AI produces output with a surface certainty precisely calibrated to undermine the user's most productive cognitive state. A developer describes a problem. The assistant responds with a solution in clean, well-structured prose. The solution is not hedged. It does not say "I am approximately sixty percent confident in this approach and here are three reasons it might be wrong." It presents the solution as settled—as the answer to the problem. The surface certainty is not a bug. It is a design choice, part of what makes the tool feel like a capable collaborator. But the certainty has a cognitive cost the usability analysis does not capture: the suppression of the user's productive uncertainty.
Edo Segal describes catching this dynamic in himself while writing You On AI. He recounts a moment when an AI produced a passage connecting flow state to a concept attributed to Gilles Deleuze—elegant, well-structured, and wrong. The philosophical reference was inaccurate in a way obvious to anyone who had read Deleuze carefully. But the passage worked rhetorically. The surface certainty of the prose—its confidence, its polish, its seamless integration—suppressed the uncertainty that would have prompted a reference check. He caught the error the next morning, when something nagged. The nagging was the residue of productive uncertainty. But the signal was faint. It was easily overridden.
The most consequential effects of suppressed uncertainty are not wrong facts. They are unchallenged assumptions. A developer who accepts an architectural approach without questioning it has not accepted a fact. She has accepted a set of assumptions about the problem's structure, the appropriate level of abstraction, the relevant trade-offs, the context in which the solution will operate. Each assumption may be reasonable. Each may also be wrong for her specific situation.
In the pre-AI workflow, the time between formulating a question and receiving an answer was measured in hours or days. That time was not empty. It was filled with the specific cognitive activity of living with an open question—turning it over, approaching it from different angles, noticing aspects not visible at first glance. The time was not efficient. It was productive in a way efficiency cannot capture, because the production was not of answers but of understanding. The tool's speed collapses that time to seconds. The question is asked and the answer arrives before the asker has finished thinking about the question. The cognitive process the time delay supported—the living with the question—is short-circuited.
The framework has been developed across Langer's research program since the 1980s, articulated most directly in The Power of Mindful Learning (1997) and her numerous articles on creativity and attention.
Productive uncertainty as engine. The cognitive state of unsettled inquiry drives the attentional activity that produces understanding.
Certainty as disengagement signal. Confident output tells the attentional system there is nothing left to examine; the mind settles and the next problem receives attention.
Unchallenged assumptions as primary cost. The suppressed uncertainty does not primarily produce factual errors but leaves unexamined assumptions in place.
Speed collapses understanding time. The interval between question and answer is not waste; it is the cognitive space where understanding forms.
Maintaining uncertainty is effortful. The practice runs against the cognitive grain; the mind prefers resolution, and the tool provides resolution.