Xinchi Yu¹ (), Pranav Raghavan¹, Shlok Amit Desai¹, Yan Bao^2,3, Ernst Pöppel^3,2, Weizhen Xie¹; ¹University of Maryland, ²Peking University, ³Ludwig-Maximilians-Universität München

While beauty is often said to be in the eye of the beholder, certain artworks consistently evoke shared aesthetic appreciation across individuals. This raises the question: to what extent does beauty inherently reside within the artworks themselves, independent of personal preferences or prior knowledge? To address this, we investigated whether pixel-level features of abstract artworks could predict consistent aesthetic judgments across people. Using abstract ink blob paintings by LaoZhu, each comprising different color variations (red, green, and gray) of the same painting, we evaluated aesthetic judgments while strictly controlling for confounding factors such as object size, image complexity, and art style. In Study 1, 617 participants rated the beauty of these paintings. Remarkably, participants demonstrated high agreement in their judgments. Interestingly, Western participants (n = 422), who were less familiar with the ink blob style, showed comparable consistency to Eastern participants (n = 195), suggesting that prior knowledge does not fully explain these shared aesthetic judgments. In Study 2, we examined whether image memorability contributed to this consistency using a continuous recognition memory task in 314 participants. The results provided minimal evidence for this, indicating that beauty appreciation is distinct from memory-related processes. To probe further, we analyzed the paintings using a pre-trained deep neural network (DNN, VGG16). We found that activation patterns in intermediate layers with a lower computational processing cost, namely lower mean activation, predicted higher aesthetic ratings. Notably, these DNN features were unrelated to memorability, highlighting distinct mechanisms for aesthetic perception and memory. Overall, our findings suggest that the consistency in perceived beauty across people may emerge from reduced information processing costs of artworks. Guided by these crowdsourced and modeling results, future research with neural recordings may help reveal the neurobiological basis of visual aesthetics.

Acknowledgements: This study is made possible by funding support from the National Institute of Neurological Disorders and Stroke (R00NS126492, PI: W. X.).