From Sage On The Stage To Architect Of Learning
Picture a university student—call her Sara—sitting at her laptop at 11 pm, three days before a midterm exam. She has watched every lecture, downloaded every slide deck, and highlighted her notes until the pages are more yellow than white. She understands the material, more or less, in the way you understand a city you have only ever seen on a map. When the exam arrives, the map will not be enough.
Sara's situation is not unusual. It is the dominant mode of learning across higher education and most online platforms. Content is abundant. Genuine understanding—the kind that survives three weeks and transfers to a new problem—is far rarer. Completion rates for online courses hover below 15%, according to MIT and Harvard researchers studying MOOCs. Students enroll with real intent, then drift away. The content was never the problem. The design was.
Generative AI has introduced something the traditional classroom never fully could: a learning environment that adapts to each individual, available at any hour, and capable of meeting a learner precisely where they are. The question is no longer whether AI belongs in education. It is whether the platforms deploying it understand enough about how humans actually learn to use it wisely.
What AI Platforms Can Do That Classrooms Struggle To
Think about what a skilled private tutor actually does. They notice when you hesitate before answering. They remember that two weeks ago, you confused two related ideas and quietly circle back to test whether that confusion has resolved. They adjust, in real time, to the specific shape of your understanding.
A teacher managing thirty students cannot realistically do all of that—not because teachers lack skill, but because structural arithmetic does not allow it. A well-designed AI system can. It tracks which learners need more retrieval practice, which have a persistent misconception, and which are disengaging—simultaneously, across an entire cohort.
The efficacy record here is meaningful. A 2016 meta-analysis by Kulik and Fletcher in the Review of Educational Research examined 50 controlled studies of intelligent tutoring systems and found effect sizes averaging 0.66 standard deviations above control conditions. Benjamin Bloom's foundational 1984 research on the "two-sigma problem" showed that one-on-one tutoring outperformed conventional classroom instruction by two standard deviations—a gap that was economically unscalable until now. AI tutoring does not fully replicate a great human tutor, but it moves the needle on an access gap that education systems have spent forty years unable to close.
This is also where AI platforms become relevant, not as a novelty but as a structural response. By allowing learners to generate courses from their own materials and providing contextual AI assistance calibrated to personal content, these platforms shift the dynamic from passive consumption to active construction—the kind of engagement that learning science consistently associates with deeper retention.
Why Motivation Is The Wrong Target—And Habit Is The Right One
Here is where most of the EdTech industry has made a consequential mistake. The dominant design philosophy in consumer-facing platforms has been engagement optimization: streaks, badges, leaderboards, notifications timed to pull you back. The assumption is that motivated learners keep learning. It is an assumption that flatters the product and fails the person.
Motivation is not a stable resource. It fluctuates with mood, stress, and circumstance. The person fired up to study on Sunday afternoon is frequently not the same person who can summon that energy on Wednesday evening after a difficult day. Designing a learning system around motivational peaks is designing for a version of the learner that does not reliably show up.
Self-determination theory, developed by Deci and Ryan, makes the problem more precise: extrinsic motivation—driven by rewards and social pressure—tends to crowd out intrinsic motivation once the external trigger is removed. The learner who studied daily to maintain a streak may find, when the streak breaks, that they have no internal reason to return.
The more durable target is habit. Research by Wendy Wood and colleagues on behavioral automaticity shows that habits—routines triggered by context cues rather than deliberate motivation—are far more stable predictors of sustained behavior. A learner who has built a consistent study habit does not require a motivational state to begin. The cue triggers the routine. The routine becomes self-sustaining.
This is the design philosophy AI platforms should be built around. Rather than competing for motivational engagement, their architecture should target the formation of sustainable study habits—behaviors that persist independently of whether the learner feels particularly energized on a given day.
Usability research conducted by Kampster with students enrolled at the London School of Economics in 2025 indicated that learners clearly distinguished between short-term engagement mechanics and systems designed for durable learning. Hence, a methodological standard the EdTech sector urgently needs: building on cognitive science first, then pressure-testing design decisions through structured research with demanding, analytically trained users.
Bjork and Bjork's work on "desirable difficulties" reinforces why this matters. Conditions that feel easy—passive re-reading, content pitched below current ability—produce weak long-term retention. Effortful retrieval and spaced repetition produce durable learning precisely because they feel harder. A platform optimized for satisfaction scores delivers the former. A platform designed around retention chooses the latter, even when it is the less immediately rewarding option.
The Educator's New Role
None of this makes the teacher obsolete. It changes what a teacher's best hours are spent doing.
If AI handles retrieval scheduling, adaptive feedback, and first-pass concept explanation, the educator's irreplaceable contribution shifts toward something harder to automate: the relational dimensions of learning, the mentorship that connects academic content to a student's sense of identity, the ability to notice that a quiet student is not disengaged but struggling. These are not peripheral to education. In many cases, they are the point.
The OECD's 2023 report Teachers as Designers of Learning Environments frames this precisely: educators increasingly functioning as learning architects, designing experiences rather than delivering content. It is a more demanding role, not a lesser one—and it requires institutions to invest in teacher development rather than treating AI as a cost-reduction instrument.
Conclusion
Return to Sara at her laptop. What she needed was not more content. She needed a system that had been helping her retrieve, space, and struggle productively with material over the preceding weeks—doing the unglamorous work of building real retention, not just the surface impression of familiarity.
That system is now technically possible to build at scale. The cognitive science behind it is not new. What has changed is the capacity to act on it accessibly, affordably, and in a way that adapts to the individual rather than the imagined average learner. The platforms taking this seriously—designing around habit over motivation, retention over engagement—are working on the right problem. So are the educators learning to work alongside them.
References:
- Ho, A. D., et al. 2014. "HarvardX and MITx: The first year of open online courses." HarvardX and MITx Working Paper No. 1.
- Kulik, J. A., & Fletcher, J. D. 2016. "Effectiveness of intelligent tutoring systems: A meta-analytic review." Review of Educational Research, 86(1), 42–78.
- Bloom, B. S. 1984. "The 2 sigma problem: The search for methods of group instruction as effective as one-to-one tutoring." Educational Researcher, 13(6), 4–16.
- Deci, E. L., & Ryan, R. M. 1985. Intrinsic motivation and self-determination in human behavior. Plenum Press.
- Deci, E. L., Koestner, R., & Ryan, R. M. 1999. "A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation." Psychological Bulletin, 125(6), 627–668.
- Wood, W., & Neal, D. T. 2007. "A new look at habits and the habit-goal interface." Psychological Review, 114(4), 843–863.
- Wood, W., Quinn, J. M., & Kashy, D. A. 2002. "Habits in everyday life: Thought, emotion, and action." Journal of Personality and Social Psychology, 83(6), 1281–1297.
- Bjork, E. L., & Bjork, R. A. 2011. "Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning." In M. A. Gernsbacher et al. (Eds.), Psychology and the real world: Essays illustrating fundamental contributions to society (pp. 56–64). Worth Publishers.
- Ebbinghaus, H. 1885. Über das Gedächtnis [Memory: A contribution to experimental psychology]. Duncker & Humblot.
- OECD. 2023. OECD education at a glance 2023. OECD Publishing.
- OECD. 2023. Teachers as designers of learning environments: The importance of innovative pedagogies. OECD Publishing.
- UNESCO. 2023. Guidance for generative AI in education and research. UNESCO Publishing.
