Active Information Gathering

Overview

The field studies how an agent can decide what information to seek next, and when to stop asking and act or defer. Common approaches include Bayesian experimental design using expected information gain or value of information, active-learning acquisition rules, and reinforcement learning policies trained to ask useful questions. These methods are tested in settings such as sequential clinical diagnosis, conversational tasks like 20 Questions, and agents that probe digital or physical environments before acting. A central challenge is that information gain does not always translate into better or safer decisions, so evaluations increasingly need to measure acquisition costs, interaction quality, and outcomes under incomplete evidence.

Latest papers

CardsList

32 papers6 papers in the last four weeks, up 100% on the four weeks before. 0.0% of all new papers.

  1. Reasoning under uncertainty in humans and language models

    Feb 9, 2026Gonçalo Guiomar, Elia Torre, Victoria Shavina +4Decision-Making under UncertaintyActive Information Gathering

  2. ProMed: Shapley Information Gain Guided Reinforcement Learning for Proactive Medical LLMs

    Aug 19, 2025Hongxin Ding, Baixiang Huang, Yue Fang +8RL for Language ModelsHealthcare