Human-Aligned Skill Discovery: Balancing Behaviour Exploration and Alignment

TL;DR

This paper introduces HaSD, a framework that uses human feedback to discover diverse, safe, and aligned skills in reinforcement learning, improving the practicality and safety of autonomous skill discovery.

Contribution

The paper proposes Human-aligned Skill Discovery (HaSD), integrating human feedback to ensure safety and alignment during unsupervised skill discovery in complex environments.

Findings

HaSD discovers safer, more aligned skills in 2D navigation and SafetyGym environments.

HaSD maintains alignment throughout the skill discovery process.

Extended HaSD learns configurable skills with varying diversity and alignment trade-offs.

Abstract

Unsupervised skill discovery in Reinforcement Learning aims to mimic humans' ability to autonomously discover diverse behaviors. However, existing methods are often unconstrained, making it difficult to find useful skills, especially in complex environments, where discovered skills are frequently unsafe or impractical. We address this issue by proposing Human-aligned Skill Discovery (HaSD), a framework that incorporates human feedback to discover safer, more aligned skills. HaSD simultaneously optimises skill diversity and alignment with human values. This approach ensures that alignment is maintained throughout the skill discovery process, eliminating the inefficiencies associated with exploring unaligned skills. We demonstrate its effectiveness in both 2D navigation and SafetyGymnasium environments, showing that HaSD discovers diverse, human-aligned skills that are safe and useful for downstream tasks. Finally, we extend HaSD by learning a range of configurable skills with varying degrees of diversity alignment trade-offs that could be useful in practical scenarios.