How a children’s educational clip and a spinning chair led me to Anchored Recalibration Learning, a method for tackling complex self-directed learning by starting from curiosity and adjusting the difficulty until it sticks.
I recently watched an animated clip and could not stop thinking about it. Two avatars stand at the base of identical steep blocks. A banana sits on top of each one—the reward. Each avatar holds the same long log and nothing about the challenge differs between them. What differs is how each avatar chooses to use what it was given.
The first avatar plants one end of the log on top of the block before walking straight to the top. It reaches the banana and eats it. The second avatar puts on a helmet, picks up a saw, and cuts the log into equal pieces to build a staircase before climbing to the banana.
Neither avatar is wrong. The first is solving a problem it does not need to own. The second is slower and only makes sense for someone who can afford to wait and cares what happens after the banana is gone. The real failure, in an AI-assisted learning context, is not choosing speed over depth or vice versa. The failure is not knowing which one is being chosen and why.
A few hours before watching the clip, I was at a local science museum, sitting in a spinning-chair exhibit, photographing a plaque on gyroscopic precession written in Japanese, a language I cannot read. I admit that I had skipped that physics class in high school, but I still remember how interesting the physics lessons were.
So I asked an AI model, using the photo, to explain why I spun when I tilted a spinning wheel by its axle. The answer described a wheel that resists changes to its orientation, so a push on its axis does not simply tilt it but produces a reaction redirected sideways, perpendicular to the push.
To be honest, those years of physics classes were far behind me, so I needed to make sense of what the model was telling me—I needed to break down the long log.
I asked for an image showing the perpendicular relationship, then further explanations about angles, reaction, and force components. And, as I slowly grew more confident in my new vocabulary and understanding, I started proposing hypotheses and testing them. I was wrong many times before I was finally right.
I enjoyed that conversation more than I expected. It taught me more about torque, precession, and angular momentum than I ever learned in high school. And the whole experience was more than I had anticipated when I unsuspectingly stepped into the museum with my children on that hot summer Sunday.
My children, now old enough to explore museums on their own, gave me nearly an hour to be wrong and curious. But, more importantly, that conversation left me with questions: Could I have become a physicist if I had access to AI in high school? Would I have aced every test? Could I have become the next Stephen Hawking or even Newton?
A Name for the Slower Path
To be honest, I am a dreamer and AI allows me that. But I know my limitations. I doubt access to AI alone would have made me the next Einstein. But perhaps my friend sitting next to me, the one who always learned fast and got it right on the first try, could have.
In any case, within my own area of interest, which is forming theories about learning, I would like to call this AI-assisted conversational learning experience “Anchored Recalibration Learning.”
Anchored means the learning starts from something a person can see, touch, or find compelling enough to spark inquiry. Recalibration means being able to step back deliberately, without fear of judgment, in a way that matches the learner’s own pace. The term also embeds the idea of scaffolding for new or high-stakes entry points into understanding.
The novelty of this method is that one learner, alone, can now work through difficult cognitive processes much faster and more easily. But neither the method nor the AI tool itself is enough. The learner’s own demand for it is essential.
The learner needs to be genuinely interested in starting and sustaining a focused conversation with a model, sometimes for hours. That is the anchor, and only the learner can drop it. The teacher’s role is to set conditions, time, and rewards, not to force the learning itself. Forced standardized learning, chasing a grade, has little to do with real understanding since it prizes uniformity over the organic curiosity that produces lasting growth.
So everything begins with allowing learners to find a real, meaningful anchor, then letting and guiding them to ask for simpler answers, test acquired knowledge, and let the difficulty rise on its own once the foundation holds.
A Guideline Any Teacher or Self-Taught Student Can Test
Here is the five-step guideline I would suggest to any teacher or curious high school student.
1. Find something genuinely confusing. Photograph, record or just input it, then ask a model to explain it in clear terms. Read carefully, but do not let it overwhelm you. Ask for simpler or more illustrative explanations until the basic idea is solid.
2. Ask for a visual, if shape or motion is involved. Images anchor abstract relationships in memory better than words alone.
3. Push back when something feels dense. Do not just repeat “explain more simply,” engage with what has already been said. Break ideas into smaller pieces or show the model what you already understand so the conversation builds rather than loops.
4. Form and test hypotheses. Try: “Here’s my guess about why this happens, am I right?” Stay close to the concepts you have already learned before venturing further. A model, unlike a teacher managing a full classroom, can test your understanding through real conversation.
5. Generalize without fear. Ask whether the same idea explains a different, unrelated situation. You will either confirm your understanding, expose a gap, or land somewhere unexpected.
Yet even this five-step climb is worth questioning since not every steep block deserves an hour of your Sunday. I would have made the wrong choice in building an AI-assisted staircase if the hour I spent making up for lost time during my high school years could have been better used learning something else or just spending it with my children.
The real failure, then, is failing to weigh what you stand to lose, not only in the future, no longer dreaming of discovering new rules governing the universe, but also right now, missing the chance to jump into a refreshing pool with my children on a scorching day.
Learners willing to slowly cut sticks into pieces should also remember that Anchored Recalibration Learning, as a staged approach to challenging, complex learning, may not always be what they need. Saving time on a task outside your real interest is reason enough not to feel obligated to own every part of a problem.
But if you can commit to it, it produces something worth returning to, the way the second avatar could climb the block faster on a later attempt, because the log had already been cut and understood.
Also, through repetition and practice, this sequence can be encoded and become almost second nature for reaching other rewards as quickly and deliberately as possible. And it will keep proving useful for whatever, adaptive challenge comes next.
Think of coming down from the block. The first avatar in the clip has to jump and might keep experiencing hard falls each time the next log runs out. The second avatar, instead, can rely on the experience of walking back down effortlessly because it took the time to break down complexity and build a heuristic in its cognitive toolbox, one that will prove handy the next time a challenging task appears.
The banana disappears the moment you eat it. The staircase stays and so does everything you learned building it. But the real question is: What do you need, the banana, the staircase, or both? When? Why? How?
ABOUT THE AUTHOR
Marvin Starominski-Uehara
Marvin Starominski-Uehara teaches at Temple University Japan and researches emerging intelligence in complex systems. He is the principal investigator of the Stigmergy Network Theory project and a guest editor for an upcoming special issue of Interface Focus (Royal Society) on collective coordination across biology and digital society. His research sits at the intersection of how individuals and organizations make decisions under uncertainty and what it takes to design conditions where genuine capability becomes visible. He is the creator of Trace Pedagogy and holds a Ph.D. in Risk Management and Public Policy from the University of Queensland and a Master’s in Public Administration with a certificate in Disaster Management and Humanitarian Assistance from the University of Hawai‘i at Mānoa. He is an alumnus of the East-West Center. Marvin’s research findings, applications and insights have been featured in World Economic Forum, The Conversation, LSE Blogs, The Scholarly Kitchen, EDUCAUSE Review, The Diplomat, among others.
Website: https://www.marvinuehara.com
LinkedIn: https://www.linkedin.com/in/marvin-starominski-uehara-0799b8415/
This article is featured in Canadian Teacher Magazine’s Fall 2026 issue.
