Explainability via Responsibility

TL;DR

This paper introduces a method for making machine learning-based game content generation more explainable by providing human-understandable explanations of AI actions, improving human-AI collaboration.

Contribution

It proposes a novel approach that offers training instances as explanations in co-creative game design, enhancing interpretability of black box models.

Findings

Training instances improve user understanding of AI actions

Explanations facilitate more efficient human-AI cooperation

Approach demonstrates potential for explainability in PCGML systems

Abstract

Procedural Content Generation via Machine Learning (PCGML) refers to a group of methods for creating game content (e.g. platformer levels, game maps, etc.) using machine learning models. PCGML approaches rely on black box models, which can be difficult to understand and debug by human designers who do not have expert knowledge about machine learning. This can be even more tricky in co-creative systems where human designers must interact with AI agents to generate game content. In this paper we present an approach to explainable artificial intelligence in which certain training instances are offered to human users as an explanation for the AI agent's actions during a co-creation process. We evaluate this approach by approximating its ability to provide human users with the explanations of AI agent's actions and helping them to more efficiently cooperate with the AI agent.