AGENT: A Benchmark for Core Psychological Reasoning
arXiv:2102.12321
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.
arXiv commentICML 2021, 12 pages, 7 figures
References in corpus (19)
- Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments
- Autonomous Agents Modelling Other Agents: A Comprehensive Survey and Open Problems
- Cooperative Inverse Reinforcement Learning
- One-Shot Imitation Learning
- Emergence of Grounded Compositional Language in Multi-Agent Populations
- ThreeDWorld: A Platform for Interactive Multi-Modal Physical Simulation
- One-Shot Imitation from Observing Humans via Domain-Adaptive Meta-Learning
- Charades-Ego: A Large-Scale Dataset of Paired Third and First Person Videos
- On the Utility of Learning about Humans for Human-AI Coordination
- Configurable 3D Scene Synthesis and 2D Image Rendering with Per-Pixel Ground Truth using Stochastic Grammars
- CLEVRER: CoLlision Events for Video REpresentation and Reasoning
- Watch-And-Help: A Challenge for Social Perception and Human-AI Collaboration
- Task-Embedded Control Networks for Few-Shot Imitation Learning
- IntPhys: A Framework and Benchmark for Visual Intuitive Physics Reasoning
- Probing Physics Knowledge Using Tools from Developmental Psychology
- Baby Intuitions Benchmark (BIB): Discerning the goals, preferences, and actions of others
- Learning Latent Representations to Influence Multi-Agent Interaction
- Continuous Relaxation of Symbolic Planner for One-Shot Imitation Learning
- Occlusion resistant learning of intuitive physics from videos