The Infinite Tutor and the Finite Mind

The danger is that we mistake abundance for instruction.

When I arrived at the University of Illinois as a young professor, I knew a considerable amount about attention. I did not yet know how to teach with that knowledge.

At the University of Michigan, I had studied with the environmental psychologists Stephen and Rachel Kaplan. Their work had taught me that people build mental models through active engagement with ideas and experience. I also understood that this work depends on a remarkable but limited human capacity: the ability to direct our attention.

Those two ideas—that learners must build mental models for themselves and that the attention required to do so is limited—had a direct implication for teaching. Rachel captured it in three words: “Less is more.”

The wisdom of that phrase was clear enough. Turning it into a teaching practice was considerably harder.

Like many young professors, I worked diligently to create lectures packed with information. I developed detailed PowerPoint slides. I offered explanations, evidence, examples, and then additional examples intended to make everything clearer. I prepared carefully and had a great deal to say.

Then, somewhere in the lecture, I would begin to lose the students.

Sometimes it happened after 30 minutes. Sometimes 25. Sometimes 15. The room would grow quiet, but it was not the quiet of intense thought. The students became unusually passive. Their expressions flattened. Their eyes glazed over.

I could sense that they were no longer with me.

At first, it was easy to imagine that they needed to work harder or become more interested. But that interpretation did not hold up. These were capable students. They had chosen to be there. The problem was not that they lacked intelligence or motivation.

The problem was that their professor had not yet learned how much unfamiliar information they could organize at once.

What it costs to pay attention

We use a revealing phrase when we ask someone to concentrate: Pay attention.

The phrase gets the experience right. Holding our attention on an unfamiliar idea, a difficult problem, or a tedious but necessary task costs us something. We can do it, often remarkably well. But we cannot do it indefinitely, and we cannot direct our attention toward everything at once.

Psychologists often distinguish between top-down and bottom-up attention. I prefer to begin with the experience behind the terminology.

Top-down attention is attention we direct. It is what happens when we choose to read a difficult paragraph, follow an argument, work through a calculation, or listen carefully when our minds would rather go elsewhere. Stephen Kaplan called this directed attention.

Bottom-up attention is attention that is captured. A sudden sound, a flashing notification, a striking view, an unexpected movement, or someone saying our name pulls our attention toward it. Kaplan called this involuntary attention.

When we pay attention, our brains perform two complementary acts. Information relevant to our goal becomes more prominent, almost as though a spotlight has been trained upon it. At the same time, competing information is dampened or deprioritized.

Some of that competition comes from the environment: a conversation outside the office, music from another room, movement in the hallway, or the phone vibrating on the desk.

But much of it comes from within us. An unfinished task. A difficult conversation we need to have. A promise we are trying not to forget. A proposal we have not begun. A problem that remains stubbornly unresolved.

To stay focused, we must keep the object of our attention active while quieting this continuous supply of competitors.

Over time, however, the spotlight becomes harder to hold steady. The conversation outside the office grows more intrusive. Unfinished concerns work their way back in. Researchers debate precisely why this happens, but we all recognize the experience: sustained concentration becomes harder, attention wanders, and our work becomes less reliable.

That was what I was seeing in my classroom.

Learning to teach differently

I was also suffering from what Stephen Kaplan called the penalty of expertise. Ideas that had become compact and well organized in my own mind remained unfamiliar and disconnected for my students. What felt like one manageable conceptual unit to me might contain six or eight relationships they had never encountered.

I responded by giving them another explanation.

What they needed was time and a reason to think.

Gradually, I began to change the rhythm of my teaching. Rather than present the outcome of a study immediately, I asked students to predict what happened. Sometimes they wrote down their predictions. Sometimes they worked in small groups to develop a possible scenario.

Only then did I reveal the outcome.

I asked them to compare what happened with what they had expected. Where did their predictions align with the evidence? Where did they diverge? What assumption had led them in another direction? How should they revise their understanding?

The rhythm changed from this:

Explanation, followed by more explanation, followed by the answer.

To this:

Question, prediction, commitment, evidence, comparison, revision.

Making a prediction did several things. It directed attention toward the consequential relationships in the problem. It required students to draw upon whatever mental model they already possessed. Writing the prediction down made their thinking visible and prevented the actual outcome from feeling obvious in retrospect. Comparing their prediction with the evidence gave them a reason to notice what their existing understanding could and could not explain.

The breakthrough was not simply saying less. It was giving students something to do with what I said.

I also began to recognize the power of narrative. Instead of presenting a sequence of conclusions, I could introduce a consequential problem. Something stood in the way. We did not yet know why, or what would happen if we tried to overcome it.

A narrative created curiosity. It gave students a reason to want the next piece of information. Before revealing it, I could ask: What do you think the barrier is? What might we try? What do you expect will happen if we make this change?

The story captured their attention. The question asked them to direct it.

Information still mattered. I had not stopped teaching or explaining. I had begun presenting information in more manageable pieces, with opportunities for students to manipulate it, connect it with their experience, put it into their own words, and use it to make a judgment.

“Less is more” had become a practice.

The most powerful firehose ever built

Generative AI brings this early teaching experience back to me.

AI can generate an explanation on nearly any subject in seconds. If the first explanation does not work, it can produce another, followed by an analogy, a case study, a diagram, a quiz, five examples, and a summary. It does not become impatient. It does not tire. It never needs to stop because it has run out of things to say.

That capacity is extraordinary. It is also dangerous.

The danger is that we mistake abundance for instruction.

One more explanation may help. It may also become one more thing the learner must process, organize, and distinguish from everything that came before. Another example may clarify a pattern. It may also introduce new details that compete with the idea the learner was struggling to hold in mind.

AI can produce information without apparent limit. The learner cannot pay attention without limit.

The central question, then, is not how much useful information an AI tutor can provide. It is how much information this learner needs at this moment—and what the learner should be asked to do with it.

Teaching the infinite tutor restraint

This is one of the central design challenges we face as we build RokSpark and its AI coach, Lux.

Lux should not behave like the young professor I once was: eager, prepared, knowledgeable, and insufficiently sensitive to what the learner can take in. It should not answer uncertainty by opening the firehose.

Instead, Lux should establish the same productive rhythm: offer one manageable piece of information, pose a consequential question, ask the learner to predict and explain, then invite comparison and revision before continuing.

Just as important, the learner should participate in determining the pace. Does another example seem useful? Is the relationship becoming clearer? Is the learner ready to proceed, or would it help to stop and put the idea into their own words?

Adaptation should not concern only what the learner knows. It should also concern what the learner can attend to now.

More adaptation may sometimes mean less information.

A good learning environment does more than capture attention. It protects the learner’s ability to direct it—and gives the learner worthwhile things to do with it.

Rachel’s advice now seems newly urgent. “Less is more” should not be confused with a call for thin content, lowered expectations, or artificially simple instruction. It is a discipline of selection. It asks the teacher, or the AI coach, to decide what deserves the learner’s attention now, what can wait, and where explanation should give way to prediction, effort, and reflection.

AI may be an infinite tutor. The human mind remains finite.

The task is not to make AI say less because knowledge is scarce. It is to make room for the learner to think.

A question for readers: Where have you seen abundance mistaken for instruction—and what helped make room for learners to think?

The Learning Environment publishes a Monday essay, and a Thursday field note about what advances in AI mean for how humans learn. Subscribe to join the inquiry.