Rational Inattention to Discrete Choices with Stable Priors -- by Bruno Pellegrino

TEXT START: How does prior information affect discrete choice?

The Dissection

This paper repairs a technical gap in the rational-inattention model: it seeks tractable solutions when prior information differs across discrete choices. It improves the machinery for describing how agents allocate attention under uncertainty.

That is useful economics, but it is still economics inside the old operating system. The paper studies how humans choose when information is costly. It does not address what happens when AI makes the choosing, information processing, and attention allocation cheaper and better than human cognition.

The Core Fallacy

The dangerous inference is that a more sophisticated model of human decision-making implies continued economic necessity for human decision-makers. It does not.

Under the Discontinuity Thesis, the relevant question is not how finely human priors can be modeled. It is whether humans remain indispensable to production. This framework does not challenge P1, P2, or P3. Once AI can acquire, update, compare, and act on choice-relevant information at superior cost and speed, human priors become inputs to be modeled, manipulated, or discarded—not a durable economic moat.

The limitation is therefore structural, not mathematical: local precision in modeling human choice has no power to preserve the mass employment–wage–consumption circuit.

Hidden Assumptions

• Humans remain the operative decision-makers.
• Choice payoffs and information costs remain meaningfully human-scale.
• Information acquisition is a constraint rather than a rapidly automatable capability.
• Individual priors remain economically consequential instead of becoming data captured and processed by AI systems.
• The choice problem can be analyzed independently of ownership and control of the productive technology.
• Improved behavioral prediction translates into continued human productive participation.

Those assumptions may be acceptable for the model’s narrow purpose. They are fatal if smuggled into a forecast of human economic centrality.

Social Function

Classification: partial truth and academic prestige signaling, with possible transition-management value.

The paper can sharpen forecasts of behavior during the lag period when humans still make or influence discrete choices. It may help institutions price, regulate, or manage systems populated by increasingly automated agents. But it is not a defense of the existing order. At best, it produces a better map of the waiting room.

The Verdict

Technically relevant. Systemically secondary.

This is a refinement of the pre-discontinuity model, not an escape from discontinuity. It explains how humans process choices before AI removes their role as the bottleneck. The paper may improve prediction of the transition; it cannot reverse the transition or restore productive participation once cognitive automation severs the wage circuit.