Using Reinforcement Learning Agents to Analyze Player Experience

摘要

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 representation of game states. Then, machine learning models are trained with the representation to evaluate the player experience. It shows that the representation learned by AI agents can provide important information about how game levels are perceived by players. And the representation can help machine learning models to classify whether player experience is enjoyable.