Skip to main content Skip to navigation

Generating Levels That Teach Mechanics

Project Overview

The document explores the integration of generative AI in education, particularly through the development of game tutorials using the Mario AI framework. It emphasizes the use of evolutionary algorithms to create engaging game levels that effectively teach specific mechanics by guiding players to perform targeted actions for progression. The research underscores the significance of designing well-structured tutorials that not only enhance the gameplay experience but also address the unique challenges of catering to novice players. By leveraging generative AI, the findings demonstrate the potential for personalized learning experiences that adapt to individual player needs, thereby fostering deeper understanding and engagement in educational contexts. Overall, the application of generative AI in educational game design showcases its potential to revolutionize how complex concepts are taught through interactive and adaptive learning environments.

Key Applications

Game tutorial generation using evolutionary algorithms

Context: Educational context is video game design and development, targeting game developers and designers.

Implementation: Automatic generation of levels using a feasible infeasible 2-population evolutionary algorithm which tests various AI agents with different limitations.

Outcomes: Generated levels that teach specific mechanics effectively, providing a structured learning experience for players.

Challenges: Levels generated may be too difficult for novice players due to the superhuman reflexes of the AI used in testing.

Implementation Barriers

Technical Barrier

The AI agents used to evaluate the levels possess superhuman reflexes, making the levels too challenging for human players.

Proposed Solutions: Using human-like agents instead of perfect agents to create more accessible game levels for novice players.

Project Team

Michael Cerny Green

Researcher

Ahmed Khalifa

Researcher

Gabriella A. B. Barros

Researcher

Andy Nealen

Researcher

Julian Togelius

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Michael Cerny Green, Ahmed Khalifa, Gabriella A. B. Barros, Andy Nealen, Julian Togelius

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies