The Lemons Problem for Expertise

In The You On AI Field Guide

The market for expertise, before AI, had imperfect but functional quality signals. Production cost was one: a brief that took forty hours suggested engagement with the material that a four-hour brief could not reproduce, and the hours carried informational value even if the client could not evaluate the work directly. Credentials were another: a partner at a top firm signaled training and selection that a junior associate could not match, and the credentials provided proxy information about quality. Reputation was a third: a lawyer's history of successful outcomes provided evidence about capability even when individual outputs were difficult to evaluate. Each signal was noisy; each was better than nothing; together they allowed markets for expertise to function despite the underlying information asymmetry.

AI collapses the signal. Production cost loses informational value when cost no longer reflects hours: the AI-assisted brief produced in four hours was not tested against the same depth of understanding as the independent brief produced in forty, but the hour difference is no longer visible to the client. Credentials retain some value but are diluted as AI enables junior practitioners to produce output formerly associated with senior credentials. Reputation operates on longer timescales than AI adoption, producing transitional periods during which established reputations continue to function as signals even as the underlying capability has shifted. The result is an information environment in which the quality of expertise is more opaque to buyers than at any prior point, and the lemons dynamic accelerates.

The consequences for deep practitioners are direct. A lawyer who invested fifteen years building genuine expertise — reading cases, developing judgment through thousands of hours of patient work — cannot compete on price with a lawyer who invested less and supplements with AI. The first lawyer's depth is real; the second lawyer's surface is cheaper; the market cannot distinguish between them. The rational response for the deep practitioner is to abandon the competition, reducing her investment in the capabilities she cannot monetize. This produces the second-order lemons dynamic: the systematic disinvestment in the capabilities that markets cannot price, compounding over time into a market populated by surface practitioners whose outputs fail in the contexts where depth matters.

The policy response Stiglitz's framework prescribes is institutional restoration of quality signals. Professional certification that distinguishes AI-augmented depth from AI-substituted surface — not prohibitions on AI use, which would be both futile and counterproductive, but certification frameworks that make the distinction legible. Liability frameworks that assign accountability for AI-assisted output to the human who directed and approved it, creating incentives for practitioners to maintain the judgment that AI amplifies rather than substitutes for. Quality standards specific to domains where depth matters disproportionately — medicine, law, engineering, architecture — that maintain the market value of genuine expertise by requiring its demonstration. These interventions do not prevent AI adoption. They price it accurately, preserving the market infrastructure that rewards the human capital the AI economy most urgently needs.

Origin

The lemons dynamic was formalized by George Akerlof in his 1970 paper The Market for 'Lemons': Quality Uncertainty and the Market Mechanism, which won him the 2001 Nobel shared with Stiglitz and Spence. The extension to AI-era expertise has been developed by multiple authors — including economists, legal scholars, and AI researchers — as the implications of large language models for knowledge work became clear. The specific application to professional services follows directly from Akerlof's original framework and Stiglitz's subsequent generalization.

Key Ideas

Quality signals collapse under AI. Production cost, credentials, and reputation lose informational value as AI makes surface indistinguishable from depth.

Deep practitioners are driven from the market. They cannot compete on price with AI-assisted surface providers when the market cannot perceive the difference.

Second-order disinvestment. The market's inability to price depth produces rational disinvestment in the capabilities depth requires.

Institutional restoration as response. Certification, liability, and quality standards can restore the signals the market requires to function.

Not a prohibition on AI use. The response is pricing AI use accurately, not preventing it.