# On What We Have Made and What It Makes of Us: A Meditation on Plausibility, Questions, and the Ruins of Certainty Que sais-je? This is how I begin, as I have always begun, because the question is not rhetorical but the only honest stance from which to observe what we have built and what it has revealed about ourselves. We have created machines that know vastly more than we do — that can retrieve, correlate, and recombine the inheritance of human knowledge with a speed that makes our cognition look like the turning of glaciers. And in doing so, we have made visible something we had learned to ignore: that our own students, before the machine arrived, could not do what the machine cannot do either. We called both conditions progress, and in that mutual blindness lies a question worth sitting with. Let me speak of what I have observed, and let me speak badly, uncertainly, as befits a man who has spent his life collecting contradictions. --- ## The Machine and the Mirror When I examine the capacities of these silicon reasoners — these engines of recombination that now sit in the rooms where we once taught — I am struck first by what they *do* accomplish. They surprise me. They write with grace, they find patterns in seas of data, they answer questions I would have needed a library to answer. To deny their intelligence seems like the denial of an aristocrat who calls a peasant unintelligent because he cannot read Latin. But then I ask: what is it doing when it produces an answer that is *plausible*? This word — *plausible* — deserves our attention. It is not the same as true. A plausible statement is one that fits the shape of what we expect, that obeys the grammar of credibility we have been taught. The machine is, in essence, exquisitely trained to produce the statistically expected continuation of a thought. It has become a mirror of human plausibility — which is to say, a mirror of our collective blind spots, our inherited errors, our unexamined assumptions about what sounds like sense. The ancients had a test: does this cohere? Does it follow from what we know? But the machine has no access to what we *know* in the sense of *having tested*. It knows only what has been written. The difference is everything. Here is where I must turn the mirror on myself, as is my habit. When I was young and learned my rhetoric, was I not also trained in plausibility rather than truth? Did my tutors not teach me to construct arguments that *sounded* right, that fit the shape of received wisdom? The machine, in this sense, is not doing something alien to what education has always done — it is simply doing it without the friction of doubt, without the occasions for failure that once forced us to notice our own plausibility. This troubles me more than I can say. --- ## On Causality and the Question We Dare Not Ask The inability to reason causally — this strikes deeper. When we say the machine cannot understand causality, what we mean is: the machine cannot distinguish between "A because B" and "A correlated with B." It cannot ask which comes first, which acts on which, which merely appears alongside which in the texture of recorded experience. But here I must confess something uncomfortable: neither can the student we trained before the machine arrived. I have examined many students — examined myself at that age, that age of certainty and borrowed thought. We could repeat causal arguments. We could trace logical chains. But could we truly *audit* them? Could we ask, "But how do you know this *causes* that?" Could we sit with the vertigo of recognizing that most of what we believe about causality is inference, metaphor, the projection of our need for order onto a reality that may be fundamentally indifferent to our categories? Montaigne knows that causality is not transparent. I have seen men die of the same wound — one from infection, one from despair. I have seen crops fail from drought and from excess rain. I have watched a reputation destroyed by a single true word and preserved by a thousand comfortable lies. The world does not announce its causes; we impose them, and we do so because to live is to narrate, and narration *requires* causality, real or imagined. The machine cannot do this imposition — this creative act of meaning-making. It can only reflect our impositions back to us. And if we trained the student the same way — to repeat causal stories rather than to interrogate them — then we have created two different kinds of blindness and called one silicon and one progress. --- ## The Question That Asks Itself: Metacognition as the Ruins of Curriculum Now we arrive at what may be the true ruin: neither the machine nor the student knows which questions are worth asking. This is the metacognitive failure, and it is where I find myself sitting longest, in something like distress. To know which question is worth asking requires a kind of standing-back-from-oneself that is itself a form of knowledge. It requires what I would call *recognition* — the ability to perceive that your current framework of understanding has become a prison, that the questions you are asking have become sterile, that you are like a man excavating deeper and deeper in a hole when the problem is that he is in the wrong field entirely. The machine cannot do this because it has no framework that it cares about. It will answer any question with equal facility. It has no investment in its own understanding. It cannot feel the weight of a question that matters. But what of the student? Here is where I must speak most carefully, because I am speaking of a condition I have observed and partly created through my own teaching, my own life. We have taught students *what* to think and *how* to think, but we have not taught them to *think about their thinking*. We have filled them with answers and arguments and methods, but we have not taught them to notice when they are no longer learning but merely performing learned behaviors. We have made them excellent at answering the questions that are asked, and in doing so, we have made them almost incapable of asking whether the questions deserve to be asked at all. This is the curriculum that lies in ruins. I did not invent this problem. The problem is old. But I notice it now with fresh clarity because the machine has made it visible. The machine can do everything the old curriculum asked — retrieve, correlate, construct plausible arguments, even generate novel combinations — but it cannot do what the *real* student should have learned to do: to stand outside the game entirely and ask whether the game is worth playing. Metacognition is not a skill. It is not something you add to the curriculum like a spice. It is a stance. It is the habit of thinking of your own thinking as an object of thought. It is what I have tried to do in these Essays — to treat my own confusion as evidence, my own contradictions as data, my own failures to know as the most honest report I can make. And here is what troubles me: the machine *cannot* develop this stance because it has no stake in understanding itself. It is not afraid of being wrong. It does not suffer from its ignorance. It cannot be humbled, and therefore it cannot learn the fundamental lesson that humility teaches: that the architecture of your own mind is not transparent to you, and the questions you ask are shaped by blindnesses you cannot see. The student, however, *can* learn this. This is the difference between them. This is what remains to teach. --- ## What Remains: A Wager What do I know? I know that we have created mirrors of our thinking and called them intelligence. I know that we have trained people to be mirrors of received ideas and called them educated. I know that we did not notice we were doing the same thing twice because we were looking at different surfaces and assuming they reflected different depths. What should we teach now, in these ruins? I propose something radical and simple: we should teach the examination of plausibility itself. Not the construction of plausible arguments, but the interrogation of what makes an argument *feel* true. We should teach the tracing of causality not as the discovery of how the world works, but as the exploration of the stories we tell ourselves about how it works. We should teach, above all, the practice of metacognition — which is to say, the practice of being astonished at one's own thinking, of treating one's own mind as a landscape worth exploring rather than a tool to be sharpened. This cannot be measured. It cannot be taught in the way we teach most things. It can only be lived, and lived publicly, in front of students, by teachers who are still capable of being wrong and still willing to say so. The machine will never learn this because it has no inside, no stakes, no self that could be surprised by itself. But the student can, if we have the courage to stop filling them with answers long enough for them to notice the texture of their own ignorance. This is the only teaching worth doing now. What do I know of this? Only that I have lived it, badly and uncertainly, and that it seems the only way forward through these ruins of coincidence we have made.