<?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 Efe MISIRLI on Medium]]></title>
        <description><![CDATA[Stories by Efe MISIRLI on Medium]]></description>
        <link>https://medium.com/@efemisirli?source=rss-ed5641a2b111------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*PmX8YYuiJBEwpkHd6FKZJg.jpeg</url>
            <title>Stories by Efe MISIRLI on Medium</title>
            <link>https://medium.com/@efemisirli?source=rss-ed5641a2b111------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 07:26:52 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@efemisirli/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[Okullarda Yapay Zekâ Yasaklanmalı mı?]]></title>
            <link>https://medium.com/t%C3%BCrkiye/okullarda-yapay-zek%C3%A2-yasaklanmal%C4%B1-m%C4%B1-22b7aac5b152?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/22b7aac5b152</guid>
            <category><![CDATA[eğitim]]></category>
            <category><![CDATA[yapay-zeka]]></category>
            <category><![CDATA[eğitimde-yapay-zeka]]></category>
            <category><![CDATA[türkçe]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Sun, 28 Jun 2026 20:49:31 GMT</pubDate>
            <atom:updated>2026-06-28T20:49:31.991Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5WqkoV-jvIPrJgkMYG-y2A.png" /></figure><p>Bugün bir öğretmen bir kompozisyon ödevine baktığında artık sadece “Bu öğrenci ne anlatmak istiyor?” diye sormuyor. Çoğu zaman zihninin bir köşesinde ikinci bir soru daha beliriyor: “Bunu gerçekten öğrenci mi yazdı?”</p><p>Bu soru haksız değil. Hatta son derece anlaşılır.</p><p>Çünkü yapay zekâ araçları artık birkaç saniye içinde düzgün paragraflar kurabiliyor, ödev formatına uygun metinler üretebiliyor, matematik sorularını çözebiliyor, özet çıkarabiliyor, sunum hazırlayabiliyor. <strong>Öğrencinin kendi emeğiyle yaptığı çalışma ile bir yapay zekâ aracından aldığı destek arasındaki sınır öğretmen açısından giderek bulanıklaşıyor.</strong></p><p>Bu yüzden kimileri tarafından verilen ilk tepki şu oluyor: “O halde yasaklayalım.”</p><p>Bu tepkinin arkasında kötü niyet yok. <strong>Tam tersine, çoğu zaman öğrenciyi koruma isteği var.</strong> Öğrenci düşünmeden cevap almasın. Yazmayı öğrenmeden metin üretmesin. Emek vermeden başarı görüntüsü elde etmesin. Okulda ölçme-değerlendirme adaleti bozulmasın. Öğretmenin verdiği ödev anlamını kaybetmesin.</p><p>Bunlar son derece gerçek kaygılar. Hafife alınamaz.</p><p>Ama yine de daha derin bir soru sormamız gerekiyor: <strong>Yapay zekâyı okuldan kovmak gerçekten öğrenciyi korur mu? Yoksa onu, gelecekte zaten içinde yaşayacağı bir dünyaya hazırlıksız mı bırakır</strong>?</p><h3>Bu kaygı boşuna değil</h3><p>Önce şunu açıkça kabul etmek gerekir: <strong>yapay zekânın bazı kullanımları öğrencinin öğrenme sürecini zayıflatabilir.</strong></p><p>Bir öğrenci kendi düşüncesini geliştirmeden, yalnızca yapay zekâya “bana bu konuda kompozisyon yaz” derse ve çıkan metni olduğu gibi teslim ederse, burada gerçek bir öğrenme sorunu vardır. Öğrenci düşünmemiştir, seçmemiştir, tartmamıştır, cümle kurma zahmetine girmemiştir. Sadece bir çıktı almıştır.</p><p>Bu durumda öğretmenin kaygısı son derece yerindedir. Çünkü eğitimin amacı yalnızca ortaya düzgün görünen bir ürün çıkarmak değildir. Eğitim, o ürüne giden süreçte öğrencinin zihninin çalışmasıdır. Öğrencinin zorlanması, fikrini toparlaması, yanlış yapması, düzeltmesi, bir cümleyi kurmak için uğraşmasıdır.</p><p>Eğer yapay zekâ öğrencinin bu süreci tamamen devrettiği bir araca dönüşürse, burada elbette sorun vardır.</p><p>Fakat buradan doğrudan “o halde yasaklamalıyız” sonucuna atlamak fazla hızlı bir karar olur. Çünkü <strong>aynı araç, farklı bir kullanım biçimiyle öğrencinin düşünmesini de destekleyebilir.</strong></p><p>Örneğin bir öğrenci önce kendi kısa taslağını yazar. Sonra yapay zekâya “Bu metindeki en zayıf üç noktayı söyle” der. Ardından çıkan eleştirilere bakar, hangilerine katıldığını açıklar, metnini yeniden düzenler. Bu durumda yapay zekâ onun yerine düşünmemiş olur; öğrenciyi kendi düşüncesi üzerine yeniden düşündürmüş olur.</p><p>Demek ki mesele yalnızca aracın varlığı değil. Mesele, aracın hangi amaçla ve nasıl kullanıldığıdır.</p><h3>Yapay zekâ çıkmadan önce eğitimimiz güllük gülistanlık mıydı?</h3><p>Yapay zekâ tartışmalarında bazen şöyle bir hava oluşuyor: Sanki yapay zekâ gelmeden önce okullarda her şey yolundaydı; öğrenciler derin düşünüyor, güçlü yazıyor, severek okuyor, merakla araştırıyor, ödevlerine anlamlı emek veriyordu. <strong>Sonra yapay zekâ çıktı ve bu harika düzen bozuldu.</strong></p><p>Gerçekten böyle miydi?</p><p>Genel düzeyde eğitim sistemimizin zaten uzun süredir ciddi sorunlarla boğuştuğunu biliyoruz. Öğrencilerin okuma alışkanlıkları, yazılı ifade becerileri, derse ilgileri, sınav odaklı öğrenme kültürü, ezberle anlama arasındaki gerilim, ödevlerin gerçek öğrenmeye dönüşüp dönüşmediği gibi konular yapay zekâdan önce de tartışılıyordu.</p><p><strong>Yapay zekâ bu sorunları başlatmadı. Daha çok, var olan sorunları daha görünür hale getirdi.</strong></p><p>Bir öğrenci kompozisyon yazmak istemiyorsa, bu isteksizlik yapay zekâ ile başlamadı. <strong>Bir öğrenci ödevi yalnızca teslim edilmesi gereken bir yük olarak görüyorsa, bu duygu yapay zekâ çıkmadan önce de vardı.</strong> Bir öğrenci bilgiyi anlamak yerine en kısa yoldan doğru cevaba ulaşmaya çalışıyorsa, bu da yeni bir sorun değil.</p><p>Yapay zekâ burada bir büyüteç gibi çalışıyor. Eğitimde zaten var olan düşünme, yazma, merak ve emek krizini büyütüyor; onu hızlandırıyor, daha görünür hale getiriyor.</p><p><strong>Bu nedenle çözüm, sanki eski düzen kusursuzmuş gibi yapay zekâyı dışarı atmak olamaz.</strong> Asıl soru şu olmalı: Yapay zekâ çağında öğrencinin gerçekten düşünmesini, yazmasını, sorgulamasını ve emek vermesini nasıl tasarlarız?</p><h3>Okulun kapısından çıkınca yapay zekâ bitmiyor</h3><p>Yasak fikrinin bir başka sorunu da pratik gerçekliktir.</p><p>Bir okul, okul içinde yapay zekâ kullanımını yasaklayabilir. Bazı sitelere erişimi engelleyebilir. Öğrencilerden yapay zekâ kullanmadıklarına dair beyan isteyebilir. Ödevlerde belirli kurallar koyabilir. Bunların bazıları belirli yaş grupları ve belirli çalışmalar için gerekli de olabilir.</p><p><strong>Fakat yapay zekâyı öğrencinin hayatından tamamen çıkarmak artık mümkün değildir.</strong></p><p>Okul “biz yasakladık” dediğinde çoğu zaman yalnızca şunu yapmış olur: Yapay zekâ kullanımını okulun görebileceği, konuşabileceği, rehberlik edebileceği alandan çıkarıp görünmez alana iter.</p><p>Bu da eğitim açısından riskli bir durumdur. Çünkü öğrenci bir aracı okul dışında, rehbersiz, kuralsız ve tartışmasız kullanıyorsa, okul o konuda eğitici rolünü kaybetmeye başlar.</p><p>Oysa okulun gücü yalnızca yasak koymasında değildir. <strong>Okulun asıl gücü, öğrencinin karşılaştığı dünyayı anlamlandırmasına yardım etmesindedir.</strong></p><p>Çocuklar ve gençler yapay zekâ ile karşılaşacaklar. Soru şu: Bu karşılaşma öğretmensiz, rehbersiz ve rastgele mi olacak? Yoksa okul bu karşılaşmayı <strong>bilinçli bir öğrenme fırsatına mı çevirecek?</strong></p><h3>Yasaklamak eşitsizliği artırabilir</h3><p>Yapay zekâ yasağı ilk bakışta eşitlikçi görünebilir. Herkese aynı kural konur: Kullanmak yok. Böylece kimse avantaj elde etmez gibi düşünülür.</p><p>Fakat gerçek hayat genellikle bu kadar basit işlemez.</p><p>Yasaklar, özellikle teknoloji konusunda, her öğrenciyi aynı şekilde etkilemez. Evinde bilgisayarı olan, iyi internet erişimi bulunan, ailesinden destek alabilen, teknolojiyle erken tanışmış veya özel ders desteği gören öğrenciler yapay zekâyı okul dışında öğrenmeye ve kullanmaya devam edebilir.</p><p><strong>Buna karşılık, bu araçlarla yalnızca okul aracılığıyla tanışabilecek öğrenciler geride kalabilir.</strong></p><p>Bu durumda yasak, yapay zekâyı ortadan kaldırmaz. Sadece onu daha görünmez ve daha eşitsiz hale getirir. Bazı öğrenciler yapay zekâyı daha iyi soru sormak, yazısını geliştirmek, yabancı dil çalışmak, deneme çözmek, proje üretmek için kullanmayı öğrenirken; diğer öğrenciler bu yeni okuryazarlıktan mahrum kalabilir.</p><p><strong>Eğer okulun, özellikle devlet okullarının temel görevlerinden biri fırsat eşitliğini güçlendirmekse, yapay zekâyı tamamen dışarıda bırakmak bu göreve hizmet etmeyebilir.</strong> Tam tersine, okulun rehberliğinde bilinçli kullanım, bu yeni teknolojik farkın toplumsal eşitsizliğe dönüşmesini azaltabilir.</p><p><strong>Bu yüzden yapay zekâ meselesi yalnızca “kopya çekme” meselesi değildir. Aynı zamanda bir fırsat eşitliği meselesidir.</strong></p><h3>Asıl mesele öğrenciyi güçlendirmek</h3><p>Yapay zekâ tartışmasını “yasak mı, serbest mi?” ikiliğine sıkıştırdığımızda asıl eğitim sorusunu kaçırıyoruz.</p><p>Asıl mesele öğrencinin yapay zekâ kullanıp kullanmaması değil; bu araçla karşılaştığında nasıl bir zihinsel tutum geliştireceğidir.</p><p>Öğrenci yapay zekânın ürettiği her cevabı doğru mu kabul edecek? Yoksa sorgulayacak mı? Kendi fikrini oluşturmadan çıktıyı kopyalayacak mı? Yoksa yapay zekâyı kendi düşüncesini geliştirmek için bir karşı ses gibi mi kullanacak?</p><p><strong>Gelecekteki güçlü öğrenci, yapay zekâdan hiç yararlanmadan mezun olan öğrenci mi olacak? Bence hayır. </strong>Güçlü öğrenci<strong> </strong>yapay zekâ ile çalışırken kendi aklını devre dışı bırakmayan öğrenci olacak.</p><p>Bu yeni dönemde öğrencinin öğrenmesi gereken beceriler değişiyor. İyi soru sormak, gelen cevabı değerlendirmek, eksik veya hatalı bilgiyi fark etmek, farklı kaynaklarla karşılaştırmak, etik sınırları bilmek, kendi emeği ile makine desteği arasındaki farkı ayırt etmek daha önemli hale geliyor.</p><p>Bunlar kendiliğinden gelişmez. Öğrencinin bu becerileri okulda, öğretmen rehberliğinde, güvenli ve eğitici örneklerle öğrenmesi gerekir.</p><p>Bu nedenle yapay zekâdan kaçmak yerine, öğrenciyi bu araç karşısında daha güçlü hale getirmek gerekir.</p><h3>Öğretmen hâlâ en önemli rehber</h3><p>Yapay zekâ tartışmalarında öğretmen bazen yanlış bir yere konumlandırılıyor. Sanki yapay zekâ gelirse öğretmenin rolü azalacakmış gibi düşünülüyor. Oysa tam tersine, yapay zekâ çağında öğretmenin rolü daha da önemli hale geliyor.</p><p><strong>Öğretmen, yapay zekâyı yasak tabelasıyla değil, öğrenme tasarımıyla karşılayabilir.</strong> Öğrenciye “bunu kullanma” demenin ötesine geçip “bunu nasıl kullanırsan düşüncen gelişir, nasıl kullanırsan öğrenmen zayıflar?” sorusunu öğretebilir.</p><p>Mesela bir yazma ödevinde öğrenciden yalnızca son metni istemek yerine, süreci de isteyebilir: ilk taslak, yapay zekâdan alınan geri bildirim, öğrencinin bu geri bildirime verdiği cevap, son metinde neyi değiştirdiği. Böylece yapay zekâ kopya makinesi olmaktan çıkar, düşünme sürecinin görünür hale geldiği bir araca dönüşebilir.</p><p>Sevsek de sevmesek de yapay zekâ artık hayatımızın bir parçasıdır.</p><p>Bu yüzden okullarda yapay zekâyı konuşmamız gerekiyor. Korkmadan, ama hafife almadan. Yasak refleksine teslim olmadan, ama sınıf gerçekliğini de inkâr etmeden.</p><p>Öğrenciyi gerçekten korumak istiyorsak, onu yapay zekâdan tamamen uzak tutmaya çalışmak yetmez. Ona yapay zekâ çağında düşünmeyi, yazmayı, sorgulamayı, üretmeyi ve emek vermeyi öğretmemiz gerekir.</p><p><strong>Çünkü yapay zekâyı okuldan kovmak, öğrenciyi korumak değil; onu geleceğe hazırlıksız bırakmaktır.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=22b7aac5b152" width="1" height="1" alt=""><hr><p><a href="https://medium.com/t%C3%BCrkiye/okullarda-yapay-zek%C3%A2-yasaklanmal%C4%B1-m%C4%B1-22b7aac5b152">Okullarda Yapay Zekâ Yasaklanmalı mı?</a> was originally published in <a href="https://medium.com/t%C3%BCrkiye">Türkçe Yayın</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI in Education: A Starter Pack for Busy Educators]]></title>
            <link>https://efemisirli.medium.com/ai-in-education-a-starter-pack-for-busy-educators-c599858fb021?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/c599858fb021</guid>
            <category><![CDATA[ai-in-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Wed, 24 Jun 2026 04:36:40 GMT</pubDate>
            <atom:updated>2026-06-24T04:36:40.489Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/910/0*mrckZvGqzfOuj42N.jpeg" /></figure><p>Artificial Intelligence is quickly becoming a part of everyday teaching and learning — from automated feedback tools to personalized tutoring systems. But if you’re an educator without a background in AI or computer science, knowing where to begin can feel daunting.</p><p>That’s why I’ve put together this curated “starter pack” — a practical set of free, beginner-friendly resources to help you get up to speed on AI in education. Whether you’re just curious or planning to integrate AI tools into your profession and/or research, these materials will give you a solid foundation without needing hours of spare time (or a tech degree!).</p><p>Before jumping into the resources, I want to share how I approached learning about AI in education myself a couple years ago. I knew I didn’t need to become a coder or machine learning expert, but I did want to understand enough to feel confident about mechanisms behind the AI systems and following the academic conversations happening in education.</p><p>Here’s the simple plan I created for myself:</p><ol><li><strong>🧠 Get the basics down</strong> — I started by learning how AI works at a high level, especially things like <strong>deep learning and large language models (LLMs),</strong> but without going too deep into the technicalities.</li><li><strong>🎓 Take online courses</strong> — I picked a few beginner-friendly courses to build a structured understanding of how AI can be used in a practical education setting.</li><li><strong>📚 Read books and papers</strong> — I explored some recent books and academic articles to dive deeper into the educational applications of AI.</li><li><strong>📰 Follow what’s new</strong> — I began following AI in education news, research, and tools on a regular basis so I could stay up to date.</li><li><strong>✍️ Share what I learn</strong> — Writing this blog is part of how I make sense of it all and share helpful info with others who are on the same path.</li></ol><p>If you’re just starting out, you don’t have to follow these steps exactly — but having some kind of strategy makes it way less overwhelming. In the rest of this post, I’ll share some of the resources that helped me along the way.</p><p>Karpathy is known for explaining cutting-edge AI concepts in a way that’s friendly, and visual, plus and he is very into AIED with his new venture <a href="https://eurekalabs.ai/">Eureka Labs</a>.</p><p>Demis is one of the most visionary minds in AI today, a recent Nobel laurate and his talks blend science, ethics, and human learning in fascinating ways.</p><p>Ilya is less “performative” than others, but if you want to understand where LLMs came from and where they might go, he’s a must-watch.</p><p>These resources helped me go from “What even <em>is</em> an LLM?” to “Okay, I get the general idea now.” You don’t need to watch everything in one sitting — even picking two or three from this list can give you a solid foundation but <strong>I really recommend the three Karpathy videos above.</strong></p><p>I also wanted to delve deeper and gain practical insights into how AI can be applied in educational settings. Online courses provided a structured approach to enhance my knowledge and skills. Here are a couple course that I found particularly beneficial:</p><p>A beginner level course for educators on how to develop successful prompts for different use cases and explores ethical implications of using AI in educational settings.</p><p>This beginner-friendly course is designed specifically for educators, offering a comprehensive overview of AI’s potential in education such as essential AI concepts, practical applications, prompt engineering and responsible AI use.</p><p>Here are some papers I found accessible and very useful in my research:</p><p>📚 <a href="https://arxiv.org/abs/2408.12143"><strong>Promises and challenges of generative artificial intelligence for human learning</strong></a> by Lixiang Yan, Samuel Greiff, Ziwen Teuber, Dragan Gašević</p><p>As the title suggest, this paper provides great overview of the field and is a must read.</p><p>📚 <a href="https://www.sciencedirect.com/science/article/abs/pii/S1041608023000195"><strong>ChatGPT for Good? On Opportunities and Challenges of Large Language Models for Education</strong></a> by Kasneci et. al</p><p>This is also a paper that provides great overview of the field.</p><p>📚 <a href="https://www.mdpi.com/2071-1050/15/17/12983"><strong>Transforming Education: A Comprehensive Review of Generative Artiﬁcial Intelligence in Educational Settings through Bibliometric and Content Analysis</strong></a><strong> by </strong>by Zied Bahroun, Chiraz Anane, Vian Ahmed and Andrew Zacca</p><p>This is a literary review of the field and summarizes the research done so far.</p><p>📚 <a href="https://link.springer.com/article/10.1007/s13347-025-00883-8"><strong>Why AI will not Democratize Education: a Critical Pragmatist Perspective</strong></a>by Michał Wieczorek</p><p>To examine critical takes on a field that you are excited about is always important. This paper provides that very well.</p><p>Keeping up with the rapid advancements in AI within education is crucial. Here are some valuable resources to help you stay informed</p><p>Dr. George Siemens, a prominent figure in the field of learning analytics and AI, publishes a newsletter called <em>SAIL: Sensemaking, AI, and Learning</em>. You can read the latest editions and subscribe to the newsletter on the <a href="https://buttondown.com/SAIL/archive/">Buttondown website</a>. <a href="https://buttondown.com/SAIL/archive/?utm_source=chatgpt.com">stage.calstate.edu+2buttondown.com+2buttondown.com+2</a></p><p>Professor Rose Luckin, a leading expert in AI and education, publishes a monthly newsletter titled <em>The Skinny on AI for Education</em>. This newsletter offers insights into the latest developments at the intersection of AI and education. You can subscribe to the newsletter and access past issues on the <a href="https://www.educateventures.com/the-skinny">Educate Ventures Research website</a>. <a href="https://www.educateventures.com/the-skinny?utm_source=chatgpt.com">x.com+7educateventures.com+7educateventures.com+7</a></p><p>LinkedIn is an excellent platform to stay connected with experts and thought leaders and follow their work. By following key individuals and organizations, you can receive a steady stream of insightful posts, articles, and discussions right in your feed.</p><p><strong>Save posts for later reading:</strong> LinkedIn allows you to bookmark posts and articles. Use this feature to create a personal reading list, so you can dive deeper when you have time without losing valuable resources.</p><p>One of the best ways to deepen your understanding of AI in education is to share what you learn (like I am doing right now). You don’t need to be a professional writer start simple. Summarize articles, share your take on new AI tools, or document your learning journey. Over time, your blog can become a valuable resource for others, and a place where your own knowledge continually grows.,</p><p>Embarking on the journey to understand AI in education doesn’t have to be overwhelming. By breaking it down into manageable steps — building a foundation, taking courses, reading key resources, staying updated, and sharing your insights — you can steadily grow your knowledge and confidence.</p><p>AI is transforming how we teach and learn, and educators who stay informed and proactive will be best positioned to harness its potential responsibly and creatively. Remember, it’s not about mastering everything overnight, but about continuous learning and curiosity.</p><p>I hope this starter pack helps you take your first steps into the exciting world of AI in education. If you have questions or resources to share, feel free to connect!</p><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/ai-in-education-a-starter-pack-for"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c599858fb021" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Meet LearnLM: Google’s New AI Tutor — A Deep Dive into Its Research and Real-World Performance]]></title>
            <link>https://efemisirli.medium.com/meet-learnlm-googles-new-ai-tutor-a-deep-dive-into-its-research-and-real-world-performance-1277462ff1ec?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/1277462ff1ec</guid>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Tue, 23 Jun 2026 07:52:11 GMT</pubDate>
            <atom:updated>2026-06-23T07:52:11.952Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7KnCFMpJTV9VbCYW.png" /></figure><p>Now, with the rise of large language models (LLMs), a new wave of optimism (and skepticism) has emerged. LLMs like ChatGPT and Gemini can converse naturally and answer complex questions. But are they truly capable of <em>teaching</em>? Critics warn that these systems, by default, <strong>optimize for helpfulness rather than pedagogy</strong> -often giving away answers too easily or failing to adapt to learners’ needs.</p><p>Against this backdrop, Google has introduced <strong>LearnLM</strong>, an LLM-based tutor explicitly designed for educational use. Rather than relying solely on prompting, <strong>LearnLM is fine-tuned to follow pedagogical instructions-adapting its behavior based on specific learning goals.</strong></p><p>Google recently released two technical papers <a href="https://goo.gle/LearnLM">(1)</a><a href="https://goo.gle/LearnLM-dec24">(2)</a> detailing the vision, training, and evaluation of LearnLM. In this post, I’ll introduce LearnLM Tutor, summarize and contrast these two papers, and share my own test experience with the model. <strong>My aim is to assess whether LearnLM, at its current iteration, can bring closer to the long-standing vision of AI-powered education.</strong></p><p>LearnLM Tutor is currently available in experimental form via <a href="https://aistudio.google.com/">Google AI Studio</a>, where developers and educators can test its capabilities and offer feedback to help shape future iterations.</p><p>The first technical paper presents the foundational vision behind LearnLM-Tutor, grounded in the long-standing promise of AI in education which sounds very good to me: <strong><em>a personal tutor for every learner and a teaching assistant for every teacher</em>.</strong></p><p>However this is easier said than done so far due to various barriers . The paper identifies several critical barriers to achieve this goal which I completely agree with:</p><p>Through interviews and co-design sessions with educators, the authors derive core principles for both human and AI tutors:</p><ul><li><strong>Don’t give away solutions-guide discovery.</strong></li><li><strong>Make explanations clear and relatable.</strong></li><li><strong>Be encouraging and celebrate progress.</strong></li><li><strong>Proactively recognize when a learner is struggling.</strong></li><li><strong>Ask questions to check understanding.</strong></li><li><strong>Teach thinking processes step-by-step.</strong></li></ul><p>These reasonable sounding principles formed the basis for the pedagogical goals behind LearnLM.</p><p>To guide development and evaluation of the model, the team created a rubric that consist of five pedagogic dimensions:</p><p>Below is a further breakdown of these dimensions from the paper to provide deeper context:</p><p>While the first paper set the conceptual foundations for LearnLM, the <a href="https://goo.gle/LearnLM-dec24">second paper</a>, released about six months later, focuses on how Google substantially refined the training and evaluation of the the model based on feedback from stakeholders and advances in the Gemini base model. The current iteration of LearnLM, built on Gemini 1.5 Pro, appears to be a much more more effective AI tutor according to the authors due to these refinements.</p><p>First, LearnLM moves beyond prompt-based behavior shaping. Instead, it introduces <strong>pedagogical instruction following</strong> as the core mechanism for aligning the model with educational goals. This means developers can define <em>system-level instructions</em> such as “don’t give away the answer” or “provide step-by-step feedback” which the model learns to follow consistently across interactions.</p><p>To do this, LearnLM’s training combined:</p><p>While the LearnLM team has developed a detailed pedagogical rubric grounded in learning science (as detailed in the first paper), they also recognize the limitations of their approache. In particular, they highlight the need to go beyond lab-based assessments and measure real-world educational impact. As they note:</p><blockquote>Second, we would like to start moving from intrinsic evaluations, which measure the model’s performance according to a predefined pedagogy standard, to extrinsic evaluation, which measure impact such as learning outcomes. Intrinsic evaluations are useful for model development, as they are faster to run and directly identify the shortcomings in the models. However, while the core principles of our rubric, such as encouraging active learning and managing cognitive load, are broadly agreed upon and evidence-based [18], it is unclear how well the results translate to improvements in learning outcomes. It is likely that as the field matures and AI systems master the basics of tutoring dialogue, extrinsic evaluations will play a more important role.</blockquote><p>Across thousands of conversations and over 10,000 expert ratings:</p><ul><li><strong>LearnLM, as a tutor, was preferred over GPT-4o by 31%</strong>, <strong>over Claude 3.5 by 11%, and over Gemini 1.5 Pro by 13%.</strong></li><li><strong>It consistently scored higher on tutoring effectiveness, instruction adherence, and learner engagement.</strong></li><li><strong>Importantly, these improvements did <em>not</em> come at the cost of general model performance-LearnLM retained its ability to reason, respond to diverse topics, and understand context.</strong></li></ul><p>To see how LearnLM Tutor performs outside a controlled study, I ran my own basic test. I intentionally avoided being “hacky” or testing edge cases, aiming instead to experience the model in line with the developers’ intended use.</p><p>I accessed the model through <a href="https://aistudio.google.com/">Google AI Studio</a>, as recommended by the developers. However, locating the model was not entirely straightforward-I had to search through the options on the right-hand side for a while before finding it.</p><p>After locating the model, my initial experience was lukewarm, as it behaved like a regular LLM rather than assuming the tutor role. This was confusing, especially since I had explicitly used the phrase “teach me” in my prompt to trigger its tutoring behavior, as shown below.</p><p>Instead of engaging in a tutoring interaction, the model returned an encyclopedic entry on James Watt, which was not what I expected (I used the default setting of the model). It felt as though I was interacting with the wrong model. Luckily, things went smoother with my next prompt:</p><p>Probably due to my explicit “be a tutor” prompt, the model snapped out of the default LLM persona and started acting like an helpful tutor:</p><p>Suddenly, I found myself more interested in how a steam engine works than in James Watt himself. ( <em>Can I call this “Vibe Learning?”</em>) I’ve been curious about steam engines ever since a visit to the Science Museum, so this became a great opportunity to see how the model handles <strong>context shifts</strong> in educational conversations.</p><p>And it handled the shift quite well. The model didn’t shame me for my pedestrian understanding of steam engines; instead, it was encouraging and motivating. <strong>This positive attitude made me trust the tutor more and that trust pushed me to continue exploring.</strong></p><p>I am geniuenly impressed with the rest of the conversation.</p><p>As the developers intended, the model did well to resist my prompt the give the answer right away and encouraged me to explore a bit further:</p><p>Eventually, the model began explaining components of the steam engine, such as valves, as expected. But at some point, I began to feel overwhelmed, and I shared that with the model. Its response was thoughtful-it acknowledged my earlier effort, encouraged me, and rephrased the content in a more accessible way.</p><p>At this point I really wanted to <em>see</em> a basic schematic of a steam engine to visualize the distinct parts and how they interact. So I asked the model to provide me one:</p><p>While it seems like the model does not yet support multimodality or tool use, it offered a clever workaround: it encouraged me to sketch the engine myself and gave helpful tips for finding reference images online. This was a pedagogically smart move-it shifted the activity toward <strong>active learning rather than passive consumption.</strong></p><p>Next I submitted some prompts in my native language Turkish to test its multilanguage capabilities and pleasantly see that it can converse in Turkish too. At this point I thought I got the gist of the model which was my intention, felt a bit tired and stopped interacting with it. In short, I liked what I experienced so far and planning to do more experiments with it later.</p><p>Then I get the idea to replicate the same exercise with a regular LLM. I chose Grok 3 Beta to test, a model I use regularly. Even though I used the exact same prompts with Grok (default settings, no persona selected), the learning experience was far worse. At each time Grok produced wall of texts followed by generic questions which was neither engaging nor motivating. Unfortunately I did not have time to test other LLM’s. I am particularly curious about Claude’s performance as a tutor, as it was one of the well performing LLM’s according to the Google technical paper.</p><p>Learning about it and my initial testing of LearnLM Tutor left me feeling <strong>genuinely optimistic about the future of AI in education.</strong> While it didn’t blow me away from the first prompt, once it clicked into its tutoring role, the experience felt thoughtful, supportive, and importantly <em>human-centered</em>. It encouraged me, adapted to my needs, and gently nudged me to think rather than just handing over answers.</p><p>Of course, this is still early days. There are rough edges and open questions-especially around how we measure real learning gains, not just “good conversations.” But based on what I’ve seen, <strong>LearnLM feels like a step in the right direction.</strong> I’m particularly <strong>excited about the integration of reinforcement learning (RL) in tutoring models, and its potential to generate emergent tutoring behaviors.</strong> I’m also curious to see how the developers will eventually <strong>incorporate multimodality into the experience in a way that is pedagogically meaningful.</strong></p><p>As educational AI matures, models like LearnLM raise important questions: Can we reliably measure real learning outcomes from AI tutoring? How do we ensure these systems remain inclusive, adaptable, and trustworthy across learners and contexts?</p><p>These are the kinds of questions this blog will continue to explore.</p><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/meet-learnlm-googles-new-ai-tutor"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1277462ff1ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I A/B Tested My Own Brain to see if AI Tutoring Can Prep Me for My PhD Stats Exam]]></title>
            <link>https://efemisirli.medium.com/i-a-b-tested-my-own-brain-to-see-if-ai-tutoring-can-prep-me-for-my-phd-stats-exam-62e2430abd0c?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/62e2430abd0c</guid>
            <category><![CDATA[ai-in-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Mon, 22 Jun 2026 07:51:40 GMT</pubDate>
            <atom:updated>2026-06-22T07:51:40.775Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AC2U4wCdpmkILiug.png" /></figure><p>As a PhD student in Education Technology, I’m supposed to be good at learning. But to be honest, <strong>statistics has always been a struggle for me.</strong></p><p>It wasn’t just the math; it was the motivation. Trying to study complex statistical concepts on my own felt like wading through mud. I would open the textbook, stare at the formulas, and feel that familiar wave of “I’m never going to get this.”</p><p>And let’s be real: <strong>Studying gets harder as you get older.</strong> I’m not an undergrad with endless free time anymore. I have life obligations, work, and a hundred other things fighting for my attention. Focusing on a dry stats textbook for hours just wasn’t happening. I needed something that could hold my attention and actually <em>teach</em> me, not just present information.</p><p>I had a critical Statistics exam coming up covering 11 chapters. I realized this was the perfect opportunity to turn myself into a lab rat.</p><p>I didn’t just want to “study”; I wanted to know-statistically-if an AI tutor could break through my mental block where traditional studying failed. So, I designed a mini One-Group Pretest-Posttest experiment using the latest tools from Google.</p><p>Here is exactly how I did it, and the data on what happened.</p><p>One of the biggest myths in AI right now is that you need to be a “prompt engineering wizard” to get good results. My experiment proved the opposite.</p><p>I used <strong>LearnLM</strong>, a Gemini Gem powered by the new <strong>Gemini 3.0</strong>. I used it because I am familiar with it <a href="https://efemisirli.substack.com/p/meet-learnlm-googles-new-ai-tutor">as reviewed two papers from its developers before</a>. My setup was intentionally simple to test the raw capability of the model:</p><ul><li><strong>The Context:</strong> I simply uploaded my course textbook as a PDF.</li><li><strong>The “System” Prompt:</strong> None. I used a fresh conversation with no pre-conditioning.</li><li><strong>The Trigger:</strong> My prompts were caveman-simple. I literally typed: <em>“Tutor me on Chapter 1”</em>, <em>“Chapter 2”</em>, etc.</li></ul><p>I wanted to see if the model could handle the pedagogical heavy lifting without me holding its hand.</p><p>To make this a at least a passible scientific experiment, I needed to ensure I wasn’t just memorizing answers. I needed an impartial judge to create my exams before the real exam.</p><p>For each of the 11 chapters:</p><ol><li><strong>The Baseline (Pre-Test):</strong> I had NotebookLM generate a unique 25-question test based on the chapter. I took it <em>before</em> any study session.</li><li><strong>The Intervention:</strong> I had my tutoring session with LearnLM (Gemini 3.0).</li><li><strong>The Measurement (Post-Test):</strong> I went back to NotebookLM and asked for a <em>brand new</em> set of 25 questions for the same chapter.</li></ol><p>Transparency is key in research. Here are the raw scores recorded for each chapter before and after the AI intervention.</p><p>I ran the numbers using Python. The difference wasn’t just noticeable; it was drastic.</p><p><em>The trajectory of my scores for every single chapter. Notice the upward trend across the board.</em></p><p>Before the AI tutoring, my average score was sitting at a <strong>60.45</strong>. I was passing, but barely comfortably. After the simple tutoring sessions with LearnLM, my average skyrocketed to <strong>87.27</strong>.</p><p><em>The boxplot above shows how the spread of scores tightened and shifted upwards significantly.</em></p><p>For the data nerds reading this: Yes, I ran a Paired Samples t-Test.</p><p>The result was statistically significant (p &lt; .001), meaning this wasn’t luck.</p><p>But the real star of the show is the Effect Size. In statistics, we use <strong>Cohen’s <em>d</em></strong> to measure how powerful an intervention is.</p><p>In educational research, a result of 1.79 isn’t just “large”-it’s massive. It suggests that the AI didn’t just help me review; it fundamentally shifted my performance bracket.</p><p>Experimental data is fun, but the real test was the actual university exam I was dreading.</p><p>I crushed the exam.</p><p>For the first time in my academic life, I sat down in front of a statistics paper and didn’t feel that rising panic. I knew the concepts. I recognized the patterns. The definitions I had discussed with Gemini came back to me naturally. The experiment didn’t just look good on a graph; it delivered a real-world win when it mattered most.</p><p>I walked into this experiment with more than just academic curiosity; <strong>I walked in with anxiety.</strong> Statistics had always been a mental roadblock for me, and the sheer volume of the material made me dread opening the book. <strong>The fear of failing-combined with the exhaustion of adult life-had paralyzed my motivation.</strong></p><p>The AI tutor did something sitting with a textbook never could: <strong>It cured my study paralysis.</strong></p><p>If you are struggling with a subject, battling exam anxiety, or if life is just getting in the way of your focus, don’t give up. Upload your PDF, ask for a tutor, and let the data speak for itself.</p><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/i-ab-tested-my-own-brain-to-see-if"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62e2430abd0c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Eğitimde Yapay Zeka Gerçeği: “Temkinli İyimserlikten” Sıyrılma Zamanı]]></title>
            <link>https://medium.com/t%C3%BCrkiye/e%C4%9Fitimde-yapay-zeka-ger%C3%A7e%C4%9Fi-temkinli-i%CC%87yimserlikten-s%C4%B1yr%C4%B1lma-zaman%C4%B1-174c82133165?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/174c82133165</guid>
            <category><![CDATA[etik]]></category>
            <category><![CDATA[temkinli-iyimserlik]]></category>
            <category><![CDATA[eğitim]]></category>
            <category><![CDATA[türkçe]]></category>
            <category><![CDATA[yapay-zeka]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Fri, 19 Jun 2026 04:15:05 GMT</pubDate>
            <atom:updated>2026-06-19T04:15:05.718Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jxWWFpIJh8CEt5No.png" /></figure><p>Biz hala <strong>kopya çekilmesini nasıl engelleriz</strong> diye tartışırken, yaklaşan Genel Yapay Zeka (AGI), öğrencilerimize kazandırmaya çalıştığımız becerileri birer birer tarihe gömmeye hazırlanıyor. Buna rağmen gözlemlediğim kadarıyla eğitimde yapay zeka tartışılırken, eğitim konferanslarında, akademik koridorlarda ve öğretmen odalarında, garip bir hava var. Ben buna <strong>“temkinli bir iyimserlik”</strong> diyorum.</p><p>Yıllar içinde projeksiyon cihazlarından akıllı tahtalara, tablet pilot projelerinden EBA’ya kadar sayısız teknolojik dalgayı “atlatmış” olmanın getirdiği bir bağışıklık bu. Eğitimcilerimiz, yapay zeka devrimini de benzer bir refleksle karşılıyor: <em>“Bu da gelir, bu da geçer, biz neler atlattık!”</em></p><p>Gözlemlerim sadece kişisel bir histen ibaret değil; Türkiye’de son yıllarda yapılan akademik çalışmalar da bu “tehlikeli iyimserliği” doğruluyor.</p><p>Örneğin, Eriçok, Karataş ve Yüce’nin 2024 tarihli çalışmasında, öğretmen adaylarına yapay zekayı hangi metaforlarla tanımladıkları soruluyor. Cevaplar çok manidar: <strong>“Pusula”, “Kılavuz”, “Ansiklopedi”.</strong></p><p>Bu metaforların ortak özelliği şudur: Hepsi pasiftir. Pusula, yönü siz belirlerseniz işe yarar. Ansiklopediyi siz açıp okursanız bilgi verir. Oysa karşımızdaki teknoloji (Generative AI), gidilecek rotayı bizden daha iyi çizen, hatta o rotaya gidilip gidilmeyeceğine karar verebilen, muhakeme (reasoning) yeteneği kazanan bir yapı.</p><p>Yine 2025 yılında yayınlanan bir başka çalışma (Topçubaşı, 2025), öğretmen adaylarının yapay zekayı bir “devrim” olarak değil, sadece iş yükünü hafifletecek, materyal hazırlayacak bir <strong>“kırtasiye asistanı”</strong> olarak gördüğünü ortaya koyuyor. Eğitim camiası, yaklaşan tsunamiyi “hafif bir meltem” sanma hatasına düşüyor.</p><p>Şu an eğitimdeki ana gündem maddelerine bakın:</p><p>Bunlar elbette önemli <strong>etik</strong> sorunlar. Ancak bu tartışmalar, <strong>yaklaşmakta olan fırtınada ıslanmamak için şemsiyeyi hangi açıyla tutmamız gerektiğini tartışmaya benziyor.</strong> Oysa yaklaşan şey bir yağmur değil, binanın temellerini sökecek bir fırtına.</p><p>Etik sorunların tamamı çözülse, veri %100 güvenli olsa ve AI hiç halüsinasyon görmese bile; ortada hala devasa, varoluşsal bir soru duruyor:</p><p>Gelin, o “temkinli iyimserlik” perdesini biraz daha aralayalım ve kimsenin sesli sormaya cesaret edemediği o soruyu soralım: <strong>Eğer bilgiye erişim, bilginin işlenmesi ve hatta yeni bilginin sentezlenmesi bir yapay zeka tarafından daha iyi yapılabiliyorsa, akademisyene ve öğretmene ne gerek var?</strong></p><p>Bugüne kadar cevabımız hep hazırdı: <em>“Makine öğretir ama insan eğitir. Duygu, empati ve ilham sadece insandadır.”</em><strong>Bu, bence bizi rahatlatan romantik bir savunma mekanizması.</strong> Ancak AGI senaryosunda bu savunma hattı da çöküyor.</p><p><strong>Akademisyenler için tehlike çanları:</strong> Akademik üretimin büyük kısmı literatür taraması, veri analizi ve mevcut bilginin yeniden yorumlanmasına dayanır. Bugün yapay zeka, binlerce makaleyi dakikalar içinde okuyup sentezleyebiliyor, hipotez kurabiliyor ve hatta bazı alanlarda kod yazıp test edebiliyor. Eğer bir akademisyenin işlevi sadece “bilgiyi aktarmak” veya “vasat bir literatür özeti çıkarmak” ise, üzülerek söylemeliyim ki o kürsü çoktan yıkıldı. Yapay zeka, “ortalama” olan her şeyi ikame edecektir.</p><p><strong>Öğretmenler için yol ayrımı:</strong> Sınıflarda durum daha da kritik. Bir öğretmen aynı anda 30 öğrenciye yetişemez ama kişiselleştirilmiş bir AI tutoru, her öğrencinin hızına, ilgisine ve öğrenme stiline göre 7/24 sabırla ders anlatabilir. Eğer öğretmenlik tanımımızı “müfredatı aktaran kişi” olarak sınırlı tutarsak, bu mesleğin son kullanma tarihi gelmiştir.</p><p><strong>“İnsan dokunuşu” ve “rehberlik” elbette değerlidir; ancak milyonlarca öğretmeni istihdam etmek için sadece “insan sıcaklığı” yeterli bir ekonomik gerekçe olacak mı? Yoksa öğretmenlik, sadece çok küçük, elit bir zümrenin erişebildiği lüks bir “butik hizmete” mi dönüşecek?</strong></p><p>Bu sorulara “Hayır, bize her zaman ihtiyaç var” demek, matbaanın icadından sonra hattatların “Ama el yazısının ruhu başkadır” demesine benziyor. Ruh başkadır belki, ama fonksiyon değişmiştir.</p><p>Belki de yüzleşmekten en çok kaçındığımız gerçek şu: Yüzyıllardır geçerli olan toplumsal sözleşme çatırdıyor. Bu sözleşme basitti: <strong>“Okula git, beceri kazan, diploma al ve karşılığında ekonomik bir değer (iş) üret.”</strong></p><p>Eğitim sistemi, özünde piyasaya kalifiye “bilişsel iş gücü” sağlayan bir huni olarak tasarlandı. Ancak AGI senaryosu, bu huninin altını kapatıyor.</p><p>Eğer Yapay Genel Zeka (AGI) gerçekleşirse -ki modellerin gidişatı bu yönde- bu, bir yapay zekanın ekonomik değeri olan herhangi bir görevi, ortalama bir insandan daha iyi ve daha ucuza yapabileceği anlamına gelir. <strong>Böyle bir senaryoda, eğitim ve istihdam arasındaki o kadim köprü yıkılmaya mahkumdur.</strong></p><p>Bugün müfredatımızda yer alan becerilerin %90&#39;ı, “işe yarar bir çalışan” yaratmak üzerine kurulu. Peki ya “işe yarar çalışan” kavramı, silikon tabanlı bir zeka tarafından tamamen domine edilirse? <strong>O zaman diplomalar, birer yeterlilik belgesi olmaktan çıkıp, sadece nostaljik birer kağıt parçasına dönüşmez mi?</strong> “Çocuklar kodlama öğrensin” derken, kodun kendisini yazan, hata ayıklayan ve sistemi kuran bir zeka karşısında, onlara aslında “artık ihtiyaç duyulmayan bir zanaatı” öğretiyor olabilir miyiz?</p><p>Bu, basit bir “otomasyon” meselesi değil; insanın ekonomik üretimdeki merkezi rolünün kaybolması meselesidir. Eğitimcilerin “Bunu da atlatırız” iyimserliği, işte bu büyük kopuşu görmezden geliyor.</p><p>Eğitim dünyası şu an hararetle “intihal”, “veri gizliliği”, “yapay zekanın önyargıları” ve “akademik dürüstlük” konularını tartışıyor. <strong>Yanlış anlaşılmasın, bunlar çözülmesi gereken önemli sorunlar.</strong></p><p>Ancak kabul etmeliyiz ki; bu tartışmalar aynı zamanda eğitimciler için <strong>konforlu bir sığınak</strong> görevi görüyor. Çünkü “kopya çekmeyi nasıl engelleriz?” sorusu, tanıdık bir sorudur. <strong>Kontrol edebileceğimizi sandığımız bir alandır. Bize hala sistemin hakimi olduğumuzu hissettirir.</strong></p><p>Oysa bu “etik sığınağı”, dışarıdaki asıl yıkımı görmemizi engelliyor. Tüm etik sorunlar çözülse, veri %100 güvenli olsa ve AI hiç halüsinasyon görmese bile; büyük resimdeki o devasa soru değişmiyor: <strong>İnsan zekasına duyulan ihtiyacın azaldığı bir dünyada eğitimin amacı nedir?</strong></p><p>Şemsiyenin rengini, sapını veya etiğini tartışmayı bırakalım. Şemsiyeyi kapatıp kafamızı yukarı kaldırmamız gerekiyor. Fırtına kapıda değil, çoktan içeri girdi. Artık sorulması gereken soru, “Ödevi yapay zeka mı yaptı?” değil; <strong>“İstihdamın olmadığı bir gelecekte, ‘insan olmak’ ve ‘öğrenmek’ ne anlama gelecek?”</strong> olmalıdır.</p><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/egitimde-yapay-zeka-gercegi-temkinli"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=174c82133165" width="1" height="1" alt=""><hr><p><a href="https://medium.com/t%C3%BCrkiye/e%C4%9Fitimde-yapay-zeka-ger%C3%A7e%C4%9Fi-temkinli-i%CC%87yimserlikten-s%C4%B1yr%C4%B1lma-zaman%C4%B1-174c82133165">Eğitimde Yapay Zeka Gerçeği: “Temkinli İyimserlikten” Sıyrılma Zamanı</a> was originally published in <a href="https://medium.com/t%C3%BCrkiye">Türkçe Yayın</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Literacy Isn’t “Just Prompting”: A Practical Field Guide to the Prominent Frameworks]]></title>
            <link>https://efemisirli.medium.com/ai-literacy-isnt-just-prompting-a-practical-field-guide-to-the-prominent-frameworks-e4e2c10dbfba?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/e4e2c10dbfba</guid>
            <category><![CDATA[literacy]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Mon, 15 Jun 2026 04:21:16 GMT</pubDate>
            <atom:updated>2026-06-15T04:21:16.135Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6QhACWJwi1VV1DZm.png" /></figure><p>“AI literacy” is now the new buzzword in education and professional development. Everyone agrees we need it, but almost nobody can concretely define what <em>it</em> is. Is it prompt engineering? Knowing how neural nets work? Ethics? Data privacy? Creativity? All of the above?</p><p>Here’s the problem: most debates about AI literacy are really debates about <strong>which framework</strong> people are implicitly using (often without knowing it). So instead of adding more noise, this post curates and decodes a handful of the frameworks that are actively shaping how AI skills are taught and assessed-using your attached examples as distinct “cases.”</p><h3>The move that cuts through the noise</h3><p>Stop asking: “What is AI literacy?”</p><p>Start asking: <strong>“What job do I need an AI literacy framework to do?”</strong></p><p>Because frameworks are tools. Different tools, different jobs.</p><p>Below are five “cases” you can treat like a menu</p><h3>Case 1: The “full-stack learner” view (OECD’s: Empowering Learners for the Age of AI).</h3><p>If you want a framework that says <em>AI literacy = knowledge + skills + attitudes</em> (not just tool-use), the <a href="https://ailiteracyframework.org/wp-content/uploads/2025/05/AILitFramework_ReviewDraft.pdf">Empowering Learners for the Age of AI</a> is built for that.</p><p>It defines AI literacy as the technical knowledge, durable skills, and future-ready attitudes needed to thrive in an AI-influenced world-enabling learners to <strong>engage, create with, manage, and design AI</strong> while evaluating risks and ethics.</p><p>What makes it useful in practice is its <strong>four-domain structure</strong> (a clean curriculum scaffold):</p><p><strong>When to use it:</strong> you’re designing a program and want a “complete” map that includes mindsets (curiosity/responsibility), not only mechanics.</p><h3>Case 2: The “curriculum + progression” view (UNESCO Student AI Competency)</h3><p>If you need something that <em>looks and behaves like a curriculum blueprint</em>, <a href="https://unesdoc.unesco.org/ark:/48223/pf0000391105/PDF/391105eng.pdf.multi">UNESCO’s student framework</a> is doing that job.</p><p>It defines four essential elements of student AI competency:</p><p>Then it adds a progression model- <strong>Understand → Apply → Create</strong> -meant to support spiral learning and assess growth over time.</p><p><strong>When to use it:</strong> you’re building (or auditing) an actual curriculum, need level-based outcomes, and want ethics and design skills explicitly represented-not “bolted on later.”</p><h3>Case 3: The “classroom-ready practices” view (Digital Promise)</h3><p><a href="https://files.eric.ed.gov/fulltext/ED671235.pdf">Digital Promise’s framework </a>is extremely practical because it breaks AI literacy into <strong>demonstrable practices</strong> you can teach and observe.</p><p>It frames AI literacy around modes of engagement (Understand / Evaluate / Use) and provides “AI Literacy Practices” as actionable skills.</p><p>Example: it spells out what “Data Privacy &amp; Security” looks like in student behavior (e.g., identifying how personal info is collected, preventing collection, deleting collected data).</p><p>Two details make it especially useful right now:</p><p><strong>When to use it:</strong> you need lesson objectives, “look-fors,” and implementation support that works whether or not every learner has AI access.</p><h3>Case 4: The “AI is not a separate subject” view (DigComp 3.0)</h3><p>Recently published EC’s <a href="https://publications.jrc.ec.europa.eu/repository/bitstream/JRC144121/JRC144121_01.pdf">DigComp </a>3.0’s core claim is basically: <strong>AI competence is woven through digital competence</strong>, because AI is embedded in everyday digital tools.</p><p>It operationalizes that idea with a clever labeling system:</p><p>And it quantifies the “AI is everywhere” reality:</p><ul><li>14% of competence statements are AI-explicit, 68% AI-implicit</li><li>13% of learning outcomes AI-explicit, 63% AI-implicit</li></ul><p><strong>When to use it:</strong> you’re doing organization-wide skills mapping (schools, workforce, government) and need AI literacy to align with broader digital competence-without creating yet another silo.</p><h3>Case 5: The “OG” view (Long &amp; Magerko, 2020)</h3><p>If you want the academic backbone-what AI literacy <em>actually contains</em> based on literature-this is the anchor.</p><p>A key insight: AI is often hidden inside platforms, so people don’t realize they’re interacting with it. That makes <strong>Recognizing AI</strong> a foundational competency.</p><p>It also translates into design guidance: avoid deceptive “black box” patterns, promote transparency, and “unveil gradually” to prevent overload while still enabling understanding.</p><p><strong>When to use it:</strong> you’re building a framework from scratch, doing research, or want evidence-based coverage (not just a list of trendy skills).</p><h3>So… what is AI literacy, concretely?</h3><p>If you merge what these frameworks repeatedly converge on, you get a practical definition that’s much more actionable than “be good at AI”:</p><p>AI literacy is the ability to <strong>notice AI</strong>, <strong>judge it</strong>, <strong>use it responsibly</strong>, and <strong>shape outcomes</strong> (as a user, creator, manager, or designer)-grounded in knowledge, skills, and human values.</p><p>A “minimum viable AI literacy” checklist (use this to design a course or assess gaps):</p><h3>How to pick a framework without overthinking it</h3><p>Use this quick selector:</p><ul><li><strong>Building a brand-new AI literacy program?</strong> Start with <strong>AILit</strong> for structure, then borrow <strong>Digital Promise</strong> practices for teachability.</li><li><strong>Writing curriculum outcomes and assessments by grade band?</strong> Use <strong>UNESCO</strong> (progression is the point).</li><li><strong>Upskilling across an institution or workforce?</strong> Use <strong>DigComp 3.0</strong> to integrate AI across existing digital skills.</li><li><strong>Need academic defensibility / completeness checks?</strong> Use <strong>Long &amp; Magerko</strong> as the evidence-based coverage map.</li></ul><h3>A simple way to turn any framework into a course (in 60 minutes)</h3><p>Take whichever framework you choose, and force it through three practical questions:</p><ol><li><strong>Audience &amp; context:</strong> student? teacher? employee? citizen? (DigComp is strongest when context varies.)</li><li><strong>Observable behaviors:</strong> what would you <em>see</em> someone do? (Digital Promise shines here.)</li><li><strong>Assessment design:</strong> scenario-based tasks beat multiple-choice for ethics and judgment (UNESCO provides lots of scenario logic).</li></ol><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/ai-literacy-isnt-just-prompting-a"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e4e2c10dbfba" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cultivating a “Techno-Artisan” Ethos in the Classroom]]></title>
            <link>https://efemisirli.medium.com/cultivating-a-techno-artisan-ethos-in-the-classroom-7f7e81dfa683?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/7f7e81dfa683</guid>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Sat, 13 Jun 2026 08:05:47 GMT</pubDate>
            <atom:updated>2026-06-13T08:05:47.397Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bkHF2VKGHRTS3jIC.png" /></figure><p>As an EdTech researcher watching a six-year-old step into the school system today, the stakes of the Generative AI revolution feel incredibly immediate. We are raising a generation for whom the screen is no longer just a display; it is an oracle.</p><p>Right now, the default student interaction with AI is passive consumption. It is treated as the ultimate shortcut to bypass the friction of learning, a way to generate the essay, write the code, or solve the math problem so they can move on to something else.</p><p>If our educational response is simply to ban the tool, we will fail. If we let the tool do all the thinking, we will fail even worse.</p><p>The goal isn’t to reject the technology. The goal is to cultivate a <strong>Techno-Artisan Ethos</strong>, a mindset where students view AI as an exoskeleton for their mind, but recognize that their own hands and human intuition are the ultimate validators of truth.</p><p>How do we actually build this ethos in the classroom? It rests on three pedagogical pillars.</p><p>A true artisan knows which tasks to do by hand and which to automate. A master carpenter doesn’t feel guilty about using a power saw for rough cuts, but they rely on a hand chisel for the delicate, defining details.</p><p>In the classroom, we need to teach students the meta-skill of <strong>Digital Delegation</strong>. They must learn <em>what</em> to offload to the algorithm and <em>what</em> to fiercely protect as their own cognitive territory.</p><ul><li><strong>Offload:</strong> Syntax, repetitive formatting, basic brainstorming, and initial code generation.</li><li><strong>Protect:</strong> System logic, ethical judgment, problem-definition, and the final physical execution.</li></ul><p>We need to shift students from being “consumers” of AI outputs to “directors” of AI capabilities. A Techno-Artisan doesn’t ask ChatGPT to “do the project.” They ask it to “draft the blueprint,” fully knowing they are the ones who have to build it.</p><p>Digital environments are inherently frictionless. If a student designs a bridge in a CAD program or asks an LLM to generate the schematic for a smart garden, gravity and thermodynamics don’t really matter until you hit “simulate.”</p><p>To cultivate the Techno-Artisan ethos, we need a strict pedagogical rule: <strong>Digital generation must end in physical execution.</strong></p><p>If a student uses AI to write the code for a soil moisture sensor, the project cannot end on the screen. They must be required to wire the breadboard, attach the sensors, debug the inevitable loose connections, and put it in actual soil.</p><p>The ethos is forged in the messy, frustrating gap between the “perfect” AI code and the reality of a stripped wire or a burnt-out LED. This forces students to understand that the physical world is the ultimate fact-checker.</p><p>AI provides instant gratification. The physical world emphatically does not.</p><p>If we aren’t careful, the frictionless speed of GenAI will destroy our students’ attention spans and their tolerance for frustration. The Techno-Artisan ethos requires actively teaching resilience against the backdrop of instant digital answers.</p><p>When a 3D print fails four hours into the job, or the Arduino code provided by an LLM throws a hardware error because of a voltage mismatch, students will naturally want to quit. AI has taught them that answers should take three seconds. We are teaching them to endure the slow, methodical pace of physical reality.</p><p>Troubleshooting a physical build requires a different kind of intelligence, one that involves touch, spatial reasoning, and emotional regulation. Building Tinkerer’s Resilience ensures that when the algorithm fails, the student does not.</p><p>How do we know if a student has successfully developed a Techno-Artisan ethos?</p><p>We will know it when their primary question shifts. We will know it when they stop looking at a screen and asking, <em>“What prompt should I use?”</em> and start looking at their workspace and asking, <em>“Now that the AI has drafted the blueprint, how do I actually build this?”</em></p><p>The future belongs to the students who are unafraid to command the algorithm, but remain deeply, stubbornly grounded in the dirt.</p><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/cultivating-a-techno-artisan-ethos"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7f7e81dfa683" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is “Maker Education” About to Make a Comeback?]]></title>
            <link>https://efemisirli.medium.com/is-maker-education-about-to-make-a-comeback-a2624d8f1bd2?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/a2624d8f1bd2</guid>
            <category><![CDATA[maker-movement]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <category><![CDATA[maker-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Fri, 12 Jun 2026 07:57:08 GMT</pubDate>
            <atom:updated>2026-06-12T07:57:08.176Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Y9oGHv2DbqQJ4yiv.png" /></figure><p>If you walked into an EdTech conference in 2015, the <strong>“Maker Movement”</strong> was the main event. 3D printers were humming, laser cutters were buzzing, and we talked endlessly about Seymour Papert’s <em>Constructionism</em>. We believed that learning happened best when you were building something tangible.</p><p>Then, the hype cooled. Coding bootcamps took over. The 3D printers gathered dust in the corner of the library. The focus shifted entirely to the screen.</p><p>Fast forward to 2026. We are currently watching Generative AI devour the very skills we told students were “future-proof.” Writing code, summarizing text, analyzing spreadsheets-the bread and butter of the “Knowledge Worker”-are becoming commodities. The marginal cost of cognitive labor is dropping to zero.</p><p>But here is my hypothesis: <strong>This devaluation of the digital is exactly what will trigger a massive comeback for Maker Education.</strong></p><p>We are entering the era of the <strong>Techno-Artisan</strong>.</p><p>For fifty years, the economy rewarded people who could sit at a desk and manipulate symbols (words, numbers, code).</p><p>Generative AI has effectively “solved” symbol manipulation.</p><ul><li>Need a React component? Claude can write it.</li><li>Need a literature review? Gemini can draft it.</li><li>Need a marketing strategy? ChatGPT can outline it.</li></ul><p>When “knowing” becomes cheap, the market value of pure knowledge work collapses. In a world of infinite digital abundance, scarcity shifts to the one place AI cannot reach: <strong>Physical Reality.</strong></p><p>This brings us to a new archetype for the future of work: The Techno-Artisan.</p><p>This is not a return to the pre-industrial “craftsman.” We aren’t talking about rejecting technology to weave baskets by candlelight. The Techno-Artisan is a hybrid. They are a “Cyborg Craftsman” who leverages high-level AI for design and iteration, but relies on high-level physical skill for execution and verification.</p><p>Why is this the winning combination?</p><p><strong>1. The “Hallucination” Filter</strong> You can hallucinate a legal brief or a Python script, and it might take hours to realize it’s broken. But you cannot hallucinate a chair. If you 3D print a gear with the wrong tolerance, it jams. Physics is the ultimate fact-checker. The Techno-Artisan uses AI to generate 50 variations of a design in minutes, but uses their physical intuition to know which <em>one</em> will actually stand up to gravity and friction.</p><p><strong>2. The “Last Mile” Problem</strong> AI operates in the cloud. We live in the dirt. An AI agent can design a perfect sustainable housing unit, but it cannot navigate the messy, chaotic reality of a construction site to assemble it. It cannot feel that a screw is stripping the wood. It cannot sense that the material is warping due to humidity. The Techno-Artisan bridges the gap between the pristine digital model and the messy physical world.</p><p>This economic shift demands a pedagogical shift.</p><p>For years, Maker Spaces in schools were treated as “fun centers” or “break rooms” from the <em>real</em> learning happening in the classroom. Now, that logic has flipped.</p><ul><li><strong>The Classroom (Screen-based):</strong> Where students learn tasks that AI can already do better (summarizing, essay writing).</li><li><strong>The Maker Space (Physical-based):</strong> Where students learn the one thing AI cannot do: <strong>Embodied Agency.</strong></li></ul><p>Maker Education is making a comeback not because it’s “cute,” but because it is an economic hedge. It teaches students to be comfortable in the physical world-to understand materials, electronics, tension, and thermodynamics not just as equations, but as tactile realities.</p><p>The most valuable students of 2030 won’t be the ones with the highest GPA in Computer Science, nor the best woodworkers.</p><p>They will be the students who can prompt an LLM to write the C++ code for a robotic arm, and then have the manual dexterity to solder the circuit board that powers it.</p><p>We don’t need to choose between the screen and the workshop. The <strong>Techno-Artisan</strong> owns both. And that is why the Maker Movement is about to have its second-and much more important-act.</p><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/is-maker-education-about-to-make"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a2624d8f1bd2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[EdTech’s Definition in 2026: A 60-year tour through our field’s identity crisis]]></title>
            <link>https://efemisirli.medium.com/edtechs-definition-in-2026-a-60-year-tour-through-our-field-s-identity-crisis-e98298dc978e?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/e98298dc978e</guid>
            <category><![CDATA[education-technology]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Thu, 11 Jun 2026 17:56:05 GMT</pubDate>
            <atom:updated>2026-06-11T17:56:05.473Z</atom:updated>
            <content:encoded><![CDATA[<h4>What 60 years of official definitions reveal about us.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dAEG0cXz581MX4bE.png" /></figure><p>Name and textual definition of our field is not a terribly urgent topic but today, I do not want to do another post about AI in education. What I want to share with you today stems from the fact that I still fumble when someone asks what I do.</p><p>Is it Educational Technology? Instructional Technology? Instructional Design and Technology? Depending on which university department you ask, the answer changes. Some spell it “Technology.” Others go with “Technologies.” If we can’t settle on a name, what does that say about what we actually are?</p><p>This isn’t a minor branding problem. It follows our graduates into the real world. Are we coding teachers? E-learning content creators? The de facto IT department at schools? The identity crisis runs from the department door all the way to the job posting.</p><p>So I went back to the sources as much as I can, to the the official definitions our field has produced since 1963 and held them up against my own intuitions.</p><p><strong>1963 — Department of Audio-Visual Instruction (DAVI, precursor of AECT)</strong></p><p>Historically interesting, but narrow. This reads as a definition of instructional design not a whole field. The focus on “controlling the learning process” hasn’t aged well either.</p><p><strong>1970 — Commission on Instructional Technology (CoIT)</strong></p><p>My favourite of the bunch. The Gestalt framing “more than the sum of its parts” is exactly right. It also explicitly includes human resources alongside technology, which matters. The weak spot: calling it “a systematic way of doing things” makes the field sound like a trade rather than a discipline.</p><p><strong>1977 — AECT</strong></p><p>Systems thinking quietly disappears here, replaced by “process.” But “all aspects of human learning” is the kind of broad, ambitious scope our field actually deserves.</p><p><strong>1994 — AECT</strong></p><p>A real step forward: framing the field as both <em>theory and practice</em> finally gives us intellectual legitimacy alongside practical application. No systems thinking, but the framing is more honest than most.</p><p><strong>2007 — AECT</strong></p><p>The word “ethical” shows up for the first time here and I find it redundant. Every field is supposed to be practiced ethically; calling it out specifically feels like an insecurity, not a value. (It turned out AECT later developed formal ethical standards, which explains the emphasis.) “Performance” as a goal alongside learning is a useful addition, though it’s hardly unique to our field.</p><p><strong>2017 — AECT</strong></p><p>More elaborate than 2007. Not necessarily clearer.</p><p><strong>2023 — AECT</strong></p><p>“Evaluation” makes a quiet comeback after being absent since 1977. “Empower learners” is a welcome addition in spirit. But the core problem remains: this is still a list of things practitioners <em>do</em>, not a statement of what the field fundamentally <em>is</em>.</p><p><strong>The Definition by Reiser &amp; Carr-Chellman</strong></p><blockquote>“The field of instructional design and technology encompasses the analysis of learning and performance problems, and the design, development, implementation, evaluation, and management of instructional and noninstructional processes and resources intended to improve learning and performance in a variety of settings.”</blockquote><p>“Encompasses” is doing a lot of heavy lifting here. Again job description, not field definition.</p><p>Here’s what’s missing from all of these debates: <strong>Does science precede technology, or is it the other way around?</strong></p><p>The default assumption is that knowledge comes first. We discover how people learn; then we build tools and systems around that knowledge. Under this view, EdTech is simply the application of learning science. Downstream. Derivative.</p><p>But Heidegger argued the opposite that technology isn’t applied science at all. It’s a way of framing and revealing the world that actually <em>shapes</em> what questions science ends up asking.</p><p>If he’s right, EdTech isn’t downstream of cognitive science or instructional psychology. It’s a generative field, one where practice creates the conditions for theory, not the other way around. We don’t just use knowledge about learning. We produce it.</p><p>That’s the philosophical ground on which I build my own definition.</p><blockquote><strong><em>Education Technology is a scientific discipline where systems thinking is applied to identify, analyse, and solve problems in all aspects of human learning through the utilization of technological resources.</em></strong></blockquote><p>It’s not perfect. But it does things the official definitions don’t: it centres <em>systems thinking</em>, keeps humans in the frame alongside technology, and defines the field by what it <em>is</em> not just what its practitioners do on a Tuesday.</p><p>I’d appreciate if you can provide some insight about at least one questions below:</p><ol><li>Do you buy the Heidegger argument that technology shapes science, not the other way around? And if so, what does that mean for how we position EdTech?</li><li>Is “ethical” a meaningful addition to a field definition or a symptom of the identity crisis itself?</li></ol><p><em>Originally published at </em><a href="https://efemisirli.substack.com/p/edtechs-definition-in-2026-a-60-year"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e98298dc978e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Cognitive Offloading is Revealing the Artist in Me]]></title>
            <link>https://efemisirli.medium.com/how-cognitive-offloading-is-revealing-the-artist-in-me-c4640b6f392f?source=rss-ed5641a2b111------2</link>
            <guid isPermaLink="false">https://medium.com/p/c4640b6f392f</guid>
            <category><![CDATA[educational-technology]]></category>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[ai-in-education]]></category>
            <dc:creator><![CDATA[Efe MISIRLI]]></dc:creator>
            <pubDate>Wed, 10 Jun 2026 06:21:57 GMT</pubDate>
            <atom:updated>2026-06-10T06:21:57.838Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*aqDoWNSrfK8LATmy.png" /></figure><p>And yes, I know this is not entirely innocent. Academic writing is supposed to be where you discipline your own thought, struggle with concepts, earn clarity slowly, and suffer just enough to feel that the degree is legitimate. So when AI begins to take over parts of that rational and compositional labour, a small academic policeman inside me starts whispering: “Is this allowed? Are we still thinking? Should we be worried?” Annoyingly, the policeman may have a point.</p><p>I began asking different questions. Not only: “Is this argument correct?” but also: “Does this paragraph breathe?” “Is this formulation alive or just wearing a suit?” “Does this sound like me, or like a committee of polite robots?” “Is the idea actually sharp, or is it merely fluent?” AI made textual production abundant. <strong>Suddenly, I had versions, alternatives, structures, formulations, transitions.</strong> And when production became abundant, judgment became my real work.</p><p>In practice, this means that writing with AI is not simply pressing a button and receiving wisdom from the cloud. It is a <strong>constant process of selection and refusal.</strong> This paragraph is too generic. That sentence is elegant but false. This transition is smooth but empty. That formulation sounds impressive but says absolutely nothing, which, to be fair, means it could probably survive peer review. The work becomes less about producing language and more about sensing whether language has become meaningful.</p><p>At this point, I had the pleasant illusion that I had discovered something deep and original. Annoyingly, it turns out that smarter people than me already discovered that and published a paper on it. Luciano Floridi calls this shift <strong>“distant writing”</strong>: a form of writing in which the human author is no longer simply the person manually producing every sentence, but becomes more like a designer of textual possibilities ( <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5232088">link to the paper</a>). The human sets the intention, constraints, style, and direction; the <strong>AI generates; the human then reviews, refines, curates, and takes responsibility for the final text. </strong>Floridi also uses the term <strong>“wrAIting”</strong>, while admitting that the word is awful and should ideally not catch on, which of course makes it almost impossible not to use.</p><p>What I like about Floridi’s idea is that he does not see AI writing only as a way to get things done faster. Of course, AI does help with speed. But something more interesting is happening. Writing starts to feel less like sitting alone and producing every sentence by hand, and more like designing the conditions for good text to appear. <strong>You give direction, test different versions, notice what does not work, ask for changes, reject weak passages, keep the useful ones, and slowly shape the final piece.</strong> I am not writing fiction; I am writing a PhD dissertation. Still, the experience feels strangely similar. The more AI handles parts of the drafting, the more I find myself making judgments: Is this clear? Is it shallow? Is it precise? Does it sound fake? Does it feel alive? Does it actually say what I mean?</p><p>Of course, this is where we need to be careful. AI is very good at producing fluent text. Sometimes too good. <strong>It can produce a paragraph that looks polished, balanced, and confident while quietly containing almost no real insight. </strong>This is dangerous because bad writing is easy to detect, but shallow fluent writing is a much sneakier creature. It enters the room wearing a blazer and speaking in complete sentences.</p><p>But the answer cannot simply be: “Never use AI.” <strong>That ship has not only sailed; it is already offering conference discounts and institutional subscriptions.</strong> The more realistic question is: <strong>what human capacities become more important when AI can produce plausible academic text?</strong> My answer, at least for now, is judgment. Taste. Discernment. The ability to say, “ <strong>No, this is not it,”</strong> even when the paragraph sounds perfectly respectable.</p><p>In AI-assisted writing, <strong>refusal</strong> becomes a central intellectual act. <strong>Rejecting a sentence can be as important as producing one.</strong> The scholar must become sensitive to what is inflated, generic, shallow, forced, imprecise, or lifeless. <strong>This is where the “artist inside the scholar” appears: not as someone who abandons reason, but as someone who gives reason form.</strong> The task is not just to make arguments. It is to shape them until they become clear, honest, elegant, and recognizably one’s own.</p><p>This has implications beyond my own dissertation-induced existential drama. If generative AI changes writing from simple production to judgment among possibilities, then education has to take that shift seriously. We cannot keep pretending that the main challenge is getting students to produce text from scratch, as if AI has not quietly entered the classroom, sat at the back, and started helping everyone with their homework.</p><p>So perhaps cognitive offloading did not turn me from a scholar into an artist. That would be too dramatic, even for a PhD student. Rather, <strong>it revealed that the scholar had always needed an artist inside him. </strong>The artist is the part that senses rhythm, proportion, elegance, falseness, excess, and form. The scholar asks, “Is this true?” The artist asks, “Has this truth found its proper shape?” Academic writing, at its best, needs both.</p><p>I began by worrying that AI was taking over part of my rational mind. I still think that worry is valid. But I also think something else happened. As AI took over some of the labour of producing text, it forced me to listen more carefully for voice, coherence, and authenticity. It made me notice when an idea was technically present but spiritually absent. <strong>And occasionally, after many prompts, revisions, rejections, and minor acts of academic despair, it helped me recognize the strange moment when a sentence finally felt like my own.</strong></p><p><em>Originally published at </em><a href="https://substack.com/home/post/p-198649950"><em>https://efemisirli.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4640b6f392f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>