Epistemic Throughput: Fundamental Limits of Attention-Constrained Inference

TL;DR

This paper introduces a formal framework called Attention-Constrained Inference (ACI) to analyze the limits of inference when decision-makers have limited attention for verification, revealing a nonlinear scaling law for epistemic throughput.

Contribution

It formalizes the ACI regime and derives a tight scaling law (JaKoB) showing how screening quality and scarce verification interact to determine information gain.

Findings

Epistemic throughput scales as rac{rac{JKB}{ ext{verification}}}

Heavy-tailed score distributions enable greater leverage in sparse verification regimes

Expanding screening quality nonlinearly amplifies the impact of scarce verification.

Abstract

Recent generative and tool-using AI systems can surface a large volume of candidates at low marginal cost, yet only a small fraction can be checked carefully. This creates a decoder-side bottleneck: downstream decision-makers must form reliable posteriors from many public records under scarce attention. We formalize this regime via Attention-Constrained Inference (ACI), in which a cheap screening stage processes $K$ records and an expensive verification stage can follow up on at most $B$ of them. Under Bayes log-loss, we study the maximum achievable reduction in posterior uncertainty per window, which we call \emph{epistemic throughput}. Our main result is a ``JaKoB'' scaling law showing that epistemic throughput has a baseline term that grows linearly with verification and prevalence, and an additional \emph{information-leverage} term that scales as $\sqrt{JKB}$, where $J$ summarizes screening quality. Thus, expanding cheap screening can nonlinearly amplify scarce verification, even when informative records are rare. We further show that this scaling is tight in a weak-screening limit, and that in the sparse-verification regime ($B \ll K$), substantial leverage requires heavy-tailed score distributions; for light-tailed scores the amplification is only logarithmic.