Item

Abstract

The complexity of images critically influences our impression of them and our assessment of their beauty. However, there is no consensus on the best way to formalize an objective measure of complexity for images. Moreover, studies relating subjective assessments of complexity and beauty to potential objective measures are hampered by the use of hand-crafted stimuli from which it is hard to generalize. To tackle these issues, we generated 2D black-and-white patterns algorithmically using cellular automata, and collected the ratings of 80 participants of their subjective complexity and beauty. We then assessed the relationship between beauty and complexity ratings, and objective measures of complexity such as density, asymmetry, entropy, local spatial complexity, and (approximate) Kolmogorov complexity. We also introduced an “intricacy” measure that quantifies the number of components in a pattern using a graph-based approach. We found that a weighted combination of local spatial complexity and intricacy was an effective predictor (R2test = 0.46) of subjective complexity. This implies that people’s complexity ratings depend on the number of distinct elements in the pattern along with the elements’ local spatial distribution and therefore that global and local image features are integrated to determine complexity judgements. Furthermore, we found a positive linear relationship between beauty and complexity ratings, with a negative linear influence of disorder, namely asymmetry and entropy, and a negative interaction between the two (with a total predictive accuracy of R2test = 0.64). This implies that there is beauty in complexity as long as there is sufficient order in the form of low asymmetry and randomness. In addition, a moderated mediation analysis showed that subjective complexity mediates the influence of objective complexity on beauty at all levels of disorder. Lastly, we found some evidence for individual differences with different people displaying different degrees of preference towards intricacy (in their complexity assessments) and dislike of disorder (in their beauty assessments).