Mixed-Initiative Level Design with RL Brush

TL;DR

RL Brush is a reinforcement learning-powered level design tool that enhances human creativity in tile-based games by providing AI suggestions, leading to longer engagement and higher quality levels.

Contribution

This work introduces RL Brush, a novel mixed-initiative level design tool using reinforcement learning for AI suggestions in Sokoban.

Findings

Users with AI suggestions stayed longer.

Levels created were more playable.

Levels were more complex.

Abstract

This paper introduces RL Brush, a level-editing tool for tile-based games designed for mixed-initiative co-creation. The tool uses reinforcement-learning-based models to augment manual human level-design through the addition of AI-generated suggestions. Here, we apply RL Brush to designing levels for the classic puzzle game Sokoban. We put the tool online and tested it in 39 different sessions. The results show that users using the AI suggestions stay around longer and their created levels on average are more playable and more complex than without.