AGENT: A Benchmark for Core Psychological Reasoning

TL;DR

This paper introduces AGENT, a benchmark dataset designed to evaluate machine agents' understanding of core psychological reasoning, inspired by human intuitive psychology, through diverse 3D animation scenarios.

Contribution

It presents a new benchmark with a comprehensive dataset and evaluation protocol to assess AI models' ability to reason about human mental states and actions.

Findings

Models need to incorporate utility and physics knowledge to pass tests.

Current baselines show room for improvement in core psychological reasoning.

AGENT correlates well with human judgments on psychological reasoning tasks.

Abstract

For machine agents to successfully interact with humans in real-world settings, they will need to develop an understanding of human mental life. Intuitive psychology, the ability to reason about hidden mental variables that drive observable actions, comes naturally to people: even pre-verbal infants can tell agents from objects, expecting agents to act efficiently to achieve goals given constraints. Despite recent interest in machine agents that reason about other agents, it is not clear if such agents learn or hold the core psychology principles that drive human reasoning. Inspired by cognitive development studies on intuitive psychology, we present a benchmark consisting of a large dataset of procedurally generated 3D animations, AGENT (Action, Goal, Efficiency, coNstraint, uTility), structured around four scenarios (goal preferences, action efficiency, unobserved constraints, and cost-reward trade-offs) that probe key concepts of core intuitive psychology. We validate AGENT with human-ratings, propose an evaluation protocol emphasizing generalization, and compare two strong baselines built on Bayesian inverse planning and a Theory of Mind neural network. Our results suggest that to pass the designed tests of core intuitive psychology at human levels, a model must acquire or have built-in representations of how agents plan, combining utility computations and core knowledge of objects and physics.