Leuven Orthogonalized Art Dataset (LOAD): A Multidimensional Art Image Set for Aesthetic Appreciation Research

The intricate interplay between visual art, aesthetic appreciation, and human cognition has long captivated artists and scientists. One difficulty of this line of research is that artworks differ on many dimensions that influence aesthetic appreciation. While trying to keep the out-of-scope dimensions constant is a typical strategy, using a stimulus set with balanced dimensions avoids noise and bias. Although several visual art datasets exist, none have been specifically designed to control many dimensions simultaneously. In response to the need of a well-controlled and balanced stimulus set, the Leuven Orthogonalized Art Dataset (LOAD) controls and balances dimensions related to aesthetic appreciation, including style, content, emotional valence, and liking/beauty. Each artwork in the image set was first carefully selected to fit a balanced design, then annotated by 50 participants on measures of pleasure, fluency, interest, liking, emotional valence, and familiarity for further validation. This careful annotation process also aims to disentangle the often- ambiguous interrelations among several aesthetic appraisals such as pleasure, interest, and liking. By providing balanced designs across multiple dimensions, LOAD serves as a valuable and highly flexible tool for exploring the complexities of aesthetic appreciation.