Learning Causal Overhypotheses through Exploration in Children and Computational Models

TL;DR

This paper introduces a new RL environment with controllable causal structures to compare exploration strategies of children and computational models, revealing key differences and inspiring future research in causal exploration.

Contribution

It presents a novel causal RL environment and compares exploration behaviors of children and models, highlighting differences and potential directions for improving RL exploration strategies.

Findings

Children and models explore differently in causal environments.

Information-gain optimal RL exploration differs from children's exploration.

The environment enables unified evaluation of exploration strategies.

Abstract

Despite recent progress in reinforcement learning (RL), RL algorithms for exploration still remain an active area of research. Existing methods often focus on state-based metrics, which do not consider the underlying causal structures of the environment, and while recent research has begun to explore RL environments for causal learning, these environments primarily leverage causal information through causal inference or induction rather than exploration. In contrast, human children - some of the most proficient explorers - have been shown to use causal information to great benefit. In this work, we introduce a novel RL environment designed with a controllable causal structure, which allows us to evaluate exploration strategies used by both agents and children in a unified environment. In addition, through experimentation on both computation models and children, we demonstrate that there are significant differences between information-gain optimal RL exploration in causal environments and the exploration of children in the same environments. We conclude with a discussion of how these findings may inspire new directions of research into efficient exploration and disambiguation of causal structures for RL algorithms.