluable insights. And it all started with… <strong>The World’s Biggest Problems.</strong></p><p>At the beginning of 2025, I had enough. As a teenager growing up in a world that seems increasingly pessimistic, where you are bombarded with all the problems mentioned above, I wanted clarity.</p><p>So, I created a mind map of all the problems that popped into my mind — that I had heard about in the news, in documentaries, from acquaintances, or in class.</p><p>After a few minutes, I had to stop because too much negativity was filling up my mind. Still, I got this result:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fn7unnltsXYb2A6NuNVzXg.png" /></figure><p>Looking at that map made me wonder: WHY? Why do they exist? What is the root cause? Is there any way to solve or at least mitigate them?</p><p>And this is where it hit me. It all really comes down to… education.</p><p>Think about that: <strong>our world is governed by what is going on in people’s minds.</strong></p><p>The things that surround us in daily life were created by humans. The desk, the car, the phone. The systems that govern our modern life were put in place by humans. Finance. Economics. Healthcare. The beliefs we carry and the behavior we show — which shape how we treat others and experience life — originate in humans.</p><p>And while there are a myriad of factors that influence humans (our biology, our environment, God?), <strong>education is one of the most powerful ways on Earth to shape what is going on in our minds and therefore shape our lives.</strong> Ideally, shape our lives for the better.</p><p>Would educated people kill each other? Would educated people prioritize profits over our planet? Would educated people fall for short-term thinking and groupthink to the extent people do now?</p><p>Perhaps. But I think the number of people would be far lower.</p><p>And an important thing to note here is that I don’t mean “educated” only in an intellectual sense — I mean it in a holistic sense.</p><p>Take dancing, for example: knowing everything about the human body and different dance styles doesn’t make you a good dancer.</p><p>For education, this means that knowing a bunch of facts doesn’t make you educated, because pure information becomes almost worthless once life hits you with its everyday chaos and challenges.</p><p>This is an idea that I want you to keep in mind throughout this article — <strong>holistic education</strong>. Education is not what most people think it is. There is more to it than sitting in a classroom, listening to the teacher, taking a test, and being graded.</p><p>One could even argue that education is the exact opposite.</p><p><em>“If you wanted to create an education environment that was directly opposed to what the brain was good at doing, you probably would design something like a classroom. If you wanted to create a business environment that was directly opposed to what the brain was good at doing, you probably would design something like a cubicle. And if you wanted to change things, you might have to tear down both and start over.”</em> — John Medina</p><p>If this resonates with you, I highly recommend that you keep reading.</p><p>If this sounds strange to you, I highly recommend that you keep reading too.</p><h3>THE SYSTEM IS OUTDATED AND PROJECT‑BASED LEARNING WILL FIX IT</h3><p>Or at least that’s what I thought at the start of my journey. Don’t get me wrong, this statement is partly true, but it ignores the complexity of the education sector. But let’s start at the beginning. Let’s look at what is most apparent.</p><p><strong>There are certain problems in the school system¹:</strong></p><ul><li>Worsening PISA results (the OECD’s international student assessment)</li><li>Mental health crisis among youth</li><li>Disengaged and poorly behaved students</li><li>Teacher shortages</li><li>Educational inequality</li><li>Run-down school facilities</li></ul><p>I think this is what the general public hears most about or experiences themselves. And these are severe problems indeed. However, they are only the very tip of the iceberg. To dig deeper, it’s worth asking <em>why </em>these problems exist. Why is there a teacher shortage? Why are students disengaged? Why are school buildings crumbling? And here is where things are starting to get tricky. But first, the basics.</p><p><strong>Sleep </strong>is one of the most essential things in life. Poor sleep is associated with all sorts of negative outcomes, such as dementia, cardiovascular diseases, depression and low academic performance. Yet, schools systematically deprive children and teenagers of sleep and force us to wake up against our circadian rhythm.</p><p><strong>Physical activity </strong>is one of the most essential things in life. A lack of movement is associated with all sorts of negative outcomes, such as obesity, cardiovascular diseases, depression and low academic performance. Yet, schools systematically deprive children and teenagers of enough physical activity because we have to sit still half of the day. Luckily, there is break time and PE, but the majority of the time at school, students sit. On chairs. On the most uncomfortable chairs that exist. Just putting that out there.</p><p>Now, the problem is that enough movement is not only essential for a healthy life but also for good learning. How? In a nutshell, if the body is inactive, the brain will be too.</p><p><strong>A proper diet </strong>is one of the most essential things in life. Unhealthy food choices are associated with all sorts of negative outcomes, such as obesity, cardiovascular diseases, depression and low academic performance. Yet, many schools struggle to provide learners with healthy meals.</p><p>I could stop the article right here and already make my point: the school system is off, because it doesn’t even meet the basic physiological needs of young people. But there is much more.</p><p>So let’s dig a little deeper and get down to the fundamental structures of the school system — <strong>standardization</strong>.</p><p>The opposite of standardization is personalization. So let’s imagine a spectrum like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ANZ90uzKXkSR-MbSv-Jw8A.png" /></figure><p>Right now, we are on the far right end of this spectrum. Everything is heavily standardized and rigid. Now, what do I mean by that?</p><p>The system is based on a cluster of schools spread throughout a state or country.</p><p>Each school is divided into grade levels — an age-based cohort.</p><p>Each grade is divided into classes². And each class follows the same schedule. A schedule that chunks knowledge into subjects and 45-minute lessons³.</p><p>Within each subject, teachers follow a specific curriculum and students are graded on a specific scale.</p><p>You can illustrate it as a cluster that looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/751/1*FAm8bdUnQalkCF52wQtPAw.png" /></figure><p>I think that for most people this isn’t anything shocking. It simply is how schools work.</p><p>The problem is that this is not how learning and child development work. We as a human species and society have very different needs than what the school system can account for.</p><p><strong>We are not machines. We are humans.</strong></p><p>I’m not suggesting moving to the far left end — a world where everything is precisely tailored to a person. What I suggest is a system that is less extreme and more balanced.</p><p>But why do we have such an extreme system in the first place?</p><p>Mass schooling emerged during industrialization, and the industrial world left its mark on it. The goal was not only to make education available to everyone. It was also to build a shared national identity and spread literacy. However, one cannot ignore that schools were a clear means to an end — with the end being people that suit the industrial world. <strong>Operating principles from factories like standardization, compartmentalization, and strict measuring were transferred to education for the sake of accountability and efficiency.</strong> Whatever the original intentions, the result was a system that rewarded obedience and basic competence, which suited the needs of the labor market and the economy.</p><p>This origin gives a general idea of why schools are the way they are — <strong>because they weren’t designed for true education in the first place.</strong></p><p>But now we are here with a school system that treats students almost like a product. Sometimes subconsciously and sometimes in a very direct way, you are taught to just sit, listen and obey. You don’t learn how to sense your needs, talents and interests and pursue them. You may never experience how fun learning can be, or learn how to learn in the first place. And that is detrimental, especially in today’s world.</p><p>So, who are we going to blame for that?⁴ The teachers? Aren’t they the ones who give us all the homework and bad grades and make us take exams? At first glance, yes. They decide how they teach, how they grade and how they treat students. Teachers have certain freedoms and some teachers use them for the better (e.g., using more alternative, child-centered or project-based teaching methods). Still, in Germany, most teachers are civil servants, so their main job is largely to implement what the government prescribes. So the more you empathize with teachers, the more you understand that they are stuck in the same system as well.</p><p><strong>Students would like to learn. Teachers would like to teach. But both are stifled by a system that doesn’t provide suitable conditions for that.</strong></p><h4><em>Enter AI…</em></h4><p>Somehow, AI seems to interfere with the problems mentioned above. Students use it to cheat, teachers are occupied with finding ways to prevent cheating and parents are afraid of their children either using AI or, the opposite, not knowing how to use AI.</p><p>Over the past few years it has undeniably fueled uncertainty. Many people just understand how AI works and what it can do. And wherever uncertainty grows, fear and foolishness are never far behind. I consider this one of the reasons why some school districts started banning AI completely. And the reason I think it is foolish is that it is pure symptom patching. So there is not just a lack of expertise. People fail to recognize that <strong>the overlap of what AI can do and what schools do is too big.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ai2MsRzP6IrOYxYiJVJIeA.png" /></figure><p>Both AI and schools can provide information and assess. That overlap is where the tension begins: when schooling is reduced to delivering content and measuring what students remember, AI can seem like both a shortcut and a threat. But sharing some functions doesn’t make AI and school interchangeable. The deeper question is what education should offer beyond information, practice, and assessment. We should work out which repetitive, transactional tasks AI can take on and redesign schools to focus more on meaningful learning, human guidance, relationships, and collaboration. The goal isn’t to replace school with AI, but to use AI in ways that give schools more room to do what matters.</p><p>Just keep in mind that AI can be a problem, but it can also be part of the solution, especially when project-based learning is used in schools.</p><h4>Project-based learning</h4><p>Project-based learning (PBL, which is similar to problem- or inquiry-based learning) means that learning happens by doing projects instead of listening to lectures, filling out worksheets and taking exams.</p><p>A project starts with a <strong>driving question</strong>, e.g.: How can a nearby grocery store encourage customers to buy more organic products? How can we design a playground that is accessible to people with disabilities? Or are humans superior to animals?</p><p>Through <strong>sustained inquiry</strong>, students explore the question in depth and practice their curiosity. In this case, they’ll learn about marketing, sustainability, construction, medical conditions, ethics, social responsibility, and so much more.</p><p>Then students decide how they want to tackle the project, learn how to delegate tasks and take responsibility, developing their collaboration and communication skills along the way.</p><p>The issue and the work process are demanding, interdisciplinary, and authentic.</p><p>In the end, having practiced the 8 C’s⁵, students will have a <strong>tangible product or outcome</strong>: a proposal for the grocery store, a self-built mini playground prototype, and a panel discussion, respectively.</p><p>Doesn’t PBL seem like the perfect solution? We remove all the boring classes, stressful exams and annoying homework. There are still certain guidelines and criteria that projects need to follow, but essentially students are put in the driver’s seat of their learning experience.</p><p>This practice makes learning more meaningful too because you either learn something that you are interested in or you learn something that you can apply immediately.</p><p>And this is where AI comes into play as well. It can support students during the project, e.g., during problem analysis, ideation, or prototype building. And it can also support teachers since it can track the process and make it more transparent to them. Consequently, teachers can focus on their actual job again, which is connecting with and supporting students as they learn.</p><p>Education crisis solved, right? Right?</p><p>Well, if it’s so simple and suitable, why hasn’t it been implemented everywhere yet? And should PBL be implemented everywhere in the first place?</p><h3>OKAY, EDUCATION IS NOT THAT SIMPLE</h3><h4><strong>There is no panacea</strong></h4><p>There are certain objections one can raise against PBL, or at least, against making learning entirely project-based.</p><p><strong>As always, true virtue is created when balance exists</strong>. The same applies to learning. Right now, it’s heavily lecture-based. That’s not good. Making learning entirely project-based isn’t perfect either. A balance of these two approaches would be nice. But what would be even better is more variety in the first place. Sometimes, it’s best to just listen to someone and have them explain the topic to you, sometimes you want to do reading on your own, sometimes you can just go out in nature and observe a phenomenon, sometimes you can build a prototype with your hands or on a digital device, and sometimes you want to engage in a debate with others. Learning can take various shapes and forms and look different for everyone. <strong>Learning happens best when you have different options and can choose the mode that best suits a certain topic.</strong></p><p>When you look into the neuroscience and psychology of learning in particular, it becomes apparent what a complex phenomenon learning is. And surprisingly, it’s not just the brain that is involved in it. The body plays an important role too. This ties back to why it’s paramount that basic physiological needs are met and that the body reaches a certain state of activation as well.</p><p>As you might imagine, there is so much more that influences our learning and education in general.</p><h4><strong>Architecture</strong></h4><p>Just look at a typical school building. The architecture and furniture — rows of desks, uniform classrooms, and long, corridor-based layouts — reflect standardized assumptions about teaching and learning. To me, my school resembles a prison or a hospital rather than an inviting place for us to learn and socialize.</p><p>But school should feel like a place where you want to spend time. Warm, welcoming spaces with comfortable places to sit, natural light, and room to move could make a real difference. Different areas could support quiet focus, group work, input sessions, physical activity, hands-on projects, and conversation.</p><h4><strong>Family Background</strong></h4><p>Evidently, our family background and environment massively influence our education too. Did your parents receive proper education? Could they afford special programs or a private tutor? Did you live in a city or in the countryside?</p><p>Currently, schools tend to amplify certain inequalities among students. Especially in Germany, when you compare it to other OECD countries, the socioeconomic background of a student plays a huge role in their academic and professional careers.</p><p>Ideally, schools should be a place that does the exact opposite…</p><h4><strong>Countries &amp; Culture</strong></h4><p>Can we even speak of <em>the</em> education system or <em>the </em>school? Broadly speaking, yes. Because most school systems and therefore most schools operate on the same core principles (standardization, grading, lecture-based learning) or consist of and practice similar elements (classrooms and long hallways, homework, exams). Nevertheless, one cannot ignore the vast diversity of how systems are designed and how schools work. In Germany alone, there are basically 16 different systems since each state is responsible for the school system. And when you compare a German system to the American system, or even the Vietnamese system, you will observe various differences. Not only in how schools work, but also in the quality and value of education.</p><p>Imagine you are born in Somalia. You probably wouldn’t even think about complaining about school, about sitting in the classroom all day and listening to lectures. Not because school doesn’t deserve criticism, but because you know how incredibly rare and valuable attending school is. Education is your ticket out!</p><p>I would love to mention examples from other countries, but I honestly don’t have a lot of experience with that. All I can say is that education is shaped by the country and the culture. And one will never be able to change the system if one doesn’t account for regional and national differences.</p><p>Once I started exploring the complexities of the systems, I also realized how much education is tied to other branches and systems. And trust me, it’s a blessing and a curse at the same time. More on that below.</p><h4><strong>Politics</strong></h4><p>Surprise, surprise… politics and education are related. The obvious reason is that politics sets the laws and standards that communities, schools, and teachers need to follow. And I want to point out three problems that I’ve noticed here:</p><ol><li>Too many politicians (especially in Germany, and Bavaria in particular) carry mindsets and worldviews like “students need strict rules and authority,” “performance and pressure are important,” or “the system works — be grateful for your education.” Don’t get me wrong, every person can think whatever they want, but when you examine the statement “pressure is important for learning,” it ignores what we know about stress and learning: too much pressure hurts, and challenge only helps when students feel able to meet it. <strong>People can have their own opinion, but not their own facts.</strong> Generally speaking, this rather conservative mindset paired with a lack of imagination and openness about what school could and should be, contributes to why the educational landscape is changing so little.</li><li><strong>A system is doomed to fail once those who make the decisions aren’t those who carry the consequences.</strong> In this case, policymakers make the decision and everyone else (teachers, parents, students) has to carry the consequences… That’s why I’m a huge advocate for giving students a seat at the table. Not only when it comes to deciding on the prom theme, but also when it comes to what should be taught and how. You know, the big decisions. Because we are the ones who know school from the inside out and who the system is made for (at least at first glance). My hypothesis is that the decision-makers have very little idea about what it actually feels like to be a student, what it feels like going to school and sitting in a classroom all day. I’m convinced that there is far more discourse and exchange needed between all the stakeholders that are somewhat involved in the school system.</li><li>Even when a respectable and beneficial policy is put into place, this doesn’t guarantee that the implementation will work as well. One example is the new exam policy in my state. Teachers are allowed to base grades on alternative formats such as projects and debates. That’s nice. But will every teacher in the state all of a sudden be incredibly innovative, have all the resources and know-how to actually make use of that policy? Or will the majority just continue with business as usual? Unfortunately, I think the latter will be the case.</li></ol><h4><strong>Economy</strong></h4><p>I’m lowkey afraid that the economic system and especially work life will be even more rigged and annoying than the school system. Because even now, as a student focusing on how to improve the school system, it’s hard to ignore how the economy shapes schooling. If you think back to why the school system was designed that way in the first place, you’ll realize that economic incentives play a huge role in that. And currently, we are basically stuck in the following flow:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y6dhxuLgLj9ZCUDOy943eg.png" /></figure><p>I think that most humans need a system that keeps them stuck in the mundane world, but even these systems and the incentives change. However, it’s a rather slow process, since raising the bar of mediocrity takes time.</p><p>Currently, I see AI as a catalyst. Because as the economy and the job market change, schools will eventually be forced to adapt too. That gap between what people need in life and what school provides them with is only growing. And I’m hoping that it’ll be fixed before things get ugly.</p><p>What is important to remember, as always, is that every system we live in was put in place by humans. So if enough humans decide to change something, it is very likely to change. Tying that to the economy, which is quite a powerful branch since it governs a lot of aspects of our lives, imagine the 500 most influential business people advocating for a better school system.⁶ The system would change faster than you can say Quidditch. Obviously, this remains a dream, but it’s important for me to get across how the economic demands influence how schools work and what students are taught.</p><p>This also explains why the interplay of systems is a blessing and a curse at the same time. It increases complexity and makes change even harder, but it can also do the exact opposite. If the school system has too much inertia (which it does), you can use other systems that undergo rapid change as leverage to create momentum.⁷</p><h4><strong>Stakeholders</strong></h4><p>Everyone, duh…</p><p>Parents, teachers and students (direct)</p><p>Principals, admins, counselors, janitors, school boards (institutional layer)</p><p>Government and policy makers (governance layer)</p><p>Employers, taxpayers, higher-ed institutions (economic layer)</p><p>Community organizations, media, researchers, EdTech companies (peripheral)</p><p>It’s important to keep all these stakeholders in mind because they are the ones who make up the system. Just as important, we’re dealing with an entire system. It’s not about launching a product or providing a service (although this is part of systemic change). It’s about pushing forward the evolution of a system. And at its core, a system is upheld by people. Therefore, this endeavor will require a lot of convincing work. Again, just putting a new policy into place won’t work successfully if parents, teachers and researchers are not on the same page. Unfortunately, <strong>convincing people and making them embrace change is one of the hardest things to do.⁸</strong> And I hope that you are ready to either be convinced or convince others.</p><p>At the same time, just convincing people isn’t enough. Or at least, there is more depth to convincing people than just telling them facts.</p><p>Successful change requires a narrative that people can believe in and a vision that people can aspire to. And in order to gain clarity about that vision, one has to start at the beginning, tackle the root cause and rethink it from the ground up aka first principles.</p><h3>RETHINKING THE PURPOSE OF EDUCATION</h3><p>Addressing the elephant in the room sucks, but it is necessary. In this case, the elephant consists of the most basic, yet the most challenging questions, such as “what is the purpose of education?” “who is responsible for education?” and “what should be taught and how?”</p><p>Deeply exploring those questions will eventually lead you to metaphysics, but let’s keep it more surface level for the sake of this article. To rethink something, it makes sense to completely erase what we hold to be true and start from first principles or some other forms of axioms to establish the foundation. I have to admit that I haven’t taken such a systematic approach yet, but I’ll still put in my two cents:</p><p>Education is one of the most fundamental and most human things.</p><p>Education encompasses much more than pure knowledge acquisition and employability.</p><p>Education should focus on the wholeness of humans — the head, the hands and the heart.</p><p>Education is a transformative experience that allows us to live a more aligned and fulfilled life by giving us the capabilities we need to fulfill our needs and realize our potential.</p><p>And these capabilities don’t just have to be cognitive. They can be physical, emotional, spiritual.</p><p>And I think I would have a hard time if I had to explain it to an average person, especially if that person was brought up in the Western world because <strong>too many people hold a rather utilitarian and materialistic view of the world.</strong> They equate an educated person with someone who has a lot of degrees and knows a lot. They think that education has to be useful for one’s career. And I’m not blaming anyone for thinking like that, partly because these perceptions are ingrained in me as well, but it’s essential to move away from this sort of extremism. Education isn’t solely materialistic — and it shouldn’t be purely idealistic. It shouldn’t be focused purely on cognition — and it shouldn’t be focused purely on the body either. <strong>We need to embrace both sides of the spectrum, restore the middle and recognize the interconnectedness, e.g., of the mind, body and heart.</strong></p><p>I get it — this might seem very vague, but it gives a good guiding principle. From there, one has to develop a concept of what children need. Luckily, the sciences and humanities like psychology, sociology, pedagogy and neuroscience have made some good progress for the past 200 years. And one major shift that I’d advocate for is to recognize that we humans are innately curious and that, especially in children, this curiosity can be used as a powerful engine for learning. This is also why school shouldn’t be a mere information dispenser but rather a place that provides the right conditions for learning.</p><p>So what would be important to do right now, on an individual as well as a collective level, is to rethink not only the purpose of education but also, from there, the roles of the institutions and actors in education.<strong> The question of purpose and of the assigned role and the responsibility that comes with it must be examined for kindergartens, schools, colleges, teachers, parents, students, and possibly further institutions.</strong> For example, a teacher’s job shouldn’t consist purely of behavior management and information transmission. It should be more about supporting students in their learning journey by challenging them and providing guidance and inspiration. And students aren’t just there to absorb information and spit it out on a test — the role of students should shift more toward that of a creator and actor, not just a consumer. And as mentioned above, the role of school is to provide the right conditions for exactly that to happen. In my ideal world, schools are just one of many places that together make up an education system. A school is a place that children enjoy going to — a place of socializing, learning and growing. A place that welcomes not only students and teachers but also parents, professionals and seniors, so that they can give mentorship or pass on their knowledge to the next generation. What I also advocate for is a bigger focus on competencies with the most important one being <strong>learning how to learn.</strong> In an uncertain and ever-changing world, the one who stays adaptable is the one who inherits the future. And to stay adaptable, one has to be willing to keep learning.</p><p>The challenge is that such a role shift would require quite a few things. Schools would need to decouple themselves, at least to some extent, from the logic of the labor market — shifting the focus away from employability. What’s also required is a different kind of accountability, one that acknowledges that each student would acquire different kinds of knowledge and skills, not all of which can be measured through standardized curricula or testing. At the same time, parents would need to be relieved of some of the systemic pressure so they can truly focus on raising their children rather than just placing them in front of screens.</p><p>Unfortunately, the act of rethinking (and thinking in the first place lol) and rebuilding something on a broad and collective level is arduous.⁹ But that’s by no means a reason to refrain from doing it. Sooo, how are we gonna get there?</p><h3>As you start to walk on the way, the way appears</h3><p>That’s what I truly believe in. This might sound like a laissez-faire method, but don’t forget that you need to start walking in the first place aka TAKE ACTION.</p><p>I’d love to hand out a manual that gives step-by-step instructions on how to rewire the education system, but sadly that’s not the case. <strong>This year, I spent a lot of time trying to discover or figure out a master plan, but there is none.</strong> In that sense, no matter what action you take, it’s the right one.¹⁰ No one will change an entire system alone — it requires all hands on deck, steering in the same direction, with wind in their sails and enough resources for the long and arduous journey.</p><p><strong>There are no absolutes in education.</strong> In fact, there is a lot of complexity, nuance, and ambiguity. A policy that works at school A won’t work at school B, a method that works with student A won’t work with student B, and a fact that convinces person A won’t convince person B. But, again, that shouldn’t stop us from taking action.</p><p>Here are some examples of <strong>action items</strong> that you (yes YOU) can do right after reading this article:</p><ul><li>Research organizations and initiatives near you (or online) that advocate for change in education</li><li>Join a parents, teachers or students association at your school/in your state</li><li>Reach out to someone who is involved in the education sector (family, friends, social media)</li><li>Organize a local roundtable or event</li><li>Read further articles or books about topics like learning, schools, systemic change, and educational history.</li><li>Send this article to someone you know</li><li>Research education events near you and attend one</li></ul><p>And there is so much more you can do!</p><p>Changing an entire system seems overwhelming and close to impossible. But we need to start somewhere.</p><p><strong>Think about what matters to you. Take ownership. Take action. And keep going.</strong></p><h3>Closing Remarks</h3><p>This has been an incredible year — very busy, but very rewarding too.</p><p>Exactly one year ago, I published my first article as part of <em>The Knowledge Society</em> program. From there, things got kind of messy, but I eventually built a platform called Passion Fruit that encourages students to start building projects.</p><p>At the same time, I joined the Young Founders Network and the Bavarian Student Movement, and began creating content on social media.</p><p>During the summer, I interned at an international school and an EdTech startup, and had the opportunity to give a TEDx Talk (check it out <a href="https://www.youtube.com/watch?v=X160tb96PtM">here</a>).</p><p>While working on education projects, I also deepened my love for philosophy.</p><p>Last month, I took part in a nationwide education protest, and now I’m organizing an education event in November while continuing to create content.</p><p>Through reading, thinking, talking with smart and wise people, and taking action, I’ve gained many insights into the education system — and I can’t wait to learn more!</p><p>From the title, you can probably tell this won’t be the last article like this. My work in education is definitely no longer just an interest — it’s a commitment.</p><p>Thank you to everyone who’s been part of my journey ❤</p><p>Let’s see where this year takes me!</p><h4>Notes</h4><p>¹ I live in Germany, so the problems I describe may vary by country and context.</p><p>² In Germany, “class” and “subject” mean different things. A class is a group of 20–30 students who share a schedule and stay together over several years.</p><p>³ This can vary across schools and states.</p><p>⁴ “Blame” is a stylistic device here. I don’t mean that blame should be assigned; I mean that responsibility is either met or not met.</p><p>⁵ The 8 Cs are character, critical thinking, curiosity, creativity, connection, collaboration, communication, and creation.</p><p>⁶ You can define “better” however you want. For this thought experiment, assume the business leaders act in your interest.</p><p>⁷ Very abstract, I know…</p><p>⁸ Maybe I should just found a B2B SaaS startup and call it a day.</p><p>⁹ And that’s still an understatement…</p><p>¹⁰ Assuming you use common sense and have good intentions.</p><h3>About me</h3><p>I’m Laura Schroeder, and I’m on a mission to rewire education!</p><p>I’m driven by curiosity and a desire to make a meaningful impact by empowering people to live more self-aligned lives aka fulfill their purpose.</p><p>To me, education is one of the most powerful and beautiful ways to help people unlock their true potential.</p><p>As a high school student, I’m working on various education projects and sharing my journey online.</p><p>I’m always happy to connect with others, so feel free to reach out. And thank you for reading my work and supporting me.</p><p><strong>Let’s connect on </strong><a href="https://www.linkedin.com/in/laura-schroeder07/"><strong>LinkedIn</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8b998c086e75" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Passion Fruit: How I Turned a Messy Idea Into a Real Platform — TKS Create]]></title>
            <link>https://medium.com/@laura7.schroeder/passion-fruit-how-i-turned-a-messy-idea-into-a-real-platform-tks-create-163adcc8de7b?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/163adcc8de7b</guid>
            <category><![CDATA[tk]]></category>
            <category><![CDATA[education-reform]]></category>
            <category><![CDATA[students]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[building]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Thu, 28 May 2026 09:55:35 GMT</pubDate>
            <atom:updated>2026-05-28T09:55:35.007Z</atom:updated>
            <content:encoded><![CDATA[<p>If you want good grades, you need to understand the system. And once you do, you might become a straight-A student. But so are roughly 100 million other students.</p><p>Good grades might get you a nice certificate. But a certificate doesn’t teach you how to think across disciplines, figure things out on your own, or build something from nothing. It tells people you were good at school — that’s it.</p><p>The problem is that school was never really designed to prepare you for real life. It was designed to make you compliant, not capable. And while the world has changed dramatically, the system mostly hasn’t.</p><p>If you actually want to grow and develop real skills, discover what you’re passionate about, and create something that matters, you need to build projects.</p><p>By project I mean real ones — not the kind with a rubric and a due date. The kind where nobody tells you what to do next.</p><p>But here’s the thing nobody warns you about: <strong>building is messy.</strong></p><p>It doesn’t start with a clean idea and end with a polished product.</p><p>It starts with a vague feeling, a half-baked concept, a pivot and a scrappy prototype. You’ll spend hours on something that doesn’t work. You’ll feel lost. You’ll question whether it’s even worth continuing.</p><p>I know it because I’ve been there.</p><p>If you are a student builder, you know what I’m talking about. And even if you are not, you know how hard it is to get started. What am I interested in? What should I build? How do I learn the skills I need?</p><p>Let me tell you something: you don’t lack aspiration, you lack inspiration. And most importantly, you lack a community who is there to support you.</p><p><strong>That’s why I built Passion Fruit.</strong></p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FTyh9c_Vi0ho%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DTyh9c_Vi0ho&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FTyh9c_Vi0ho%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/e2728abcff18447789eba13c929e9890/href">https://medium.com/media/e2728abcff18447789eba13c929e9890/href</a></iframe><h3><strong>How does Passion Fruit work?</strong></h3><p>Passion Fruit is a platform that encourages students to start building projects by connecting curious and ambitious students with each other and lowering the bar to actually getting started.</p><p>Instead of polished portfolios, Passion Fruit focuses on inspiration and iteration. How?</p><p>The<strong> project gallery</strong> allows students to explore what other students have been working on. But you don’t just see the final outcome — you also see how it started as an idea. To emphasize progress over perfection, Passion Fruit is built on a phase system: <strong>Seed → Building → Iterating</strong>. That way, beginners aren’t intimidated by polished products. Even just starting with an idea is appreciated.</p><p>If you feel motivated to start a project too, you can join the <strong>Discord community</strong> and connect with like-minded student builders. You can ask for help, exchange advice, and use the resources linked in the <strong>resource hub.</strong> Once you’re ready to start, you just need to apply via a form to become a builder. This ensures that everyone in the community is genuinely passionate and trustworthy. Once verified, you can start uploading your projects to the gallery — show your work, receive feedback from other users, and inspire other students.</p><p>Whether you’ve built ten projects or zero — if you’re between 13 and 19 and curious enough to try, Passion Fruit is for you.</p><h3><strong>How did I build Passion Fruit?</strong></h3><p>After finishing the prototype, I realized I would need to make some changes — both to implement it technically as a full-stack platform and to better emphasize the core principles behind Passion Fruit.</p><p>Using AI, I mapped out what building a full-stack platform would actually look like. After deciding to use Cursor, I started with the PRD. This required me to be really intentional about my vision, the user flow, and other technical details. Before opening Cursor for the first time, I set up accounts on Supabase and Vercel.</p><p>Then came the part I thought would be the hardest: the pipeline for signing in and signing up, editing profiles, and uploading projects. Turns out, with the help of Claude and Cursor, it only took a few hours and was fairly straightforward. After surprisingly little time, the backend of Passion Fruit was done — and I was optimistic about what was ahead.</p><p>The next part was fun: brainstorming and choosing the UI. I searched online for inspiration and used Figma to sketch out a few ideas. And then… it got complicated. How do you actually translate UI ideas into a real platform? I had never really thought about that.</p><p>I figured out I could use v0 to vibe code the frontend and then feed the generated code into Cursor. Easier said than done. I’ll spare you the details — but what ultimately made the biggest difference was getting the Claude subscription and using Claude Code to implement the v0 design in no time.</p><p>I had to tweak a few things and fix gaps I hadn’t anticipated — a clear vision and user flow are absolutely essential. But in the end, it came together quite nicely.</p><p>Beyond the platform itself, I also had to think about the community side. I set up a Passion Fruit Discord server and thought carefully about its structure. I didn’t want it to feel rigid or pressuring, but I still wanted clear structure around community interactions — a place where everyone feels welcome and supported. To help student builders get started, I also collected useful resources in a dedicated resources hub.</p><p>The last piece was building a waitlist. I vibe coded a waitlist page using Lovable to gauge interest and see how many students I could reach. The response was clear — people genuinely liked the idea of Passion Fruit and were ready to sign up.</p><p>Now, I have a finished platform (although I’m sure it can be improved), a waitlist with 50+ interested students, a resources hub and a Discord server.</p><h3><strong>And here’s what I’ve learned</strong></h3><p>At this point I feel like a broken record — but again: clarity is key.<strong> </strong>When ideating and putting together your PRD, you need to think step by step and constantly ask yourself whether what you’re doing is coherent and logical. By that, I don’t mean being overly rational about your product or your vision. I mean making sure the user flow actually makes sense and thinking through technical implementation before you start building.</p><p>Another thing I’ve learned is the importance of having a plan with clear goals and deadlines. But what’s even more important is the ability to just start and adapt as you go. I made the mistake of spending a day or two planning out the next weeks and months for Passion Fruit in detail — when in reality, I would have moved faster if I had simply started.</p><p>Everyone talks about how powerful AI is and how it can 10x your productivity. I never doubted it. But after using Claude Code, Claude Design, Cursor, and other AI tools, I finally understand what people actually mean by that. Honestly — you have to experience it yourself to fully grasp it. And if I’m being honest, I’ve probably only unlocked about 1% of what AI can really do. The code may not be as clean as human-written code, and it does make mistakes — but to someone without a strong technical background, AI genuinely feels like a superpower.</p><h3><strong>To infinity and beyond</strong></h3><p>Obviously, Passion Fruit would need A LOT of improvement to go so far, but that’s not the goal.</p><p>It started out as an idea I couldn’t get out of my head, which then turned into a building journey with ups and downs and now…</p><p>I don’t know. Not because I’m lost, but because I still have to figure out what the right path is for me.</p><p>Perhaps Passion Fruit will just remain my TKS focus project, or it will turn into a side project, or I might create a startup out of it.</p><p>I know you might want to hear all the crazy plans I have for it, but I don’t want to make any promises.</p><p>But one thing is clear:</p><p>I want to empower students. I know there are a lot of incredible young people out there, but not enough. Why? Because the education system systematically suppresses potential.</p><p>However, by staying curious, going after your passions, connecting with others and learning how to create value you can unlock what you are truly capable of.</p><p>Passion Fruit is my contribution to that.</p><p>If you’re a student and waiting for a sign to start — this is it. Click <a href="https://passionfruit-waitlist.lovable.app/"><strong>here</strong></a> to join the waitlist and come build with us.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZMM2UPIM9YOLgEYz1Yftnw.png" /></figure><h3>About me</h3><p>I’m Laura Schroeder and I’m on a mission to rewire education!</p><p>I’m an Innovator at The Knowledge Society (TKS) and high school student, driven by curiosity, a desire to make meaningful impact, and the belief that every single person has the potential to accomplish incredible things!</p><p>To truly empower students to unlock their potential, education must be personalized, accessible, and grounded in real life. When designed well, it enables people to live happier, healthier, and more fulfilled lives.</p><p>I’m always eager to gain new insights, connect with and learn from others, and expand my perspectives — with the ultimate goal to understand myself and the world.</p><h4>Let’s connect on <a href="https://www.linkedin.com/in/laura-schroeder07/">LinkedIn</a></h4><h4>Check out my <a href="https://lauraschroeder.lovable.app/">personal website</a></h4><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=163adcc8de7b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Most Important Skill You Were Never Taught: How Learning Actually Works Pt. 1]]></title>
            <link>https://medium.com/@laura7.schroeder/the-most-important-skill-you-were-never-taught-how-learning-actually-works-pt-1-309dbb003718?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/309dbb003718</guid>
            <category><![CDATA[cognitive-science]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[adaptability]]></category>
            <category><![CDATA[neuroscience]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Fri, 27 Mar 2026 07:01:01 GMT</pubDate>
            <atom:updated>2026-03-27T07:01:01.278Z</atom:updated>
            <content:encoded><![CDATA[<p>The world has changed.</p><p>And it will change even more in the coming years.</p><p>What is novel one day may be outdated the next.</p><p>Especially with technological innovation like AI, a matter of days is not unrealistic.</p><p>This unprecedented pace of change increases uncertainty, creating a kind of quiet crisis, especially for students. What knowledge and skills will actually matter in 5 years let alone in 20 years? This shows that the ability to continuously update what you know and how you think is no longer a nice-to-have. It’s the foundation. It’s the skill that makes all other skills possible.</p><p>Therefore, adaptability is no longer optional.</p><p>Unfortunately, we as humans naturally dislike change because it always requires more energy than pure maintenance and routine.</p><p>To some extent, life forces us to change every now and then, e.g. when you are moving to a new city, when your morning routine gets disrupted, when you lose someone close, or it can be as simple as seeing the updated interface on your device. Still, some people cope better with uncertainty and change than others because certain strategies are more effective.</p><p>In this article, I want to talk about the most effective way of dealing with change. Put simply, it is about training your individual pipeline of incorporating new information into your system aka…learning.</p><p>Yes, that can mean sitting at your desk and studying for a test, and I suppose that’s what most people associate with learning. However, learning is such a complex and often invisible process. So, reducing it to a study session at your desk would be like saying that swimming is the same as touching something wet — it’s not.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Mfk4eceUin76TMNwZay8XQ.png" /></figure><h3><strong>What is learning and how does it work?</strong></h3><p>The more we try to define a concept, the more we realize how broad and multifaceted it is. Unfortunately for us, the same holds true for learning…</p><p>But let’s give it a shot!</p><p>Before we zoom into the mechanics of how learning works, it’s worth asking a more fundamental question: <strong>why does learning exist at all?</strong></p><p>Genetic adaptation is powerful, but slow. It takes generations for a species to change. Learning solves that problem. It is adaptation within a single lifetime, and depending on the situation, it can happen within hours. From an evolutionary perspective, learning is the ability to update behavior based on experience without waiting for evolution to catch up. It’s one of the most effective survival strategies in existence because it allows you to adapt to new environments and situations as they come.</p><p>At its core, <strong>learning is how any system updates itself based on experience to handle the future better.</strong></p><p>Now, what does that actually look like inside us?</p><h4>What happens in your mind when you learn something new</h4><p>Cognitive science sees the mind as an information processor, comparable to a computer, though as we’ll see, that analogy only gets us so far. The focus is on how we incorporate and process new information in relation to past experience and future actions.</p><p>At the start of any learning experience is the registration of incoming information, which enters our <strong>working memory.</strong> It’s worth distinguishing this from short-term memory, which is simply passive, temporary storage. Working memory is active — it’s where we hold and make sense of information in real time. And it is one of the biggest bottlenecks in learning since we can only hold around four chunks of information in working memory at once.</p><p>How information travels from working memory into long-term memory deserves an article of its own, but one of the central mechanisms is <strong>encoding</strong>. To understand this concept, imagine your brain as a large desk with a lot of different drawers.</p><p>A new piece of paper arrives. Before it can be put away, the brain automatically searches for a drawer it might belong to, for example, a related concept. When it finds one, the paper gets filed quickly and easily. When we have preexisting knowledge and mental frameworks that are closely related to the new information, it’s very easy to store it. When we seamlessly incorporate new information into a preexisting mental framework or category, it’s called <strong>assimilation or schema updating.</strong></p><p>But sometimes there’s no fitting drawer meaning that new information is in conflict with preexisting categories. Then, we need to modify or build new mental frameworks called <strong>accommodation or knowledge reconstruction</strong>. The latter certainly requires more effort. The problem is that most of us skip that effort without a second thought. Instead of restructuring, we unconsciously bend the new information until it fits what we already believe. That’s the root of confirmation bias. And it’s why the kind of learning that actually shifts how you see something, is often uncomfortable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_lBISqWKJ4LRGOJs1MuUzA.png" /></figure><p>In order for new knowledge to be truly solidified, certain neurological and biological processes need to take place, which I’ll get to shortly.</p><p>Because working memory is so limited, learning effectiveness depends heavily on how well that limited capacity is managed. This framework is known as <strong>Cognitive Load Theory (CLT)</strong>. It identifies three types of mental effort at play during learning.</p><p>First,<strong> intrinsic load</strong> is the inherent difficulty of the material itself. Intrinsic load is not fixed for everyone; it depends on a learner’s prior knowledge. Also, if there are many elements of the material that need to be understood simultaneously rather than in isolation, intrinsic load increases as well. This characteristic is called element interactivity. Understanding a single mathematical formula is easier than understanding the entire mathematical proof following it, where every line depends on the one before.</p><p>Second, <strong>extraneous load</strong> is the mental effort required to process non-essential and distracting information. When instructional design is suboptimal, extraneous load can be quite high. A pixelated image, background noise, a flipped diagram — all of these consume working memory without contributing anything useful to understanding.</p><p>And third, <strong>germane load</strong> is the productive mental effort devoted to “making sense” of information and mental frameworks. This means forming connections, constructing schemas and integrating knowledge. And this can only happen when there is enough room for it.</p><p>Both teachers and learners need to be aware of cognitive load theory! Understanding cognitive load is why studying from a messy slide deck feels so much harder than studying from a clean and well-organized set of notes. The information might be identical, but one version is eating your working memory before you’ve even started thinking.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8WQKCeVm_2QxbqZExn9ggQ.png" /></figure><p>Once knowledge has been stored in long-term memory, it can be pulled back into working memory when needed, known as <strong>retrieval</strong>. But retrieval isn’t just evidence of learning. Every time you successfully recall something, you strengthen the neural pathway. It’s important to be aware of this neurological mechanism since it can become one of your most powerful learning techniques. (I’ll soon publish an entire article on learning techniques grounded in science.)</p><p>And the deepest test of learning is <strong>transfer</strong>: the ability to take what you know and apply it in a situation you’ve never encountered before. Let’s take the analogy of a football match. No matter how often you practice a strategy or run plays during training, game day will be different from anything you’ve done before because of its inherent unpredictability. That’s why you need to transfer your skills and knowledge to the real situation. The same applies to the classroom. You might do a debate in English class where you learn how to structure an argument and communicate persuasively, but it means little if you can’t transfer those abilities to real-life contexts.</p><h4>What learning looks like inside the brain</h4><p>Now let’s go one level deeper — from how the mind processes information to what that actually looks like inside the brain.</p><p>At the most fundamental level, learning is physically realized in the brain through <strong>synaptic plasticity</strong>. Synaptic plasticity describes the ability of connections between neurons to strengthen or weaken based on experience. The brain is not a fixed structure. It is constantly reshaping itself in response to what you do, think, and encounter. Knowing that underlines why continuous exposure to new knowledge and situations is crucial. You will never reach a state of “my brain is now fully developed and won’t change anymore.” Neuroplasticity allows our brain to adapt to new circumstances even as one enters late adulthood. Especially when learning, we should use the fact that the brain isn’t a fixed structure as a superpower.</p><p>One of the most important mechanisms behind this is <strong>Long-Term Potentiation.</strong> You may have heard the phrase <em>neurons that fire together, wire together</em> — that’s exactly it. Repeated activation of a neural pathway strengthens it. Every time you repeat a movement, a fact, or a skill, you are literally reinforcing a physical structure in your brain. That’s the biological basis of practice.</p><p>Two complementary processes shape how this plays out over time. <strong>Synaptic strengthening</strong> builds up the connections you use. <strong>Synaptic pruning</strong> eliminates the ones you don’t, which is part of what we experience as forgetting. Evolution didn’t design pruning to upset you. It exists to increase efficiency and reduce energy consumption.</p><p>But synaptic strengthening and pruning don’t just happen while you’re awake and actively engaged. A lot of the most important work happens when you’re not doing anything at all, namely, when you’re asleep. During <strong>deep sleep</strong>, the brain replays patterns of neural activity from the day, consolidates what’s worth keeping, and transfers it to long-term memory. This frees up capacity in the hippocampus so the whole cycle can begin again the next day.</p><p>Even if you can’t remember anything else from what you’ve read here, remember this: sleep is not a nice-to-have. It’s not something to neglect, delay, or cut short. It is the biological foundation of everything you do, and for learning especially! Not a single learning technique is as effective and essential as sleep.</p><h4>Rethinking what learning means to you</h4><p>By now you should have a broad overview of what learning is from an evolutionary, cognitive and neurological perspective. Knowing these perspectives allows you to understand the hardware of learning, so to speak. But that’s only half of the picture. It doesn’t explain what role motivation and social interaction play in learning. It doesn’t explain why two people can sit in the same classroom, hear the same lecture, and walk away with completely different levels of understanding. Or why sitting in a classroom and listening to a lecture might result in everything but learning in the first place.</p><p>That’s why in the next article we’ll look at popular learning theories such as constructivism, social learning theory, and self-determination theory — a deeper look at the conditions under which human beings actually learn.</p><p>Understanding what the underlying principles of learning are is essential for every educator and every learner. And yet, when going through regular education, most people will never truly understand how learning works — and more importantly, how they as individuals learn best. One of the biggest tragedies of the current education system is that students leave school believing that learning is something arduous, boring, and mandatory. That it was something done to them, not something they can do for themselves. I’m not saying learning is fun and easy all the time. But it can be one of the most fulfilling, enjoyable, and useful things you’ll ever do.</p><p>Wherever you are in life, it’s never too late to become a beginner again. Starting out as a novice can be hard and intimidating — but think about what it can unlock. A new skill, a new language, a new way of seeing something you thought you already understood.</p><p>The future will invite every single one of us to become a learner again. The question is whether you show up to that with resistance — or with curiosity.</p><h3>About me</h3><p>I’m Laura Schroeder and I’m on a mission to rewire education!</p><p>I’m an Innovator at The Knowledge Society (TKS) and high school student, driven by curiosity, a desire to make meaningful impact, and the belief that every single person has the potential to accomplish incredible things!</p><p>To truly empower students to unlock their potential, education must be personalized, accessible, and grounded in real life. When designed well, it enables people to live happier, healthier, and more fulfilled lives.</p><p>I’m always eager to gain new insights, connect with and learn from others, and expand my perspectives — with the ultimate goal to understand myself and the world.</p><h4>Let’s connect on <a href="https://www.linkedin.com/in/laura-schroeder07/">LinkedIn</a></h4><h4>Check out my <a href="https://lauraschroeder.lovable.app/">personal website</a></h4><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=309dbb003718" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing Passion Fruit — Prototyping a Platform That Lowers the Bar to Project Building]]></title>
            <link>https://medium.com/@laura7.schroeder/introducing-passion-fruit-prototyping-a-platform-that-lowers-the-bar-to-project-building-44eae07f9207?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/44eae07f9207</guid>
            <category><![CDATA[projects]]></category>
            <category><![CDATA[students]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Mon, 23 Feb 2026 07:51:05 GMT</pubDate>
            <atom:updated>2026-02-23T17:13:38.947Z</atom:updated>
            <content:encoded><![CDATA[<p>When you graduate, what do you get? A diploma.</p><p>So do the other 200 people in your year. Maybe you ranked in the top percentile, but it’s still the same piece of paper.</p><p><strong>So what actually differentiates you?</strong></p><p>Sure, everyone has a different personality, different interests and different strengths. But the chances of getting hired <em>just</em> because of your personality are low. And the chances that grades alone will take you as far as you think? Pretty low.</p><p>Anyone who zooms out and looks at the current education system should see it clearly: being a good student, academically speaking, won’t prepare you for the future ahead.</p><p>Especially as AI advancements accelerate, the old playbook is becoming obsolete. Rote memorization, rigid subjects, cramming 20 pages the night before, making ugly Power Points, waiting to be told what to do next…these skills haven’t aged well…</p><p>The people who thrive are those who…</p><ul><li>can look across disciplines</li><li>show high agency</li><li>have the ability to figure things out</li><li>stay adaptable</li><li>discover what truly excites them</li><li>collaborate well</li></ul><p>That’s nothing new. These were the skills that set people apart a couple years and decades ago already. The difference now is that they aren’t nice-to-have anymore because we all need to master them.</p><p>Unfortunately, the main focus of schools hasn’t shifted enough yet.</p><p>That’s why it’s essential for students to take responsibility for themselves and start taking action.</p><p>And the most direct way to build these skills? Projects. Not the kind your teacher assigns with a deadline and a rubric. I’m talking about real ones. Easier said than done, though.</p><p>The hardest part of doing something is right before you start. First football game, first kiss, first project — it applies to all of it.</p><p>It really helps to remind yourself of Nike:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LWqxQqkb7srKMOd8iI5WQg.png" /><figcaption>sorry, had to…</figcaption></figure><p>Besides the mental barrier and self-doubt, there are other friction points too.</p><p>After talking to a lot of students and reflecting on my own journey, two friction points kept coming up in almost every conversation: the lack of ideas and the lack of community.</p><p>Most students don’t know what projects to build and they lack the community that could support them along the way. That’s where my project (Replicate #2) steps in.</p><h3><strong>Introducing Passion Fruit</strong></h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.loom.com%2Fembed%2F08e95e8ff7904c6dabf1e06a96023533&amp;display_name=Loom&amp;url=https%3A%2F%2Fwww.loom.com%2Fshare%2F08e95e8ff7904c6dabf1e06a96023533&amp;image=https%3A%2F%2Fcdn.loom.com%2Fsessions%2Fthumbnails%2F08e95e8ff7904c6dabf1e06a96023533-b410dac14bb14aaf.gif&amp;type=text%2Fhtml&amp;schema=loom" width="1910" height="1432" frameborder="0" scrolling="no"><a href="https://medium.com/media/fd1f3778f184eda331c19477bbb1352c/href">https://medium.com/media/fd1f3778f184eda331c19477bbb1352c/href</a></iframe><p>If you want to <strong>explore Passion Fruit yourself</strong>, check it out <a href="https://passionfruit.lovable.app/"><strong>here</strong></a>!</p><p><em>(Note: Currently, there is no integrated backend. This is just a prototype to illustrate the design and platform concept.)</em></p><p><strong>Passion Fruit is a platform that encourages students to build projects</strong> by showing real-world examples of what peers have built and by connecting students with each other. Instead of polished portfolios, Passion Fruit focuses on <strong>inspiration and iteration.</strong> The project gallery shows you what other high school students have actually built.</p><p>For project building, there are three categories: <strong>Seed, Building, Iterating.</strong> The expectation isn’t to upload a finished project, instead, you can just start with an idea and post about it as a Seed Project.</p><p>And because building alone is hard, Passion Fruit lets you <strong>exchange feedback and advice with other students and connect with an accountability buddy.</strong> Both of you set a goal, a deadline, and how often you’ll check in. Then you keep each other moving by sending updates and nudges to hold each other accountable. I imagine Passion Fruit as a place that focuses on the <strong>power of community and the process instead of perfection.</strong></p><p>Imagine you’re 16, you want to build something but have no idea where to start. You open Passion Fruit and see a 17-year-old who posted a half-baked idea about an app for language learners. It’s messy and unfinished, but because of that the bar feels reachable now.</p><h3><strong>How I Created Passion Fruit</strong></h3><p>I’d love to walk you through a clean, step-by-step process. But honestly? It was messy.<br>When starting with my second replicate I had a different idea in mind.</p><h4>Laying the groundwork for Passion Fruit — Introducing Project Prism</h4><p>Here’s the idea I had initially:</p><p><strong>Project Prism brings together students, educators and employers.</strong><br><strong>High school students</strong> can upload projects and give and receive feedback and advice, while also browsing the project gallery to get inspired and connect with each other. <br>Think: LinkedIn × Pinterest × TKS.life.<br><strong>Educators</strong> from schools facilitating project-based learning could create school profiles to showcase student work and inspire other schools.<br><strong>Employers</strong> could browse projects, discover talent, and launch real-world challenges — giving students problems to solve and prototypes to build.</p><p><strong>Check out the final product </strong><a href="https://good-mauve-64440173.figma.site/"><strong>here</strong></a><strong>!</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lq0yr7ljuQc7V24oFaojpw.png" /></figure><p>At the beginning:</p><p>I had an idea + I knew about vibe coding + I had Figma Pro thanks to TKS (which is why I chose Figma initially)</p><p>So I started.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ix15P8qKG0CVpK-y3eZ8Dg.png" /><figcaption>First try…quick brainstorming with ChatGPT, refining the prompt and copy pasting it into Figma Make.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KAu3Bbe3rHtCIIg-Zkgj7g.png" /><figcaption>Second try..I refined the prompt a bit more.</figcaption></figure><p>After experimenting and realizing how much excitement I felt about the idea and vision, I actually got started.</p><p>I taught myself <strong>how to use Figma</strong>, which seemed complex at the beginning but, as always, after a bit of experimenting and watching tutorials, I got the hang of it.</p><p>Another thing I had to teach myself was <strong>UI/UX design</strong>. I’ve heard of it before, yet never really understood what it actually meant. I learned about user flow, emotional design, color schemes, and much more.</p><p>For the foundation, I created frames in Figma, which made me realize how much thought goes into turning an idea into something tangible. In fact, <strong>gaining clarity on what to build</strong> and implementing that in the design and functionality while always thinking about the user perspective was challenging. Additionally, <strong>balancing simplicity with functionality</strong> is harder than it looks. Most people underestimate how much intention, effort, and detail sit behind a product — yet those details shape the entire user experience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H65ZA-rT-zuyCCpFiLnG8A.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ULjCgSuh9D_kNR6YOGiHnw.png" /><figcaption>Examples of the frames I designed for Project Prism to paste into Figma Make.</figcaption></figure><p>Once I had the frames done, I pasted them into <strong>Figma Make</strong>, which is their vibe coding tool, and started <strong>polishing up</strong>. This took up a lot of time. One, because the AI didn’t always execute the prompt the way I wanted. And, two, because I had to design and implement new pages and expand on the functions and features because I realized that the user flow wasn’t as logical and smooth as it could be. Seeing how this platform came together step by step and me doing things I’ve never done before made me feel very proud and excited.</p><h4><strong>The Turning Point</strong></h4><p>What I thought:<br>Idea → Figma design → Figma Make → Connect with Supabase → Platform ready :)</p><p>Turns out: things are a little more complex…</p><p>When I tried connecting Project Prism to Supabase, I hit a wall fast. The platform was far too complex to actually build with the skills, time, and vision I had.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/798/1*DVxtKmdMEwFECL3EM8vItQ.png" /><figcaption>Setting up a database in Supabase to store project reflections from students uploading a project.</figcaption></figure><p>I used ChatGPT to guide me through the process and somehow ended up building parts of a website from scratch using React.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/303/1*f2ypuT-LqBOtJu0RPiqsiQ.png" /><figcaption>whaaat am I doing here?!</figcaption></figure><p>I didn’t even know localhost, Windows PowerShell, or VS Code existed. I felt like I had unlocked a new layer of the Matrix — a confusing one. I was lost and lacked clarity.</p><p>What would be my Replicate? Could this already be something bigger (my Create)? Does it make sense to build a platform from scratch? Will anyone even use it? After trying out different options and talking to more people, I decided to pivot.</p><p>I cut the educators and employers part and really focused on the student community. And since I know that most students don’t build projects, I want to lower the bar to getting started and making it less intimidating.</p><p>That’s where Passion Fruit came from.</p><h4>Building It — Round Two</h4><p>Essentially, I was going through the exact same process again. <br>Idea → Vision → Design → Vibe Code → Polish up<br>The difference was that I moved faster this time since the lessons and skills I learned from Project Prism were already in my hands.</p><p>First, I had to get clear on what exactly I wanted Passion Fruit to be. How should it feel? What should it convey? Who is this for?</p><p>Then I envisioned it from a design perspective. For both platforms, that was the hardest part since there are so many possibilities for arranging form and function. What I learned from the first time was that I didn’t have to plan out every detail in the frames because vibe-coding tools (Lovable especially) usually do a decent job with UI/UX.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q1vrXySIz0QO56D57G9hEQ.png" /><figcaption>Creating Figma frames to make the vibe coded result more precise.</figcaption></figure><p>After creating a few frames, I continued working on the prompt.</p><p>I made the prompt as precise and detailed as possible and then started vibe coding in Figma first. Since I’d been working on a project alongside Lovable (aka TKS Challenge) at the same time, I tried vibe coding with Lovable too. I preferred the design, which is why I switched to Lovable eventually. In Lovable, I kept iterating on the functions and design, which helped me boost my vibe coding skills and design awareness.</p><p>It isn’t built out fully yet because I wanted to leave space for variance and iteration when building the actual platform. What I mean by that is that my intention wasn’t to create <em>the</em> perfect design because, depending on the feedback and technical constraints, I’ll probably have to make some changes anyway. Overall, despite the messy start and pivot, I’m content with how it turned out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*n0XPB8sqlsTxPoouDSSduA.png" /></figure><h3><strong>What’s Next</strong></h3><p>Passion Fruit is a prototype right now, but I’m genuinely excited about where this could go. The goal is a full-stack platform with an improved design, a real backend and actual users.</p><p>My plan is to talk to more people, especially students, to validate the idea. I also still need to figure out what vibe coding tool I will use because despite Lovable and Figma being great for design and prototyping, they feel too limited for a full-stack platform. I’m hoping to refine my vision, the functionality and user flow of the platform as well.</p><p>Now I’ve built two platform prototypes, pivoted once, and learned more through messy trial and error than sitting in a classroom.</p><p>The diploma doesn’t differentiate you. What you build when nobody asked you to does.</p><p>If you’re a student stuck at the starting line — Passion Fruit is for you.</p><h3>About me</h3><p>I’m Laura Schroeder and I’m on a mission to rewire education!</p><p>I’m an Innovator at The Knowledge Society (TKS) and high school student, driven by curiosity, a desire to make meaningful impact, and the belief that every single person has the potential to accomplish incredible things!</p><p>To truly empower students to unlock their potential, education must be personalized, accessible, and grounded in real life. When designed well, it enables people to live happier, healthier, and more fulfilled lives.</p><p>I’m always eager to gain new insights, connect with and learn from others, and expand my perspectives — with the ultimate goal to understand myself and the world.</p><h4>Let’s connect on <a href="https://www.linkedin.com/in/laura-schroeder07/">LinkedIn</a></h4><h4>Check out my <a href="https://lauraschroeder.lovable.app/">personal website</a></h4><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=44eae07f9207" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Rewiring Education]]></title>
            <link>https://medium.com/educate-pub/rewiring-education-why-traditional-schooling-fails-and-how-project-based-learning-creates-20f1a6310036?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/20f1a6310036</guid>
            <category><![CDATA[lifelong-learning]]></category>
            <category><![CDATA[k-12-education]]></category>
            <category><![CDATA[21st-century-skills]]></category>
            <category><![CDATA[project-based-learning]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Sun, 04 Jan 2026 16:16:33 GMT</pubDate>
            <atom:updated>2026-07-28T16:41:23.028Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why traditional schooling fails and how project-based learning creates lifelong learners</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hwZ-31azAITJK3tdsBR_zg.png" /></figure><p><strong>Imagine this:</strong></p><p>Every day, your alarm goes off early in the morning.</p><p>You get up — not because you’re motivated or rested — but because you’re expected to. You pack your bag, follow the same route, and stress about being on time. And then…the bell rings.</p><p>You look around you. Row after row of identical desks. Tired students sitting in their seats. An adult stands at the front of the room. The adult speaks. You listen. You write. You try to stay alert.</p><p>The bell decides when you sit, when you move, when you leave — sometimes even when you’re allowed to pee. It rings again. You stand up, squeeze into a crowded, windowless hallway, and walk to the next room. <br>And you do it all over again. Every hour, every day, every year. Welcome to school!</p><p>Am I exaggerating? Maybe. Is there still a lot of truth to this scenario? Absolutely!</p><p>Too many students still experience school this way. Their lives are defined by school schedules, homework, and obsolete expectations. They are fed information that is irrelevant and disconnected from their own lives. They memorize what they’re told, repeat it when asked, are measured by grades — and forget most of it soon after.</p><p><strong>You don’t need to be an expert to know that this is not how children learn and thrive.</strong></p><p>In fact, most people are already aware that something is wrong with education. Burned-out teachers, run-down schools, insufficient digitalization, and distracted students. However, these are just the symptoms. So the real question is: what is the root cause?</p><p><strong>In a nutshell, schools were designed for (1) people who don’t exist, (2) for a world that no longer exists, (3) and with methods that shouldn’t exist.</strong></p><p>Let me explain:</p><ol><li>Schooling is fundamentally teacher-centered. Even when we claim to focus on students, schools are still built around the idea of the <em>average student</em>. Who is that!? Right, the average student doesn’t exist because each human is incredibly individual and therefore has different needs — especially when it comes to learning.</li><li>In most European countries and in North America, industrialization heavily shaped the education system and laid the groundwork for how it looks today. Back then, life was more predictable. Society did go through many changes and experienced major advancements in science and the economy, but this is not comparable to the speed of today’s world. Two hundred years ago, some of the main goals of education were to educate the masses and produce reliable factory workers. The former was achieved through widespread standardization, the latter through “education”. These approaches worked back then. But guess what? The world has changed — and it will change even more in the coming years and decades. Not just a little, but drastically. Policymakers, teachers, students, and parents need to stay adaptable. The world a student enters in first grade is already different by the time they graduate — and that gap will only continue to widen. <strong>Adaptability is no longer optional.</strong></li><li>Let’s take a closer look at the methods. Obviously, each teacher, each school and each country uses different methods. However, similar patterns can be seen across all of them. Passive, lecture-based instructions is one of them. The teacher talks for an hour straight in front of the class, asks students from time to time, and scribbles barely legible words on the board. Listening is important, but as a dominant teaching method, it drastically reduces the likelihood of real learning. Students are being talked at and not talked to.</li></ol><p>The same is true for note-taking and rote memorization. Memorization without context is closer to forgetting than to understanding. We memorize, we take the test, we forget.</p><p>Actually, at least from a student’s point of view, this process is more nuanced. You are being told how important grades are and how they will impact your life. Pressure and extrinsic motivation build up. You stress, you study, and freak out eventually. You take the test. You are nervous before getting the results, and more often than not, you are disappointed and discouraged after getting the results. You begin to despise learning and school altogether, but since your future depends on it, you better get ready to do this cycle all over again. Learning becomes a performance, not an exploration, and certainly not something enjoyable. It is widely known that grades don’t always mirror true understanding and competence, are rarely objective, and are never good for the mental health of young people. <strong>We shaped education to make it measurable — and forgot that quality matters more than quantity.</strong></p><p>This mismatch between how schools expect students and the world to be, and the actual reality, is largely caused by the dominant approach used by most education systems worldwide: <strong>standardization.</strong></p><p>Again, the extent depends on the system. When I compare a U.S. high school to a German high school, German schools tend to be more standardized. For example, most students are required to take the same courses and have limited choice in what they study, whereas course selection is generally more flexible in the U.S. Additionally, individual schools and school districts in the U.S. often have more autonomy than their German counterparts. If we broaden the comparison to include East Asian countries such as Japan or South Korea, it becomes clear that these systems face even more nationwide standards and restrictions.</p><p>Standardization doesn’t only include things like curriculum-based state standards. It appears in many other areas as well. Assessment standardization determines how learning is measured and compared. Governance standardization defines rules, regulations, and the degree of autonomy. Outcome standardization ensures that certificates, grades, and qualifications are the same across regions. Standardization is also embedded in the physical and organizational design of schooling itself. School architecture and furniture — rows of desks, uniform classrooms, and long, corridor-based layouts — reflect standardized assumptions about teaching and learning. Lesson formats, bell schedules, subject segmentation, age-based cohorting…you name it!</p><p>All of this is based on the assumption that all students are the same, but this assumption couldn’t be further from the truth.</p><p>Everyone is different. We all have our own strengths, weaknesses, passions, backgrounds, preferences, beliefs, routines, needs, values and goals. Not just students, but teachers and parents as well. The education system (and other systems as well…) need to take that into account!</p><p><strong>Education should adapt to students, not the other way around.</strong></p><p>Education needs to be personalized. We must step back and return to first principles. What is the purpose of education? Why do we learn in the first place? How does learning actually happen? What does it mean to be a teacher?</p><p><strong>The traditions of the past are not a justification for the future.</strong></p><p>Don’t get me wrong — removing standardization entirely would not work.</p><p>Let me explain it with an analogy. In health, a common standard is “eat healthy.” It’s a great guideline — but it looks different for different people. But eating healthy means something different for everyone. For some, it might be intermittent fasting or quitting sugar and alcohol; for others, it might be switching to a vegetarian diet or only eating home-cooked meals. In education, personalization doesn’t mean getting rid of standards, guidelines, and recommendations. Instead, the degree and form of development and design needs to be adjusted to each individual. Another useful analogy is bowling. Beginners often need gutter guards (aka guidelines). But how the player throws the ball, how it moves, and what the outcome is remains individual. Structure enables learning; it does not dictate it.</p><p>So, how should education be? Obviously, that’s a tough question where everyone will give different answers. Frameworks such as the 4 C’s from the Partnership for 21st Century Skills (P21) and the 6 C’s from New Pedagogies for Deep Learning (NPDL) really resonate with me. Those are a good starting point. The C’s lay the groundwork for what education should look like. Here’s my vision:</p><h4><strong>Character — educated people decrease their chances of being an a**hole.</strong></h4><p>Students should know what it means to be empathetic, patient, helpful, grateful, respectful, confident, and trustworthy. By exposing students to diverse and sometimes challenging situations, teaching disciplines such as philosophy and psychology, and giving them the freedom to explore without immediate punishment, schools can support children in forming their character. However, schools shouldn’t tell students how they should be, what they should like, and how they should behave. We don’t force a plant to grow into an exact shape; we only provide the right conditions so it can grow on its own.</p><h4><strong>Critical thinking — don’t you dare outsource your own thinking.</strong></h4><p>Critical thinking holds us true to ourselves and is the filter and compass in a world full of noise. Although students are increasingly asked to evaluate statements or decisions, schools don’t treat critical thinking as an essential skill. Critical thinking can’t just exist as an evaluation task on a test that intends to assess how deeply one has understood the concept and can argue about it. If we have a classroom full of critical thinkers, will the teacher be able to go on with the lesson as usual? — No. There will be questions, hesitations, and discussions. And that’s what we want: students who can think independently. Students should be equipped with the skill to take in information, cross-check it against their prior knowledge and values, stay open-minded and adapt when necessary.</p><h4><strong>Curiosity — there’s always more to explore.</strong></h4><p>Every single human is curious by nature. We are born into this world and the only way to make sense of it all is through exploration driven by curiosity and inquiry. Curiosity enables continuous learning and fuels growth — a foundational skill for life. Obviously, constant shallow curiosity and critical thinking will stop you from getting things done. The older we get, the more precise, intentional, and deep our questions should go.</p><p>But since schools teach us that it is wrong to be wrong, standardize learning, and limit autonomy, there is no way for curiosity to thrive. That’s where personalization comes in. Students need to be exposed to a broad range of topics and knowledge, while also being allowed to go deep into what genuinely excites them. It’s okay to go off the path, and it’s okay to make mistakes — that’s what teachers and students need to embrace. <strong>When schools cultivate curious minds, education no longer ends at graduation — it becomes lifelong.</strong></p><h4><strong>Creativity — connecting things that haven’t been connected before.</strong></h4><p>Creativity is not just for artists. Creativity lies in us, humans. Creativity allows us to see and create things in novel and unprecedented ways. When integrating creativity in schools, it’s not about having more art classes (although that can be part of it). It means equipping students with the skill of figuring things out and knowing that there is always another way. There are a hundred ways to solve a problem, yet only through creativity and curiosity are we able to uncover them. At school, students should be allowed to follow their own path and not be penalized, e.g., through bad grades, when their solutions differ from what was taught in class. Creativity is what makes us humans, collectively and individually; it’s a way of thinking, behaving, and seeing the world, and creating something new.</p><h4><strong>Connection — everything is interconnected.</strong></h4><p>Subjects don’t exist in the real world; principles and phenomena do. By separating knowledge into subjects, students only get to know half the story.</p><p>Students get distracted because they are not engaged. We are not engaged because the material we are being taught is not relevant to us. It is not relevant to us because it doesn’t connect to our life and the real world. Schools shouldn’t exist as a parallel reality. They should be embedded within their communities, addressing real problems and creating mutual value.</p><h4><strong>Collaboration — try to succeed in life without others, and you’ll fail.</strong></h4><p>At times, students are forced to work alone, and during projects (I’ll go into more detail on the difference between doing projects and project-based learning later), are suddenly required to work within groups. Effective and fruitful collaboration is a form of art. Unfortunately, most people fail. Therefore, when it comes to collaboration, students should learn two key things: <em>how </em>to collaborate and <em>when </em>to collaborate. People should be able to handle tasks for themselves, but they should also know when to call for help. And once they decide to work together, they should know how to do so.</p><h4><strong>Communication — you are only as good as you can communicate.</strong></h4><p>One key, if not the key, to successful collaboration is communication. Expanding vocabulary, deepening language understanding, practicing reading, writing, and talking, and learning about communication, e.g., from a psychological perspective, is essential.</p><p>Currently, the biggest focus is on spelling and grammar. Yup, that’s important! But do students really need to know that “<em>go boldly”</em> is a split infinitive? Probably not. What we do need is the ability to know how to express ourselves and how to engage with others. We cannot think and discuss complex and profound topics if the only adjectives one can use don’t go beyond “weird”, “cool” and “sus” and half of our sentences are filled with uhms and ahs.</p><p>By explicitly teaching communication strategies and creating constant opportunities to practice — through discussion, presentation, writing, and feedback — students learn to articulate ideas with confidence and clarity.</p><h4><strong>Creation — we can all consume, but very few can create.</strong></h4><p>Learners consume information, memorize it, and spit it out on the test. Our memory is limited, and it’s an illusion if policymakers assume that by the time of graduation, students will know most of what they have learned. Indeed, a large portion of school knowledge is forgotten. Education therefore needs to change in two ways: first, learning must be designed to improve retention. Second, it must shift from passive consumption to active creation. <strong>Creation turns knowledge into something usable, personal, and meaningful — and that’s what lasts.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fQsFQYIhH-o3c2VLchqpoA.png" /></figure><h3><strong>Project-Based Learning — Turning the vision into reality</strong></h3><p>We can keep patching the symptoms and create cute add-ons to education with little impact. Or we can rethink the way education is done and actually help students understand themselves, the world, and create their place in society. We need a student-centered approach that equips learners with what they need to thrive in the 21st century. That’s where project-based learning comes into play.</p><p>Project-based learning (which is similar to problem- or inquiry-based learning) puts the student at the center of the learning experience. At the beginning of this experience is a driving question, allowing learners to follow their curiosity and explore how different topics are connected to one another. Through sustained inquiry, students explore the question in depth, practice creative and critical thinking, and refine their understanding.</p><p>For example: How can a nearby grocery store encourage customers to buy more organic products? How can we design a playground that is accessible to people with disabilities? Or are humans superior to animals? They’ll learn about marketing, sustainability, construction, medical conditions, ethics, social responsibility, and so much more.</p><p>Students decide how they want to tackle the project, learn how to delegate tasks and take responsibility, developing their collaboration and communication skills along the way.</p><p>The issue and the work process are demanding, interdisciplinary, and authentic.</p><p>In the end, students will have a tangible product or outcome: a proposal for the grocery store, a self-built mini playground prototype, or a panel discussion, respectively. This ensures that learning becomes visible and meaningful, while practicing the 8 C’s.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wF3nSvaMNGoEQLewfe_iig.png" /></figure><p>This was a very brief run-through of project-based learning. Don’t worry — over the next few articles, I’ll go deeper into PBL. What is the definition and process of PBL? What frameworks and best practices? What are the benefits and challenges concerning PBL? Stay tuned!</p><p>Project-based learning is not a nice-to-have add-on. It’s a necessary shift in rewiring education for the 21st century. Education can no longer rely on standardization, isolation, competition, passive instruction, and memorization. Especially in the age of AI, so-called “soft skills” must move to the center of learning. Those who can use AI, know how to communicate and collaborate, follow their creativity and curiosity, think critically, and create instead of merely consume, are the ones who will inherit the future. Those who think that good grades, empty knowledge, and obedience to pre-defined schedules and paths alone will lead to a successful, let alone fulfilled life will be left behind.</p><p>Either the education system starts to change and adapts intentionally, or societal and scientific advancements will accelerate so rapidly that disruptive forces will push the system to change. Giving every student a Chromebook or adding a single project day at the end of the year are steps in the right direction, but just a drop in the bucket.</p><p><strong>If we want a true paradigm shift, we need to redefine the purpose of education, rethink how schools and teachers operate, empathize with the student perspective, and remain adaptive.</strong></p><p><strong>Imagine this:</strong></p><p>Every day, your alarm goes off in the morning.</p><p>Your body feels rested, and your mind is curious about what the day will bring.</p><p>You arrive at a place, feel the welcoming and colorful atmosphere of the entrance hall, and walk past the makerspace, where younger students are excitedly sketching ideas for a robotic arm. After checking in with your mentor, you join your peers to evaluate the interview responses from employees at a nearby nursing home. You can’t wait to send your proposal to the local health department.</p><p>You decide when to sit, when to move, when to take a break.</p><p>You are trusted. You are challenged. You are heard.<br>And over time, you are becoming a capable, compassionate, creative, and curious part of society.</p><p><strong>Welcome to school — reimagined.</strong></p><h3>About me</h3><p>I’m Laura Schroeder and I’m on a mission to rewire education!</p><p>I’m an Innovator at The Knowledge Society (TKS) and high school student, driven by curiosity, a desire to make meaningful impact, and the belief that every single person has the potential to accomplish incredible things!</p><p>To truly empower students to unlock their potential, education must be personalized, accessible, and grounded in real life. When designed well, it enables people to live happier, healthier, and more fulfilled lives.</p><p>I’m always eager to gain new insights, connect with and learn from others, and expand my perspectives — with the ultimate goal to understand myself and the world.</p><h4>Let’s connect on <a href="https://www.linkedin.com/in/laura-schroeder07/">LinkedIn</a></h4><h4>Check out my <a href="https://lauraschroeder.lovable.app/">personal website</a></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*75X_AixoZXocR_3Irk_vEw.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=20f1a6310036" width="1" height="1" alt=""><hr><p><a href="https://medium.com/educate-pub/rewiring-education-why-traditional-schooling-fails-and-how-project-based-learning-creates-20f1a6310036">Rewiring Education</a> was originally published in <a href="https://medium.com/educate-pub">Educate.</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How an AI chatbot can foster critical thinking and deep understanding — fine-tuning Phi2 for…]]></title>
            <link>https://medium.com/@laura7.schroeder/how-an-ai-chatbot-can-foster-critical-thinking-and-deep-understanding-fine-tuning-phi2-for-a9292a807fc7?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9292a807fc7</guid>
            <category><![CDATA[qlora]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[fine-tuning]]></category>
            <category><![CDATA[socratic-dialogue]]></category>
            <category><![CDATA[education]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Mon, 03 Nov 2025 10:22:57 GMT</pubDate>
            <atom:updated>2025-12-12T14:06:55.212Z</atom:updated>
            <content:encoded><![CDATA[<h3>How an AI chatbot can foster critical thinking and deep understanding — fine-tuning Phi2 for socratic dialogues</h3><p>Will AI make us dumber by replacing our thinking? And should students be able to use AI in the first place? It is essential to ask such questions and critically reflect on how, when and why we use AI.</p><p>Especially when time is limited, people tend to use AI to replace hard work, which usually involves critical thinking, idea generation, and deep reflection. Obviously, this is an example of what AI <em>shouldn’t </em>be doing. But how can we avoid this from happening? This question becomes even more relevant when using AI in education. If students use AI incorrectly, it will undermine their cognitive and skill development. Therefore, we need to improve AI literacy among students, while also ensuring that tool providers take responsibility for building AI systems that encourage learning and thinking rather than replacing them.</p><h3><strong>My idea: fine-tuning a Language Model to engage in Socratic dialogue</strong></h3><p>With the help of<a href="https://towardsai.net/p/machine-learning/from-quantization-to-inference-beginners-guide-for-practical-fine-tuning-qlora-with-mistral-7b"> this tutorial</a>, I fine-tuned the Microsoft Language Model Phi2 to create a chatbot that doesn’t just answer questions but asks the student about their current knowledge and train of thought instead. This ensures that students think critically and truly understand the topic, instead of letting AI do all the work.</p><p>First of all, let’s clarify some important terms:</p><h4><strong>Socratic Dialogue</strong></h4><p>We have all experienced it: we read ChatGPT’s answer, but after a couple of seconds have already forgotten what we just read and therefore didn’t learn anything. We are blinded by the fast, well-rounded, and nicely structured answer. This is where Socratic Dialogue comes into play. It is a method that can be used in discussions, especially in exchanges between learners and educators. Instead of giving the answers right away, the educator asks carefully chosen questions to guide the learner on the journey to finding the answer. By focusing on examining one’s knowledge and assumptions, this method fosters critical thinking, reflection and deep understanding.</p><h4><strong>Fine Tuning</strong></h4><p>Fine-tuning is a method used to train a model on a specific task. Think of the following analogy: a person already knows how to speak English, but is taught how to talk like a lawyer. One is taking pre-existing knowledge and, by training the person on new information, they can perform new or more specific tasks. The same thing happens to the language models.<br>Before fine-tuning, the model has already been trained and performs a wide range of tasks thanks to NLP (Natural Language Processing) and machine learning. Hence, the parameters of such a pre-trained language model (e.g., GPT, LLaMA, or Mistral) are well-adjusted for the data and tasks it was trained on. This process is called pre-training.</p><p>Now, to make the model more customized by giving it the ability to do a specific task, we fine-tune it. First, the model is fed the new, smaller, and domain-specific dataset. Second, the model is trained on the new data. This means that the weights and parameters are adjusted accordingly to fit the new data. This process is called back propagation. The values of the newly-trained parameters need to be stored somewhere, which can take up a lot of storage. Lastly, the model has learned to accomplish a new or more specific task, such as responding in a kinder tone or adjusting the format of answers.</p><p>On the one hand, fine-tuning is a very helpful method because it enables the model to adjust without having to train a new model from scratch, thereby increasing efficiency, practicability, customization, and control. On the other hand, it is still expensive and time-consuming to train a language model, LLMs in particular, because all the updated parameters have to be stored, and there is a risk of catastrophic forgetting and overfitting as the model forgets what it has learned during pre-training. Fortunately, there is a solution.</p><h4><strong>PEFT</strong></h4><p>Parameter-Efficient Fine-Tuning (PEFT) provides an effective approach by minimizing the number of parameters that need to be updated and reducing memory usage, all while maintaining performance similar to full fine-tuning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/847/1*mjyHnGfgSjjZF4S5ToVfyQ.png" /><figcaption>This figure shows the evolution of PEFT methods over recent years , with models on the same branch sharing certain characteristics.</figcaption></figure><p>To understand how PEFT works, let’s take a look at the architecture of language models.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4agXMSXTkKyCX4f3E2wxAw.jpeg" /><figcaption>This image depicts the basic components and architecture of a transformer and adpater layer. Each adapter consists of two feed-forward neural networks.</figcaption></figure><p>An LM consists of several transformer layers. For PEFT, the adapter layer is what interests us because this is where the adjustment happens. Such an adapter layer basically consists of two feed-forward neural networks and is crucial for PEFT. The model is still fed with the new data, but the difference with adapters is that instead of adjusting all the weights of the entire model, one only adjusts the adapter layers while freezing the other weights. Therefore, only the adapter weights need to be stored; hence, PEFT takes up less storage, making fine-tuning more efficient. Surprisingly, although only a fraction of parameters is updated, the accuracy of the model output is still comparable to the original LM without adapters, underlining efficiency.</p><h4><strong>LoRA</strong></h4><p>LoRA (Low-Rank Adaptation) is one method that can be used for PEFT. Let’s walk through it.</p><p>As explained earlier, an LM consists of multiple layers, each performing matrix multiplication. The first step is to check each row of the matrix for linear dependencies. If a row is linearly dependent, this row can be calculated by knowing what the dependency looks like. The maximum number of linearly independent rows is called the rank of a matrix (r), representing the amount of unique information it contains.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/877/1*7zV6ty54XrH2D6WycAQFFQ.png" /><figcaption>This illustration shows an exemplary matrix, where column C1 and C2 are linearly independent → rank 2. It shows how the matrix is represented by two smaller matrices storing less parameters.</figcaption></figure><p>Once we know the rank, a pre-trained weight matrix W= d x k can be represented as W = A x B, where A consists of d x r and B consists of r x k under the condition that r is much smaller than both d and k. LoRA uses the rank of a matrix to remove any redundancies that a model has by reducing each matrix to two smaller matrices. This decomposition reduces the number of parameters that need to be updated and stored during fine-tuning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SMOQgIl5G5LifyrHua_MoQ.jpeg" /><figcaption>The martix W is made up of the smaller martices A and B.</figcaption></figure><h4><strong>Quantization</strong></h4><p>Quantization allows us to fit the same model in a smaller GPU by representing numbers with fewer bits. Remember that n-bits contain 2^n values and 8 bits equal 1 byte. Normally, the parameters of neural networks are stored in floating-point numbers, which is a way for the computer to represent a real number. Most LMs use 32-bit floating-point numbers (FP32) for all parameters, meaning that one number uses 32 bits. One weight can be stored as 0.123456789, for instance. By making the numbers less accurate, e.g., 1.2, we allow the model to use fewer bits for an approximated version of the stored information. In our case, Phi2 has 2.7 billion parameters (2.7B).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/873/1*dcr6c95wRrZLeofcYoIcQQ.png" /></figure><p>As shown in the grid, Phi2 has a total size of 10.8 GB, which is quite a lot. Using quantization, this size can be reduced tremendously. When compressing a model, one needs to accept the tradeoffs, such as a decrease in accuracy.</p><p>In my case, I used 4-bit quantization to load Phi2, allowing me to use it for my fine-tuning purposes while still offering acceptable accuracy.</p><h4><strong>QLoRA</strong></h4><p>QLoRA is a combination of quantization (Q) and Low-Rank Adaptation (LoRA), which helps users like me to train an LM efficiently.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1021/1*DhlySviTjfjY6i78Y24hhQ.png" /><figcaption>This figure compares different fine-tuning methods and their memory requirements. QLoRA enhances LoRA by quantizing the transformer model to 4-bit precision.</figcaption></figure><h3><strong>Implementation</strong></h3><p>I used Google Colab to fine-tune Microsoft’s model Phi2 on a dataset from OpenAI available on HuggingFace, consisting of 8,500+ grade-school-level math word problems designed to test reasoning and step-by-step problem solving. I chose the dataset configuration that is written in Socratic style to suit my purpose and I used QLoRA for PEFT.</p><p><a href="https://vimeo.com/1133418970?share=copy&amp;fl=sv&amp;fe=ci">Click here</a> to check out my video to see how exactly I fine-tuned the model and result looks like.</p><p>Here are some examples of prompt — response pairs:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QYZKsrh2SZkQ2Du87E_AlQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NfA--PKtVl1LArxHltn3nw.png" /><figcaption>Instead of answering the question right away, the chatbot generates a question to support the learner in retrieving their own knowledge and eventually getting to the answer through scaffolding.</figcaption></figure><h3><strong>Problems/Limitations I faced</strong></h3><ul><li>Finding the right LM and dataset</li><li>Debugging the code</li><li>Sometimes the chatbot still responds in an unhelpful way</li><li>Just one prompt at a time → no memory of previous conversations</li></ul><h3><strong>What’s next?</strong></h3><ul><li>Training the model on more training examples → try adding multiple datasets</li><li>Create a chatbot that can store previous conversations</li><li>Include more guardrails for the model to follow Socratic Dialogue</li></ul><h3>About me</h3><p>I’m Laura Schroeder, a high school student and Innovator at The Knowledge Society (TKS), driven by curiosity, a desire to make meaningful impact, and the belief that every single person has the potential to accomplish incredible things! My deepest passions lie in education and health — two fundamental pillars of life that are deeply interconnected. By combining the insights of the humanities with the power of AI, we can revolutionize both fields, raise living standards, transform the way we live and advance humanity as a whole. I’m always eager to gain new insights, connect with and learn from others, and expand my perspectives — with the ultimate goal to understand myself and the world.</p><h3>Let’s connect on <a href="http://www.linkedin.com/in/%20laura-schroeder07">LinkedIn</a></h3><h3>Check out my <a href="https://lauraschroeder.lovable.app/">personal website</a></h3><h3>Sources</h3><p><strong>LoRA: Low‑Rank Adaptation of Large Language Models</strong> — Hu, E. J., Shen, Y., Wallis, P., Allen‑Zhu, Z., Li, Y., Wang, S., Wang, L., &amp; Chen, W. (2021). arXiv:2106.09685.<a href="https://arxiv.org/abs/2106.09685?utm_source=chatgpt.com"> https://arxiv.org/abs/2106.09685</a></p><p><strong>QLoRA: Efficient Finetuning of Quantized LLMs</strong> — Dettmers, T., Pagnoni, A., Holtzman, A., &amp; Zettlemoyer, L. (2023). arXiv:2305.14314.<a href="https://arxiv.org/abs/2305.14314?utm_source=chatgpt.com"> https://arxiv.org/abs/2305.14314</a></p><p><strong>Parameter‑Efficient Fine‑Tuning Methods for Pretrained Language Models: A Critical Review and Assessment</strong> — Xu, L., Xie, H., Qin, S.-Z. J., Tao, X., &amp; Wang, F. L. (2023). arXiv:2312.12148.<a href="https://arxiv.org/abs/2312.12148?utm_source=chatgpt.com"> https://arxiv.org/abs/2312.12148</a></p><p><a href="https://youtu.be/Bq9zqTJDsjg?si=eP4h-nWZPXNosQCi">LoRA explained — YouTube</a></p><p><a href="https://youtu.be/HcVtpLAGMXo?si=XjkC2D154GtvRkL0">PEFT explained — YouTube</a></p><p><a href="https://towardsai.net/p/machine-learning/from-quantization-to-inference-beginners-guide-for-practical-fine-tuning-qlora-with-mistral-7b">TowardsAI fine-tuning tutorial</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9292a807fc7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How AI will make you LOVE school — integrating AI-powered PBL to personalize education]]></title>
            <link>https://medium.com/@laura7.schroeder/how-ai-will-make-you-love-school-integrating-ai-powered-pbl-to-personalize-education-02211093ac68?source=rss-63531480f361------2</link>
            <guid isPermaLink="false">https://medium.com/p/02211093ac68</guid>
            <category><![CDATA[personalized-learning]]></category>
            <category><![CDATA[edtech]]></category>
            <category><![CDATA[future-of-education]]></category>
            <category><![CDATA[project-based-learning]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Laura Schroeder]]></dc:creator>
            <pubDate>Sat, 11 Oct 2025 11:13:26 GMT</pubDate>
            <atom:updated>2025-10-13T12:21:27.835Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LRc6dhNg52hYc8-LU99EfA.png" /></figure><h3><strong>How AI will make you LOVE school — integrating AI-powered PBL to personalize education</strong></h3><p>Sounds too good to be true, doesn’t it? Considering the status quo of the education system and the applied teaching methods, I totally agree! Most people are aware of problems in the current education system, such as burned-out teachers, a lack of financing, insufficient digitalization, and distracted students. However, these are just the symptoms; therefore, we need to ask ourselves — what is the root cause of all these problems?</p><p>In a nutshell, schools prepare students that don’t exist, for a world that doesn’t exist, using methods that shouldn’t exist. This mismatch between how schools expect students and the world to be and the actual reality is caused by the widespread approach that the vast majority of education systems in the world use: standardization.</p><p>Schools nowadays are designed around the average student. Who is that!? Right, the average student as such doesn’t exist because each human is incredibly individual and therefore has different needs, especially when it comes to learning. The majority of topics taught in regular classes are irrelevant to most students. Combined with unrealistic and boring teaching methods such as rote memorization, there is no chance of engagement! They become bored and easily distracted because learning without passion and intrinsic motivation isn’t learning. As a result, it becomes even harder for educators to teach in a classroom full of 30 unique (yet disengaged) individuals. What does this leave us with? Burned-out teachers who feel inefficient and students who are unmotivated and disengaged — a vicious circle…</p><p><em>TL;DR — </em><strong><em>We. Have. A. Problem</em>.</strong></p><h3>How Project-Based Learning can break the vicious circle</h3><p>We can keep patching symptoms, or we can rethink education at its foundation. The purpose of education is to understand yourself and the world and how to create your place in it. But this can only be fulfilled through personalized education because every human has his or her own interests, strengths, preferences, and goals, and therefore, their own zone of proximal development. This is where Project-Based Learning (PBL) comes into play.</p><p>At its core, instead of dividing knowledge and phenomena into rigid subjects, and teaching each student the same thing using the same approach, teaching and learning are designed around projects (challenges/phenomena/problems or whatever you want to call it).</p><p>Please don’t confuse PBL with traditional group work at school: in PBL, ideally, it is up to the student what project focus they choose and what they want to create as the outcome. We want collaboration, not competition. We want creation, not consumption. We want understanding, not memorization. The teacher takes the role of a coach to assist learners in their projects and overall endeavours. Students should be driven by curiosity and passion to discover the love for (lifelong) learning and understand the process of building by getting hands-on experience.</p><h4>Why is PBL better than traditional teaching?</h4><p><strong>1) Cross-disciplinary thinking and creativity</strong> — shows the student the intersection of different topics (remember, subjects don’t exist in the real world).</p><p><strong>2) Preparation for adulthood</strong> — teaches skills used in real life and at work, including communication and collaboration, self-organization, problem solving, and the use of AI.</p><p><strong>3) Personalization and fostering of intrinsic motivation</strong> — allows student to choose their focus and create a project according to their interests, strengths, preferences, and goals.</p><p>As a result, students will be more engaged, retain more knowledge that can turn into wisdom and real-life outcomes, and actually enjoy school (yes, <em>enjoy!</em>)</p><h4>What stops schools from integrating PBL?</h4><p><strong>1) Rigid Curriculum → Time constraints</strong></p><p>Teachers are supposed to enable every student to meet the state standards determined by the curriculum (at least from a very pragmatic perspective). The moment the teacher decides to explore a topic more deeply, take the time to help every single learner, or add additional topics, they are behind and face time pressure if they want to cover everything required for that specific class within the school year.</p><p><strong>2) Limited resources</strong></p><p>The implementation indeed requires a lot of time, effort, and resources (digital devices, facilities, materials, and teacher training) — something that many schools lack.</p><p><strong>3) Lack of Tools</strong></p><p>Ed tech tools, such as Quizlet, Duolingo, KhanAcademy, and Kahoot can make teaching and learning less complicated, boring, and time-consuming, however, most of the time, they can only be used for high-level learning or very specific tasks, and are insufficient for the PBL approach; consequently, they only treat the symptoms and not the underlying problem — the lack of personalization.</p><p>This highlights the urgent need for an application that can generate projects from scratch, requiring minimal effort, yet adapting to the unique needs of each teacher and student.</p><p><em>Enter</em> AI-powered PBL</p><h3><strong>A framework for AI-supported Project-Based Learning</strong></h3><p>By introducing the required frameworks for an exemplary AI-powered application that helps integrate PBL by generating such projects, you’ll hopefully better understand how current tools work (more on the industry landscape at the end).</p><p>Such an application would use a Large Language Model (LLM) as its foundation. Essentially, an LLM is a deep neural network model based on a sophisticated mathematical function. They are trained on immense amounts of data, allowing them to perform various tasks. LLMs like GPT are built on a transformer architecture that uses self-attention mechanism to process words simultaneously by assigning weights between tokens based on their contextual relevance. This enables the model to capture relationships and meaning across the entire input. I can highly recommend checking out these videos if you want to learn more about LLMs, transformers, and neural networks.</p><p><a href="https://youtu.be/aircAruvnKk?si=qAY4O6Aecr72KdJc">But what is a neural network? | Deep learning chapter 1</a></p><p><a href="https://youtu.be/LPZh9BOjkQs?si=pKS1O46iK0qXVoNj">Large Language Models explained briefly</a></p><p><a href="https://youtu.be/wjZofJX0v4M?si=6zN_JYxP_2aSYTAR">Transformers, the tech behind LLMs | Deep Learning Chapter 5</a></p><h3><strong>1) User input and processing</strong></h3><p>The input can consist of two things: the prompt and the matching criteria/filters.</p><h4><strong>Prompt generation</strong></h4><p>Users (who can be either the teacher, the student, or both) generate a prompt in which they describe the project, including aspects like the general idea, specific topics, goals, the structure, and setup. In order to get better outputs, the user should know at least some basic prompt engineering skills.</p><h4><strong>Constraint management</strong></h4><p>This is accompanied by certain constraints determined by the user, which can be seen as filters.</p><p><strong>Examples of constraints:</strong></p><ul><li>Grade level</li><li>Level of proficiency (beginner, intermediate, advanced, pro)</li><li>Available resources (3D printer, camera, chemicals)</li><li>Length of project/time available (e.g., 2 weeks, 5 lessons, 3 hours)</li><li>Type of media (video, text, image, audio)</li></ul><p>Constraint management enforces limits so outputs meet predefined requirements. These guardrails increase efficiency, accuracy, and personalization. In constraint management, there are two types:</p><ul><li><strong>hard constraint</strong>s → must always be satisfied, no violations allowed</li><li><strong>soft constraints</strong> → preferences, should be satisfied if possible, but aren’t mandatory</li></ul><h4><strong>Natural language processing</strong></h4><p>NLP is a subfield of AI intersecting with linguistics and enabling machines to understand, interpret, generate, and interact with human language. The goal is to make human language accessible to machines while preserving the meaning and context of the input. NLP can be subdivided into two parts: Natural Language Understanding (NLU) and Natural Language Generation (NLG).</p><p>For now, let’s focus on NLU. How can an AI understand human language?</p><p><strong>1) Tokenization</strong></p><p>Tokens are the smallest units of text (e.g., a short word, a syllable, or punctuation) → token string. Since the computer can only process numbers, each token is assigned a token ID (a sequence of numbers). If a transformer architecture is used, each token gets assigned a weight that internalizes the context of the token compared with any other token → self-attention.</p><p><strong>2) Part of Speech Tagging</strong></p><p>Labelling each token with a grammatical role (e.g., noun, adjective).</p><p><strong>3) Syntax Parsing</strong></p><p>Determining grammatical relationships between tokens ( → supports semantic understanding).</p><p><strong>4) Named Entity Recognition (NER)</strong></p><p>Contextualize and classify data (text prompt) through identifying named entities (e.g., Apple Inc., September, Kool-Aid) → This is especially useful in the educational context since it helps extract and identify key information (historical figures, events, organizations).</p><h4><strong>What are embeddings and how does it work?</strong></h4><p>Each token ID is converted into a vector. A vector is basically a list of numbers, each number representing a dimension. Each number stands for certain attributes, although not directly interpretable as specific human-readable attributes like “color” or “size” but rather captures complex, latent semantic features learned during training, such as syntactic roles or contextual nuances.</p><p>There are two types of vectors:</p><p><strong>1) Sparse vectors</strong></p><ul><li>Most values are zero</li><li>each dimension represents a specific feature like a word → high-dimensional (we aren’t talking about x,y,z-direction, more like hundreds or even thousands of dimensions)</li><li>don’t capture deeper semantic meaning</li></ul><p><strong>2) Dense vectors</strong></p><ul><li>Most values are non-zero</li><li>each dimension contains learned features that together represent the meaning of the input → low-dimensional</li><li>capture semantic meaning</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a6HdEJVmemP3tj1eIRC-cA.png" /><figcaption>A token is assigned a token ID. The token ID then gets converted into an embedded vector represented by a list of numbers.</figcaption></figure><p>These vectors are embedded into a high- or low-dimensional space. This is called embedding. Since each number represents a dimension, tokens with similar numbers (=attributes) are close to each other (geometric proximity). After training, the embedded (dense) vectors can capture semantic similarity. So remember: geometric proximity = semantic similarity.This allows the AI model to capture the context and meaning of text → Natural Language Understanding</p><p>We have successfully converted a text input into an embedded vector, enabling the machine to process the input subsequently.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x_xXFXZDRfTlXOmh6UjSTQ.png" /><figcaption>This figure illustrates vectors as points in a three-dimensional space. Each point is part of a superordinate topic. It shows that related ideas are closer together, forming clusters.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9ViPlXLwqGZN-22n-7qR9A.png" /><figcaption>An illustration of a continuous multidimensional vector space representing individual sentences. Each sentence is represented as a dot. Pairs of related sentences are connected with arrows. This shows how sentences with similar words and semantic meaning are closely embedded in the vector space.</figcaption></figure><p>Besides embedding tokens, which is called contextual embedding, there are two other embedding types that I would like to mention (for now) because they are useful as well for a possible AI-powered PBL platform.</p><p><strong>1) Sentence embedding</strong></p><p>Those embeddings can compress whole sentences into one vector. A possible use case might be when a user asks a question, which then gets embedded as a whole vector and compared to other vectors (=previously asked questions). If the question has already been asked by someone else, the AI model can just reuse the previously generated answer or project. Thereby, sentence embeddings can increase efficiency because the model doesn’t have to generate an answer or project from scratch again.</p><p><strong>2) Multimodal embedding</strong></p><p>Multimodal embeddings can create shared vector spaces, combining different modes, such as text, image, or audio. This is necessary to process various types of media. In our case, multimodal embeddings allow the LLM to process multimodal input and generate multimodal output.</p><h4><strong>What are Knowledge Graphs and how can they assist the LLM?</strong></h4><p>We have all experienced or at least heard about LLMs hallucinating and generating inaccurate or unverifiable answers, lacking accuracy and traceable sources. These shortcomings become more significant in practical contexts, where LLMs support critical operations within organizations, or, as in our case, for educational purposes. LLMs that don’t use Knowledge Graphs (KG) might produce educational content that includes errors or lacks specificity, undermining trust and learning outcomes. To mitigate these risks and make LLMs suitable for school environments, LLMs need to be grounded in knowledge graphs.</p><p>A knowledge graph is a structured (semantic) representation of knowledge. This knowledge consists of entities connected by relationships organized in a graph database. The entities equal the nodes connected by edges, equaling the relationships. You can think of it as a big, complex network laid out like a map. There are various types of KGs focusing on different entities and their corresponding relationships, but they can all be combined in one giant map of possible projects, skills, topics, and resources.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/770/1*seO7k91FNdVjqIf6URym2Q.png" /><figcaption>This table shows examples of KG types and how one element could look.</figcaption></figure><p>Each student has their unique position on the map (unique prerequisites, interests, skills), which allows the model to adapt to each student individually by having a structured and interconnected database of each user. A KG allows the LLM to check logical dependencies instead of mere statistical probabilities, meaning semantic similarity search using contextual vector embeddings. KGs store explicit relationships, thereby enhancing the explainability of the output and increasing transparency in turn, which is essential for the sake of teaching and learning. Furthermore, due to the structured representation of knowledge, KGs enforce constraint management by allowing the system to efficiently validate whether certain criteria are met.</p><p>Formally, a KG is represented as G, defined by a set of triples (head, relation, and tail) → G= (h,r,t) | h,t E,r R, where E, R denote the set of entities and relations, respectively, as shown in the example:</p><p>(Project: Photosynthesis Experiment) → teaches → (Concept: Cellular Energy Conversion)</p><p>A KG based on triples is called a Semantic Web Graph. Such graphs are useful to reason about relationships (→ logical inferences). In our case, so-called Property Graphs should complement SWG since they store knowledge in a more pragmatic way. Nodes can have labels and properties, and edges can have properties as well, making it suitable for recommendation systems (like suggesting projects) and LLM or RAG integration. So instead of asking “How do these things connect?” (SWG), it asks “What do these connections mean?” (PG). A hybrid solution using both KG types would be most suitable for a PBL application because it would allow both logical reasoning and storing dynamic relationships, and support AI integration.</p><p>At this point, after the prompt has been sent, processed using NLP, and having KGs as a structured foundation (note that KG is at the intersection of input processing and project generation), the next step comes into play.</p><h3><strong>2) Project generation</strong></h3><h4><strong>How is information retrieved from a KG?</strong></h4><p><strong>1) Semantic Similarity Search</strong></p><p>This approach works by embedding KG into a high-dimensional vector space. These embeddings are called Knowledge Graph Embeddings. Two nodes that are close in the embedding space are semantically similar, allowing an LLM to search for, e.g., projects evolving around related concepts.</p><p><strong>2) Reasoning</strong></p><p>By applying rules, hierarchies, or transitive relationships, logical reasoning leads to facts that aren’t directly stored. Then constraint reasoning is used to ensure accuracy and relevance.</p><p>The entire previous section about KG is about augmenting the LLM. Besides KG, there are other approaches to address the limitations of pure generative models. Let’s dive into it!</p><h4><strong>What is Retrieval-Augmented-Generation?</strong></h4><p>Retrieval-Augmented-Generation (RAG) is a hybrid model architecture that combines NLG with external databases using retrieval mechanisms to enhance the model’s knowledge database. Unlike traditional (LLMs), which depend solely on pre-trained internal knowledge, RAG integrates external sources to improve accuracy and reliability by searching for specific information and using trustworthy sources, respectively.</p><p>RAG models consist of two key components: the retriever, pulling specific information from external databases, and the generator, synthesizing the retrieved data into coherent and contextually relevant responses. It’s important to understand that RAG doesn’t just pull information from the internet by doing a quick Google search; instead, the information has to be processed first.</p><h4><strong>Indexing — What happens before RAG?</strong></h4><p>Initially, external data (NASA database, Wikipedia, textbooks, etc.) needs to be fetched. After the data has been ingested (manually or using an AI), it is chunked, making it easier to handle big data sets. Then the information is embedded and stored as an embedded vector in a multidimensional vector database. This process is called Indexing.</p><h4><strong>Retrieval — How is the information retrieved?</strong></h4><p>For an LLM to access the fetched and embedded information, retriever mechanisms are needed. Let’s take a look at the most commonly used retriever mechanisms:</p><p><strong>1) TF-IDF (Term Frequency — Inverse Document Frequency)</strong></p><ul><li>It looks at how often each word appears in a document (term frequency) using sparse vectors</li><li>Then it reduces the importance of very common words like “the” or “and” (inverse document frequency), since these appear everywhere and aren’t useful for distinguishing documents</li><li>The result is a score for each document based on how well its words match the query (with rare words weighted more heavily)</li></ul><p><strong>2) BM25 (Best Matching 25)</strong></p><ul><li>An improved version of TF-IDF → prevents bias towards long documents</li><li>Once a word appears enough times, more appearances add less value → just because a document is longer and contains a term more frequently, it isn’t more valuable per se</li></ul><p>→ used due to simplicity and efficiency, yet, BM25 and TF-IDF can’t capture semantic meaning</p><p><strong>3) Dense Passage Retrieval (DPR)</strong></p><ul><li>DPR allows for semantic search that goes beyond mere keyword matching by encoding both queries and documents into dense vectors</li><li>This approach uses a bi-coder architecture where the query and documents are encoded separately</li></ul><p>→DPR outperforms traditional retrieval models like the ones above.</p><p><strong>4) REALM (Retrieval-Augmented Language Model)</strong></p><ul><li>In the training phase, REALM focuses on the generation task as well by updating both the retriever and the generator</li><li>This helps improve the accuracy and coherence of generated responses</li></ul><p>→ can outperform other retrieval mechanisms in knowledge-intensive tasks by generating more accurate and coherent answers thanks to its end-to-end training</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kxzNLG3D-dBEIkZip6q3-Q.jpeg" /><figcaption>Here we can see the workflow of RAG. The user prompts a query, which is analyzed through NLP. The retriever mechanism pulls relevant knowledge from external databases, consisting of fetched and polished information from external sources. The retrieved chunks are synthesized into the LLM’s response.</figcaption></figure><h4><strong>Generation — How is the LLM augmented by the retrieved information?</strong></h4><p>After the retriever pulls the necessary information from the external corpus, the generator synthesizes this information. In this case, the LLM acts as a pillar or framework to make sure that the produced answer is accurate, coherent, and fits the original query. Let’s explore the most commonly used generator mechanisms:</p><p><strong>1) T5 (Text-to-Text Transfer Transformer)</strong></p><ul><li>Often paired with DPR</li><li>Takes the query and retrieved documents as a single input text sequence</li><li>All tasks framed as a text-to-text transformation → high flexibility</li></ul><p><strong>2) BART (Bidirectional and Auto-Regressive Transformer)</strong></p><ul><li>Often paired with DPR</li><li>Encodes the query and retrieved documents together into a contextual representation (to understand their meaning before generating an answer)</li><li>Good at handling noisy input (typos, irrelevant words, unstructured queries)</li></ul><p><strong>3) GPT (Generative Pre-trained Transformer)</strong></p><ul><li>Treats the retrieved information as part of the query</li><li>Answers using transformer architecture (basically an LLM, but can be seen as a generator mechanism in the context of RAG)</li></ul><p>In fact, RAG can not only be used for text-based retrieval and generation, but also works across different modalities. Audio-based or video-based RAG models can incorporate both visual and textual information, allowing for applications such as speech recognition, video captioning, or conversational agents. Tying this to a possible application to integrate PBL at schools, such multimodal RAG models can be very useful since they can increase flexibility. A student asking for a project about renewable energy receives a complete multimodal package including articles, images, data charts, and explanatory videos to guide their work.</p><p>In a nutshell, by using RAG, task-specific and up-to-date knowledge can be incorporated, thereby increasing accuracy and contextual relevance, making the application suitable for educational use. Combined with KG, the model can capture complex relationships, which increases explainability and efficiency. In addition, KGs can simplify reasoning for the model due to its structured representation of knowledge. Still, there is more to the story.</p><p>The previously explained methods focus on providing the LLM with knowledge — but knowledge without thought is just data.</p><h4><strong>Adding brain to the facts — Reasoning Strategies</strong></h4><p>To help the LLM improve its reasoning abilities, a common approach is to generate intermediate reasoning steps. This framework enables models to break down complex problems into more digestible steps using reasoning strategies, such as Chain of Thought, Tree of Thoughts, or Graph of Thought.</p><p><strong>1) Chain of Thought (CoT)</strong></p><ul><li>Uses a linear reasoning strategy → breaks down the task into logical steps</li><li>One step leads to the other, allowing the model to gradually build toward the answer</li><li>Example: math calculations</li></ul><p><strong>2) Tree of Thoughts (ToT)</strong></p><ul><li>Uses a branching reasoning strategy → explores multiple paths at the same time</li><li>Chooses the best solution for each path, repeats the process to arrive at one best solution</li><li>Example: trip planning</li></ul><p><strong>3) Graph of Thought (GoT)</strong></p><ul><li>Uses a networked reasoning strategy → model can explore how paths are interconnected and combine solutions</li><li>Example: project design</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FV0Apq0PKWUB5uyW6POdWA.jpeg" /><figcaption>These schemas illustrate different LLM reasoning strategies. The input is processed by going from thought to thought. Each thought consists of its own thinking processes and search methods. Depending on the method, the thoughts can influence each other.</figcaption></figure><p>These reasoning strategies allow the model to reason and can be seen as thinking methods on a metalevel. They help with problem-solving, consistency, and explainability, but it is important to mention that they aren’t impacting the architectural design of the model. If we want to change how the model “thinks” on an architectural level, the Mixture of Experts (MoE) approach can be helpful. This would increase accuracy as well, not through reasoning, but thanks to specialization. On top of that, the LLM would be more efficient. Let me explain.</p><h4><strong>How Mixture of Experts can improve the accuracy and efficiency of LLMs</strong></h4><p>Mixture of Experts (MoE) is an architecture where the model is split into a number of different sub-models, so-called experts. Usually, these experts are simple feed-forward neural networks (one-directional flow: input → hidden layer → output) and are specialized in specific areas, for example, history or STEM. The key difference compared to other machine learning approaches is that the query is not processed by the whole network, but by selected experts instead. The process of choosing which expert to activate is done by the gating network. The outputs of the experts are then weighted and added together. This forms the final output of one layer and is fed into the next layer as input.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qwncycOgOGVONfOmnMy4DA.jpeg" /><figcaption>This image shows the basic architecture of a Mixture of Experts (MoE) layer. The gating network selects two experts. The gating function then weights and connects the outputs of the experts representing the input of the following layer.</figcaption></figure><p>The fact that only some parts of the model are used makes processing more efficient. Another benefit that comes from the architectural design is that MoE allows for specialization, consequently, improving the accuracy of the output. However, to ensure that the gating network transfers the input to suitable experts, the routing mechanism must be well-trained, which can be quite complex and time-intensive. Additionally, if experts are used unevenly, this might result in an imbalanced workload, causing some experts to be undertrained or underutilized, which can undermine accuracy. Nevertheless, as a whole, it is reasonable to consider using the Mixture of Experts method for an AI-powered PBL platform due to its contribution to accuracy and efficiency.</p><p>At this point, the AI has successfully created a project that adapts to the learners’ and teachers’ needs.</p><h3>3) Project execution and review</h3><p>I won’t go into a lot of detail here since the focus is on generating the project. Yet, I would like to mention the importance of human-in-the-loop and constraint management (yes, here we go again ;)). Using constraint management, it can be checked whether all the previously determined criteria are met. To increase accuracy and the degree of personalization even more, it is essential that a human checks the output as well.</p><h4><strong>Human-in-the-loop (HITL)</strong></h4><p>Human-in-the-loop describes the interaction of models and humans, where humans give input used by the system to improve or constrain automated behavior and output, forming a closed feedback loop. Generally, humans can be incorporated in various stages, such as data collection and labelling, or training supervision. If you are using ChatGPT, you have probably been asked at least once, “Which response do you prefer?”. This allows the model to get feedback from an actual human user → Human-in-the-loop.</p><p>In our case, <strong>HITL integration</strong> can look as follows:</p><ul><li>Applying filters (constraint management)</li><li>The teacher approves and selects the generated project</li><li>Micro-feedback through the learner (like in the ChatGPT example)</li></ul><p>Implementing HITL helps make the content more tailored to students and can increase overall quality and alignment.</p><p>So far, so good — that is the theory. But what is actually out there on the market?</p><h3><strong>What does the industry landscape look like?</strong></h3><p>Integrating online resources and AI in education has been a widely known practice. The majority of students use ChatGPT, YouTube videos, or KhanAcademy for schoolwork and studying. After researching, I came across a couple of platforms that offer live AI-tutors, such as KhanAcademy, Flint, or SchoolAI. Some platforms, such as MagicSchool and Kira, focus on making a teacher’s life easier by providing AI tools for planning and automating tasks. Others use AI to gamify learning with the goal of engaging children, especially K-5.</p><p>While they are all very helpful, there aren’t a lot of companies and platforms that use AI to power PBL specifically. Still, I would like to give an overview of a couple of companies that are already doing some great work in this area.</p><h4><strong>MasteryMate by Alayna AI</strong></h4><p>MasteryMate is a teacher-facing platform that offers AI-generated, interdisciplinary “project recipes”. Educators can get inspired by accessing the library, showcasing hundreds of pre-designed recipes. To create one’s own, the teacher can select standards or objectives, grade level, subjects, and the available time for the project. Once generated, the recipes are customizable and can be complemented by rubrics or reflection prompts. It is very adaptable and easy to navigate. However, it is less suited for multimodal content generation. The HITL editing layer, supported by constraint management techniques, enables tailored and guided project generation.</p><ul><li><a href="https://masterymate.ai/">Click here</a> to check out their <strong>website</strong></li><li><a href="https://youtu.be/ucLRnu82OpE?si=Mrou6jTxQ_VGpPLL">Click here</a> to check out their <strong>demo</strong></li></ul><h4><strong>SmartPrep.AI</strong></h4><p>SmartPrep.AI offers AI-generated lesson bundles for teachers, which can be designed around PBL. This platform also allows for criteria selection (topic, class, grade level) and generates topic outlines based on the input. The teacher can select the teaching style, such as PBL, flipped classroom, or gamification. For PBL, it creates a project outline and can even provide you with multimodal resources, like videos or images, which indicates that SmartPrep.AI is using some form of RAG. In addition, various assessment types, e.g., MCQ, FRQ, or essay writing, can be created to further support teachers.</p><ul><li><strong>Pricing</strong>: $9–15/month</li><li><a href="https://smartprep.ai/">Click here</a> to check out their <strong>website</strong></li><li><a href="https://youtu.be/74sMluHdeUE">Click here</a> to check out their <strong>demo</strong></li></ul><h4><strong>PBLplans</strong></h4><p>PBLplans provides an AI-powered PBL lesson plan generator that allows teachers to quickly create a project outline with an emphasis on the 17 United Nations’ SDGs (Sustainable Development Goals). The interface is simple and intuitive. Educators can select filters such as grade level, subject, class size, and associated SDG, and add additional project details or topics. Based on this input, the system generates a tailored PBL lesson plan that teachers can further refine by adjusting the prompt or outline to ensure alignment with learning objectives, and, importantly, sustainability.</p><ul><li><strong>Pricing</strong>: $5/month/teacher (additional pricing models available)</li><li><a href="https://pblplans.com/">Click here</a> to check out their <strong>website</strong></li><li><a href="https://youtu.be/J03ZNybNNA4?si=XRK61HP2DBrgdXBT">Click here</a> to check out their <strong>demo</strong></li></ul><h4><strong>Oboe</strong></h4><p>Oboe is a tool geared more toward learners or anyone wanting to build a short course around a topic. Based on the user’s prompt, Oboe creates a mini-course categorized into three learning types or stages, namely listening, reading, and interacting. It generates podcasts, informative texts, MCQs, or flashcards in seconds, while the UI is held simple and prioritizes textual clarity. With the upgraded plan, users can browse existing courses and upload media like images for the prompt. This platform provides fewer features concerning the course structure and content compared to the other platforms, therefore making it more suitable for individuals and self-directed learning.</p><ul><li><strong>Pricing</strong>: $15–40/month</li><li><a href="https://oboe.fyi/">Click here</a> to check out their <strong>website</strong></li><li><a href="https://youtu.be/zc0pk11ldm4?si=GbkOz3r179p8mN0O">Click here</a> to check out their <strong>demo </strong>(third-party review)</li></ul><p>If a teacher or school decides to implement PBL or similar approaches in the classroom, these companies offer great support. They illustrate how AI can catalyze the implementation of PBL in the classroom by supporting teachers and students in the process of coming up with an idea for the project, creating an understandable and flexible course outline complemented by fitting, multimodal resources, and, concerning teachers, reviewing the students’ work.</p><p>Nonetheless, regarding the ideal classroom experience in terms of PBL, as I outlined at the beginning, there is still much progress to be made — not only in terms of technological innovation and implementation, but also in the underlying structure and conditions of schools, which require improvement as well.</p><h3><strong>How to navigate the challenges and risks</strong></h3><p>I’d be wishful thinking (at least for now…) to assume that students or teachers just type in a prompt and have the outline for the next days and weeks of class. If we are being realistic, there are certain challenges and risks that come with PBL implementation in learning environments. Some of them address the platform builders, and some concern schools and the educational system overall.</p><h4><strong>Superficial Projects</strong></h4><p>The majority of platforms generate projects that lack depth and detail. They may be useful for overviews or introductory courses, but they often fall short when low-level content is required to foster a deep understanding.</p><p>=&gt; This problem could be addressed by “outsourcing” knowledge acquisition to external platforms, such as Khan Academy, YouTube, little classroom lectures, and podcasts, although this is not a fully satisfying solution. A better approach is to create domain-specific AI for each collection of topics, consequently, being more knowledgeable and accurate. By integrating sophisticated and domain-specifically trained LLMs, the user can ask any questions, no matter how deep, detailed, and complex, and is provided with a fitting response.</p><h4><strong>Accuracy</strong></h4><p>Although AI is correct most of the time, it can still get things shockingly wrong or at least not as accurate as needed and produce hallucinations. Especially in education, this should be reduced as much as possible because learners shouldn’t be exposed to incorrect or misleading information.</p><p>=&gt; By grounding LLMs in KG and using RAG, MoE architecture, and HITL, PBL platforms can increase the overall accuracy of its output, making the application more reliable and trustworthy (so that even the biggest AI-skeptics might consider using it ;))</p><h4><strong>Privacy and data protection</strong></h4><p>This perhaps remains one of the biggest issues among educators and parents — what happens to all the data? This concern is more than understandable. Often, companies lack transparency regarding privacy and data governance, making it difficult to understand what happens to the data.</p><p>=&gt; The tool providers must prioritize privacy, data protection, and transparency to the best of their abilities. This includes educating the user about AI safety, like not sharing personal information, and promising not to share user data with third parties or monetize it in other ways. Users should be able to access their stored data at any time and delete it when needed.</p><h4><strong>Accessibility and AI literacy</strong></h4><p>Another issue concerns the implementation of such platforms. Many schools, particularly those with limited funding and technological infrastructure, have difficulties affording such tools. Even when tools are available, a lack of AI literacy among educators can hinder effective implementation.</p><p>=&gt; Evidently, if the platform builders lower the prices and design intuitive and simple UIs, the problem of accessibility could already be mitigated to some extent. Still, governmental funding and teacher training initiatives are essential to implement PBL successfully.</p><h4><strong>Transition for Students</strong></h4><p>Evidently, not only teachers, but students as well will have to adapt, although I believe that most students would be more than willing to change the classroom experience and integrate new technology (okay…I might be a bit biased). Especially students who haven’t developed proficient time and self-management skills could face difficulties.</p><p>=&gt; Schools should focus on teaching essential skills such as self-directed learning, time-management, problem-solving, and AI literacy. This will not only be helpful for PBL, but for life as well. In particular, during transition and in the early stages of establishing PBL, educators with special training can serve as advisors and assist students who struggle with new learning methods and self-organization.</p><h4><strong>Curriculum constraints</strong></h4><p>Integrating AI-powered PBL into existing curricula can be difficult. Excessive standardization constrains many schools and teachers, as I have explained initially. There is little to no room for experimentation and personalization.</p><p>=&gt; By loosening obsolete and rigid standards and test requirements, there would be more room for innovative and student-centered approaches like PBL. What it all boils down to is the importance of trusting teachers and students in what they are doing. Unfortunately, since this issue isn’t a technical but rather a systemic one, it will take enormous effort to change traditional structures.</p><h3><strong>Conclusion and future outlook</strong></h3><p>The biggest problem in today’s education system lies in the widespread standardization and rigid curriculum requirements. We need to make education more personalized so that each student can discover their passion, learn real-life skills, and unlock their potential to live a happier, healthier, and more fulfilled life. By leveraging AI, teachers and students can generate project ideas and outlines accompanied by helpful resources within seconds. To scale such AI tools, companies must ensure the following:</p><ul><li><strong>Accuracy </strong>— provided information must be reliable and free from hallucinations.</li><li><strong>Adaptability </strong>— projects should flexibly adjust to diverse classrooms and users.</li><li><strong>Accessibility </strong>— tools need to be affordable and easy to use.</li></ul><p>Several companies and organizations are already working on ideas and tools to make education more personalized, therefore more relevant and engaging — paving the way for a student-centered school system and an engaging and transformative classroom experience.</p><p>Take a second and imagine a school full of energized, healthy, and curious students, each driven by their passions and the desire to understand themselves and the world. Think about the impact that these children could have with their skills, projects, and ideas!</p><p>This is not fantasy. With innovative design and the power of AI, we can create an education system that enables students to unlock their potential, live with purpose and passion, and have a positive impact on our world!</p><h4>About me</h4><p>I’m Laura Schroeder, a high school student and Innovator at The Knowledge Society (TKS), driven by curiosity, a desire to make meaningful impact, and the belief that every single person has the potential to accomplish incredible things! My deepest passions lie in education and health — two fundamental pillars of life that are deeply interconnected. By combining the insights of the humanities with the power of AI, we can revolutionize both fields, raise living standards, transform the way we live and advance humanity as a whole. I’m always eager to gain new insights, connect with and learn from others, and expand my perspectives — with the ultimate goal to understand myself and the world.</p><h4>Let’s connect on <a href="http://www.linkedin.com/in/%20laura-schroeder07">LinkedIn</a></h4><h4>Check out my <a href="https://lauraschroeder.lovable.app/">personal website</a></h4><h4><strong>Sources:</strong></h4><p><strong>Shazeer, N., Mirhoseini, A., Maziarz, K., Davis, A., Le, Q., Hinton, G., &amp; Dean, J. (2017).</strong> Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer. <em>arXiv</em>.<a href="https://arxiv.org/abs/1701.06538"> https://arxiv.org/abs/1701.06538</a></p><p><strong>Zhang, Y., Li, X., Wang, S., &amp; Chen, Y. (2025).</strong> Retrieval-Augmented Generation for Educational Application: A Systematic Survey. <em>Computers &amp; Education: Artificial Intelligence</em>.<a href="https://www.sciencedirect.com/science/article/pii/S2666920X25000578?via%3Dihub">https://www.sciencedirect.com/science/article/pii/S2666920X25000578?via%3Dihub</a></p><p><strong>Li, X., Zhang, Y., &amp; Wang, S. (2024).</strong> A Comprehensive Survey of Retrieval-Augmented Generation (RAG): Evolution, Current Landscape, and Future Directions. <em>arXiv</em>.<a href="https://arxiv.org/abs/2410.12837?utm_source=chatgpt.com"> </a><a href="https://arxiv.org/abs/2410.12837">https://arxiv.org/abs/2410.12837</a></p><p><strong>Amayuelas, A., Sain, J., Kaur, S., &amp; Smiley, C. (2025).</strong> Grounding LLM Reasoning with Knowledge Graphs. <em>arXiv.</em><a href="https://arxiv.org/abs/2502.13247?utm_source=chatgpt.com"> https://arxiv.org/abs/2502.13247</a></p><p><strong>Mosqueira-Rey, E., Hernández-Pereira, E., Alonso-Ríos, D., Bobes-Bascarán, J., &amp; Fernández-Leal, Á. (2022).</strong> Human-in-the-Loop Machine Learning: A State of the Art. <em>Artificial Intelligence Review.</em> <a href="https://doi.org/10.1007/s10462-022-10246-w">https://doi.org/10.1007/s10462-022-10246-w</a></p><p>IBM. (n.d.). <em>What is a knowledge graph?</em> IBM Think. Retrieved October 9, 2025, from<a href="https://www.ibm.com/think/topics/knowledge-graph?utm_source=chatgpt.com"> https://www.ibm.com/think/topics/knowledge-graph</a><a href="https://www.ibm.com/think/topics/knowledge-graph"> ibm.c</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=02211093ac68" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>