Fourteenth International Conference on The Image

Abstract

Our entry point is from the cognitive science perspective. Our study of the image is centered upon drawings and measuring their creativity. Determining the creativity of an image is very challenging and a definitive measure has eluded researchers. From prior research, we compare two different types of creativity measurements (online rating and criterion-based) for 45 drawings. We discover similar outcomes from the measurements and that each approach featured different trade-offs. We wondered if there is another way to evaluate the creativity of drawings by applying Artificial Intelligence (AI) heatmaps. AI heatmaps are graphical representations of data that use color-coding to show the relative intensity or importance of different areas of an image. For the study reported here, we use an online AI heatmap generator (https://zyro.com/ai/heatmap) to visualize 18 drawings (9 highest and 9 lowest ranked creative drawings) from our prior study. We coded the heatmaps based on the number of embedded figures highlighted, the total number of embedded figures in the drawings, and the number of heatmap hot spots. The results show more hot spots for low creativity drawings and more highlighting of embedded figures despite no differences in the number of figures between the high and low creativity drawings. We were surprised that heatmaps identified essentially ‘hidden figures’. Perhaps higher ranked drawings hid the embedded figures better. One high-creativity drawing had 13 figures embedded and none of them were highlighted. These results support the use of AI software to discriminate creativity levels in drawings.