This article discusses the importance of achieving human-AI alignment in multiplayer games, which is crucial for creating trustworthy AI agents. The authors propose a method to evaluate this alignment using a task-sets framework, which focuses on high-level behavioral tasks. They analyze human gameplay data from Xbox’s Bleeding Edge, train an AI agent to play the game, and then compare the behavior of the human and AI gameplay. The findings reveal significant differences in the behavior of human and AI players, highlighting the need for interpretable evaluation, design, and integration of AI in human-aligned applications. The study advances the discussion on AI alignment, offering a measurable framework for interpretable human-agent alignment in multiplayer gaming.
Publication date: 7 Feb 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2402.03575
