Using Reinforcement Learning Agents to Analyze Player Experience

Abstract

Analyzing player experience often requires collecting lots of gameplay data from human players, which is labor-intensive. In this paper, we present an approach to classify player experience using AI agents. A deep Reinforcement AI agent is deployed to learn abstract representation of game states. Then, machine learning models are trained with the abstract representation to evaluate the player experience. It shows that the abstract representation learned by AI agents can provide important information about how game levels are perceived by players. And the abstract representation can help machine learning models to classify whether player experience is enjoyable.