Value of Information

Follow this topic Ask about this topic

Value of information (VOI) asks whether an observation, question, experiment, or computation is likely to improve a decision enough to justify its cost. Bayesian decision theory measures expected utility gains from sample or perfect information, while related methods use entropy reduction, regret, or risk bounds to guide experimental design, POMDP exploration, diagnosis, and clarification. Approaches range from dynamic programming and expected value of sample information calculations to Monte Carlo estimation, learned question policies, and certified rules for allocating sensing or computation. Across settings such as robotics, health sensing, retrieval-augmented generation, and language-agent interaction, a central challenge is that reducing uncertainty is not automatically useful: information must change actions for the better, given its costs and risks.

Latest papers

37 papers8 papers in the last four weeks, level with the four weeks before. 0.0% of all new papers.

Leveraging the Value of Information in POMDP Planning
Apr 1, 2026Zakariya Laouar, Qi Heng Ho, Zachary SunbergPartially Observable Markov Decision Processes Belief-Space Planning
Department of Aerospace Engineering Sciences, University of Colorado Boulder · Kevin T. Crofton Department of Aerospace and Ocean Engineering, Virginia Tech
Cognitive Friction: A Decision-Theoretic Framework for Bounded Deliberation in Tool-Using Agents
Mar 31, 2026Davide Di GioiaTool-Using Agents Cognitive Architectures for AI Agents
Entropic Claim Resolution: Uncertainty-Driven Evidence Selection for RAG
Mar 30, 2026Davide Di GioiaRetrieval-Augmented Generation Evidence Selection
Chance-Constrained Correlated Equilibria for Robust Noncooperative Coordination
Mar 14, 2026Jaehan Im, Ufuk Topcu, David Fridovich-KeilGame Theory Decision-Making under Uncertainty
Department of Aerospace Engineering and Engineering Mechanics, The University of Texas at Austin · The Oden Institute for Computational Engineering and Sciences, The University of Texas at Austin
A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring
Feb 26, 2026Usman Anwar, Julianna Piskorz, David D. Baek +6Language Model Steganography LLM Security
University of Cambridge · Massachusetts Institute of Technology · UK AI Safety Institute +3
Approximation of differential entropy in Bayesian optimal experimental design
Oct 1, 2025Chuntao Chen, Tapio Helin, Nuutti Hyvönen +1Bayesian Experimental Design Value of Information
Unfair Utilities and First Steps Towards Improving Them
Jun 1, 2023Frederik Hytting Jørgensen, Sebastian Weichwald, Jonas PetersValue of Information Algorithmic Fairness
Dept. of Mathematical Sciences University of Copenhagen Copenhagen, Denmark · Dept. of Mathematics ETH Zurich Zurich, Switzerland

Apr 1, 2026J/K move · Enter open · S save

Leveraging the Value of Information in POMDP Planning

Zakariya Laouar, Qi Heng Ho, Zachary Sunberg

Department of Aerospace Engineering Sciences, University of Colorado Boulder · Kevin T. Crofton Department of Aerospace and Ocean Engineering, Virginia Tech

Partially Observable Markov Decision Processes Belief-Space Planning Value of Information Monte Carlo Tree Search Dynamic Programming

Open paper AskPDF arXiv

Partially observable Markov decision processes (POMDPs) offer a principled formalism for planning under state and transition uncertainty. Despite advances made towards solving large POMDPs, obtaining performant policies under limited planning time remains a major challenge due to the curse of dimensionality and the curse of history. For many POMDP problems, the value of information (VOI) - the expected performance gain from reasoning about observations - varies over the belief space. We introduce a dynamic programming framework that exploits this structure by conditionally processing observations based on the value of information at each belief. Building on this framework, we propose Value of Information Monte Carlo planning (VOIMCP), a Monte Carlo Tree Search algorithm that allocates computational effort more efficiently by selectively disregarding observation information when the VOI is low, avoiding unnecessary branching of observations. We provide theoretical guarantees on the near-optimality of our VOI reasoning framework and derive non-asymptotic convergence bounds for VOIMCP. Simulation evaluations demonstrate that VOIMCP outperforms baselines on several POMDP benchmarks.