Using Generative AI for Automated Human-Creativity Assessment and Machine Creativity in a Sandbox Game

• In Event: AERA Roundtable Session Wednesday 10:50 am Four Seasons Ballroom 2-3
  In Roundtable Session: Research on Giftedness, Creativity, and Talent SIG Roundtable (Table 1)

Abstract

This study explores the use of Generative AI, specifically ChatGPT 4o, to automate creativity assessment and game level generation in the context of a sandbox game called Physics Playground. Training ChatGPT 4o to assess the creativity of game levels created by humans and generate new creative game levels. Initial results show a strong correlation between AI and human ratings of existing game levels, indicating ChatGPT 4o’s validity in assessing human creativity in the context of Physics Playground. Our findings showed that ChatGPT 4o can assess the levels with a 0.99 correlation to human raters and generate new game levels with more success using the vector image generation method.

Authors

• Seyedahmad Rahimi, University of Florida

• Deniz Ercan, University of Florida

• Ran Gao, University of Florida

• Salah Esmaeiligoujar, University of Florida

• Hongming Li, University of Florida

• Anthony Botelho, University of Florida