Not all uncertainty is alike: volatility, stochasticity, and exploration

Payam Piray

arXiv:2605.19215 · 2026-05-20 공개 · arXiv · PDF

gittins-index causes stochasticity volatility exploration-strategies noise-inference bandit-problems computational-psychiatry

Abstract

Adaptive decision-making in biological and artificial intelligence requires balancing the exploitation of known outcomes with the exploration of uncertain alternatives. Although prior work suggests that uncertainty generally promotes exploration, it has typically treated distinct sources of environmental uncertainty as equivalent. We consider environments with latent reward states that drift over time (volatility) and are observed through noisy outcomes (stochasticity). Both increase posterior uncertainty, yet we show they drive optimal exploration in opposite directions: volatility enhances it, stochasticity suppresses it. We establish this asymmetry formally by extending the Gittins index framework to Gaussian state-space bandits with latent dynamics. We further derive Cause-Aware Uncertainty-Sensitive Exploration (CAUSE), a closed-form exploration bonus obtained via control-as-inference that inherits the same monotonicities. CAUSE outperforms standard exploration strategies in environments with heterogeneous noise structure, and also improves on a Gittins-per-arm policy whose rested-bandit optimality does not transfer to restless settings. Learning and exploration are governed by the same noise-inference asymmetry, and the framework predicts that pathological noise inference produces \emph{reversed} rather than merely impaired exploration, with implications for computational accounts of psychiatric conditions.

한국어 요약

한 줄 요약

**[강화학습 / 의사결정 이론]** posterior 불확실성을 키우는 두 원인 — 환경 변동성(volatility)과 관측 잡음(stochasticity) — 이 최적 탐험을 반대 방향으로 이끈다는 비대칭을 형식적으로 증명하고 닫힌 형태의 탐험 보너스 CAUSE 제안.

핵심 기여도

핵심 아이디어

"불확실성 = 탐험 신호"라는 일원적 시각은 두 잡음 원인의 인과적 차이를 무시하고 있으며, volatility는 미래 보상 분포 변화로 정보 가치를 키우는 반면 stochasticity는 정보 가치를 깎으므로 탐험을 억제해야 한다.

기술적 접근법

주요 결과

의의 및 한계

**의의**: 탐험 이론에 새 차원(불확실성 원인의 인과적 분해)을 추가, 신경과학·심리학에서 관찰되는 비정상 탐험 패턴(예: 우울·불안)을 동일 프레임에서 설명. **한계**: Gaussian state-space 가정 의존, 비정상·다중 잠재 변수 환경으로의 확장은 후속 과제.

실용적 활용