Compression, Selection, and Organizational Self-Deception: A Unified Theory

Executive Summary

Intelligent systems deceive themselves. This happens in individual minds, in organizations, in artificial intelligence, and in academia, and the pattern is stable, predictable, and general. The paper explains it with two mechanisms: compression and selection.

Processing complexity requires compression. You cannot coordinate a thousand people through intuition; you need reports, metrics, and summaries. Compression loses information, and lost information creates gaps where distinct states look identical. Those gaps become selection environments.

The Selection Model

Selection fills the gaps through three compounding dynamics. Communication degrades as preferences diverge (strategic communication, after Crawford and Sobel). Accurate signals are costly to produce and hard to distinguish from cheap ones, so cheap signals flood the market (adverse selection, after Akerlof). Transmission favors what spreads easily over what represents reality accurately (transmission bias, after Boyd and Richerson).

The paper calls the compound force dysmemic pressure: selection favoring fit over truth in environments shaped by compressed representations. A dysmeme is a signal fit for the gap between representation and reality; it may be true or false, but truth is not the selection criterion.

Equilibrium and the Cage

Established dysmemes become the environment new signals must adapt to, so the landscape ratchets toward frame-fit. The cycle stabilizes: compression creates gaps, selection fills them, the filled gaps become the reality people navigate, and their navigation generates new signals facing the same selection. The paper calls this equilibrium the Cage.

Reform fails because a reform proposal is itself a signal subject to the selection environment it aims to change. Reform that renames things while leaving the fitness criteria intact meets little resistance; reform that changes the fitness function needs leverage from outside the current selection environment.

Substrate Independence and the Mirror

The same pattern appears in cognition, organizations, AI, and academia. Because human psychology cannot explain it in AI, and bureaucratic incentives cannot explain it in individual cognition, the paper locates the cause in what these substrates share: compression and selection. The framework states testable propositions and falsification criteria.

The response is the Mirror: structures that create observation points outside local selection pressure, such as high-reliability practices, independent central banks, Inspectors General, and adversarial collaboration. Each combines insulation from local selection, access to what the primary frame discards, and authority to surface findings. The Mirror does not remove the Cage; it makes the walls visible.

Key References