AuDrA: An automated drawing assessment platform for evaluating creativity

Area of Science:

• Psychology
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Visual creativity is crucial for human expression and ideation.
• Current visual creativity assessment relies heavily on time-consuming human raters.
• Automated methods for verbal creativity exist, but not for visual creativity.

Purpose of the Study:

• Introduce AuDrA, an Automated Drawing Assessment platform.
• Develop a machine learning model to automatically score visual creativity from drawings.
• Validate AuDrA's performance against human creativity ratings.

Main Methods:

• Trained an automated drawing assessment (AuDrA) model using line drawings and human creativity ratings.
• Tested AuDrA's generalizability across different drawing sets, raters, and tasks.
• Compared AuDrA scores with human ratings and drawing elaboration (ink on page).

Main Results:

• AuDrA scores showed high correlations with human creativity ratings (mean r = .76) across four datasets.
• Performance generalized to untrained drawing sets, raters, and tasks.
• AuDrA's sensitivity exceeded simple measures of drawing complexity.

Conclusions:

• AuDrA provides a reliable and efficient automated solution for visual creativity assessment.
• The platform overcomes limitations of traditional human-based rating systems.
• AuDrA enables researchers to efficiently assess creativity in large drawing datasets.