Key points are not available for this paper at this time.
Creativity is now universally recognized as an essential, 21st century competency. However, there are many practical barriers to the development of creativity in schools, universities and workplaces. Chief among these is a problem of the fitness-for-purpose of creativity assessments. Highly valid and reliable creativity tests, necessary to support the development of individual creativity, are typically slow and labor-intensive to administer and score. However, recent advances in machine learning are opening up solutions to the fitness-for-purpose problem, across a range of different modes (e.g. verbal and figural) of testing. This paper describes the development of an open-source, large, image-based, multi-label classification model capable of automatically scoring the figural Test of Creative Thinking – Drawing Production (TCT-DP) with accuracies comparable to those of typical human raters. The growth of this, and similar machine learning models, promises to transform the development of creativity in education, offering rapid, accurate and cost-effective assessments that can be seamlessly integrated into teaching and learning.
Cropley et al. (Thu,) studied this question.