Skip to content
← Back to research

LLM-Based Agents for Player-Type-Aware Automated Playtesting

A research direction exploring how LLM-based agents can support automated playtesting while reflecting different player behavior patterns.

LLM Agents/Automated Playtesting/Player Modeling/Game AI

This project studies how LLM-based agents can be used in automated playtesting without treating agent success as the only meaningful signal.

The public framing is intentionally conservative. The first version of the site describes the research agenda, methods, and motivation, while leaving detailed findings and figures for a future reviewed release.

Current Focus

  • Building agent behaviors that can be compared with human gameplay traces.
  • Studying whether player-type assumptions can shape automated playtesting behavior.
  • Creating analysis workflows that make agent behavior easier to inspect and explain.