![[2602.09127] Epistemic Throughput: Fundamental Limits of Attention-Constrained Inference](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.09127] Epistemic Throughput: Fundamental Limits of Attention-Constrained Inference
Summary
This paper explores the concept of 'epistemic throughput' in attention-constrained inference, analyzing how generative AI systems can manage decision-making under limited attention by formalizing the screening and verification processes.
Why It Matters
Understanding the limits of attention-constrained inference is crucial as AI systems increasingly handle vast amounts of data. This research provides insights into optimizing decision-making processes, which is vital for improving AI efficiency and effectiveness in real-world applications.
Key Takeaways
- Introduces the concept of epistemic throughput in AI decision-making.
- Establishes a scaling law that describes the relationship between verification, prevalence, and screening quality.
- Highlights the importance of heavy-tailed score distributions for effective amplification in sparse-verification scenarios.
- Demonstrates that expanding cheap screening can significantly enhance verification processes.
- Provides a formal framework for understanding the limitations of attention in AI systems.
Computer Science > Machine Learning arXiv:2602.09127 (cs) [Submitted on 9 Feb 2026 (v1), last revised 12 Feb 2026 (this version, v2)] Title:Epistemic Throughput: Fundamental Limits of Attention-Constrained Inference Authors:Lei You View a PDF of the paper titled Epistemic Throughput: Fundamental Limits of Attention-Constrained Inference, by Lei You View PDF 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-v...
