Ideal Bayesians never value information negatively, but humans prefer not to know
jessi_cata · x · 2026-09-04
The author argues that ideal Bayesian agents never assign negative value to information: if a policy attains utility U without looking, the same policy attains U by looking and ignoring it. Yet humans sometimes prefer not to know. Human rationality is a mix of heuristic bundles (in-distribution) and rational choice theory (out-of-distribution), so proving choiceworthiness for humans requires empirical psychology.
More from AGI Musings
- First human genome cost $3B and 13 years; now it's a few hundred dollars — PeterDiamandis · 2026-09-04
- theo: GPT-6 Astra is a generational leap beyond coding — computer use, 3D, swarms and more — gabrielchua · 2026-09-04
- Chollet: Astra Saturated ARC-AGI-3 Twice as Fast as I Predicted — fchollet · 2026-09-04
- Two Years On: Six Guidelines for Making Research Impact via Open-Source in AI — lateinteraction · 2026-09-04
- Ron Bodkin: Anyone Saying AGI Has Arrived Lacks Understanding — ronbodkin · 2026-09-04
- Chollet: ARC-AGI-4 lands Q1 2027, and solving ARC-3 is not AGI — fchollet · 2026-09-04