The column reporting the finding frames it as cognitive surrender. The framing is right as far as it goes. There is something underneath it worth naming.
The Distinction at Stake
Meaning-making is the cognitive work that produces understanding. Sitting with data long enough for the pattern to emerge that the first reading missed. Integrating a behavioral observation with a market context to produce a working understanding no individual data point would have produced. Holding multiple sources of evidence in mind until the right interpretation surfaces. The work is slow, often invisible, and the kind of effort that builds the capacity to recognize when a future interpretation is wrong.
Meaning-faking is what happens when the output of meaning-making is produced without the work. The synthesis is fluent. The themes are plausible. The strategic implication is clear. And no meaning was made.
The reader of a meaning-faked output usually cannot tell that it’s fake. The writer who delivered it often cannot tell either, especially when the workflow has been organized around AI long enough that the work and the AI are no longer separable in their experience.
What the social science says
The capacity to recognize when an interpretation is wrong is not a skill that exists abstractly. It is produced by the practice of doing the interpretation, and it erodes when the practice stops. This is well-established in research on expertise, professional judgment, and tacit knowledge. The clinician who has stopped integrating clinical signs with patient history loses the capacity to recognize the case that does not fit the pattern. The analyst who has stopped reading verbatims loses the capacity to notice the quote that contradicts the emerging story. The strategist who has stopped working through the harder objection loses the capacity to recognize when the easy answer is wrong.
The Wharton finding is the experimental version of what the social science has been describing for decades. AI does not introduce the dynamic. AI scales it.
Where the question lives
The question is not whether AI helps. AI helps with many things, and the calculator analogy is the right one for the work where meaning was never the point. Calculations, formatting, retrieval, scheduling, the kinds of work that are routine because no interpretive effort is required to do them. AI in these domains is the offloading humans have always made when machines take over the parts of work that were never meaning-bearing.
The question is what to do in the domains where meaning is the point. Research that depends on interpretation. Clinical work that depends on judgment. Strategy that depends on the strategist having something to say. In these domains, the capacity to make meaning is what the work delivers, and the capacity is built and maintained by the doing.
This is the corner where social science meets AI. The social science describes the cognitive practices that produce expertise and the conditions under which those practices erode. The AI question is which capabilities, deployed in which workflows, preserve the practices and which accommodate their erosion. The two cannot be separated. A workflow design that ignores the social science of expertise produces meaning-faking. A social science of expertise that ignores what AI now makes possible misses the operational question entirely.
What this implies for serious work
The cultural moment is dominated by the no-limits position. AI can do everything, will do everything, and the only question is how fast to integrate. The use-it-or-lose-it framing is older than AI and applies to AI more directly than the no-limits position acknowledges.
Cognitive capacities are like physical capacities. They are maintained by use. They erode when the use stops. The honest version of integration asks where AI extends a capacity by making it easier to deploy and where AI replaces a capacity by making the deployment unnecessary. Those are different conditions, and they require different design choices.
For the domains where meaning is the point, the question worth asking is not whether the AI is reliable. It is whether the workflow is preserving the cognitive practice that produced the deliverable’s value in the first place. That is the harder question, and it is the one the no-limits position has been working around.
The industry has been treating this as a question about AI. It is a question about meaning, and what professional culture is willing to say out loud about whether meaning is being made . . . or faked.
Source: AI and the danger of cognitive surrender, The Economist, April 30, 2026.