Learning Intuitive Policies Using Action Features

TL;DR

This paper explores how attention-based neural network architectures can enable AI agents to learn intuitive, human-interpretable policies by exploiting semantic relationships between actions and observations in multi-agent coordination tasks.

Contribution

It demonstrates that attention-based architectures improve the learning of intuitive, interpretable policies without human data, highlighting the importance of network design for semantic understanding.

Findings

Attention-based architectures outperform other models in learning intuitive policies.

Agents coordinate effectively with humans without prior human data.

Policies learned are human-interpretable and leverage semantic relationships.

Abstract

An unaddressed challenge in multi-agent coordination is to enable AI agents to exploit the semantic relationships between the features of actions and the features of observations. Humans take advantage of these relationships in highly intuitive ways. For instance, in the absence of a shared language, we might point to the object we desire or hold up our fingers to indicate how many objects we want. To address this challenge, we investigate the effect of network architecture on the propensity of learning algorithms to exploit these semantic relationships. Across a procedurally generated coordination task, we find that attention-based architectures that jointly process a featurized representation of observations and actions have a better inductive bias for learning intuitive policies. Through fine-grained evaluation and scenario analysis, we show that the resulting policies are human-interpretable. Moreover, such agents coordinate with people without training on any human data.