Developing a comprehensive model of aesthetic choice
Final Report Abstract
People do not always make decisions based on their monetary or fitness-related implications. Every day, we make decisions based on how the options look, sound, taste, or feel. In other words: We make decisions based on the options’ sensory (so-called aesthetic) value. My project aimed to study such aesthetic choices using a computational approach that is strictly comparable to those used to understand decision-making in other domains like money or food. I set this ambitious goal together with my supervisor Prof. Peter Dayan because it represents a unique marriage of our core competencies: the psychology of aesthetics on my side and computational cognitive science of reward learning on his. Most of the project was devoted to developing and testing a computational model of the value that we assign to sensory experiences. The model rests on the fundamental idea that sensory experiences are valuable to the extent that they help improving our sensory-cognitive system to process the sensory environment both now and in the future. Thus, sensory experiences are valuable to the extent that they are 1) already easy to process and 2) shaping the sensory-cognitive system in a way that improves processing sensory input we expect to experience in the future. The model operationalizes 1) as the probability of the sensory input given the observers generative model (system state) and 2) as the change in the distance between the system state and a second generative model that represents the observers long-term expectations about the properties of her sensory environment. We realized a simple model based on this framework and showed that the model can explain the occurrence and exact shape of the ‘mere-exposure effect’ – the phenomenon that people tend to like objects more that they have experienced more often but only up to a certain number of encounters. During this process, we also found out that the decrease in liking, the reported aesthetic value, over time could feasibly be called boredom. Thus, we also proposed a fleshed-out theory of how boredom applies to sensory experiences and how it can be characterized within our theoretical framework. At the same time, my first Master student Max Berentelg and I developed an experiment and a suitable stimulus set to test the model. We found that a simple realization of the model can predict a single participant’s liking ratings better than the ratings of other participants can. What is more, we also found that one of our core assumptions – that learning occurs and systematically alters liking ratings, thus rendering trial order relevant – held true; model predictions are best for the true trial order and do not apply to randomly generated alternatives. We have published all of the above results along with the code that lets any researcher run the model and fit it to existing data.
Publications
-
A computational model of aesthetic value. Talk at the 2nd International Symposium on the Mathematics of Neuro-Science, Technology and Engineering, Rhodes/virtual.
Brielmann, A. A. & Dayan, P.
-
A computational model of aesthetic value. Talk at the 43rd European Conference of Visual Perception
Brielmann, A. A. & Dayan, P.
-
Panel Discussion The concept of Aesthetic Emotions in the age of neuroasthetics: a roundtable discussion. Talk presented at XXVI Congress of the International Association of Empirical Aesthetics (IAEA 2021). 2021-09-01 - 2021-09-03.
Chatterjee, A.; Fingerhut, J.; Nadal, M.; Skov, M. & Brielmann, A.
-
A computational model of aesthetic value.. Psychological Review, 129(6), 1319-1337.
Brielmann, Aenne A. & Dayan, Peter
-
A computational model predicts humans' aesthetic judgments based on deep neural network feature values. Invited talk at the NeurIPS workshop Shared Visual Representations in Human & Machine Intelligence, New Orleans, Louisiana, USA/online.
Brielmann A. A.
-
A computational model predicts individual aesthetic judgments. Talk at the 64th Tagung experimentell arbeitender Psychologen, Cologne, Germany / online.
Brielmann A. A.; Berentelg, M. & Dayan, P.
-
Boredom in aesthetic experiences. Talk presented at The Third International Symposium on the Mathematics of Neuroscience. Heraklion, Greece. 2022-09-24 - 2022-09-25
Brielmann, A.
-
Is aesthetic value special? Invited talk at the Workshop on Aesthetic Values and Decision-Making in the Brain, Paris, France.
Brielmann A. A.
-
Is pleasure enough?. Invited talk at the Meeting of the International Association for Empirical Aesthetics, Philadelphia, USA.
Brielmann A. A.
-
Learning changes aesthetic value over time. Invited talk at the Interdisciplinary perspectives on Beauty and Change Workshop, Turin, Italy/online.
Brielmann A. A.
-
Predicting individual aesthetic judgments over time based on deep neural network features. Invited talk at the Meeting of the International Association for Empirical Aesthetics, Philadelphia, USA.
Brielmann A. A.; Berentelg, M. & Dayan, P.
-
Spatial complexity and intricacy predict subjective complexity. Poster at the Visual Sciene of Art Conference, Amsterdam, Netherlands.
Nath, S.S.; Braendle, F.; Schulz, E.; Dayan, P. & Brielmann, A. A.
-
The intrinsic reward of sensory experiences. Talk presented at Annual Meeting of the Society for NeuroEconomics (SNE 2022). Arlington, VA, USA. 2022-09-30 - 2022-10-02.
Brielmann, A.; Berentelg, M. & Dayan, P.
-
Modelling individual aesthetic judgements over time. Philosophical Transactions of the Royal Society B: Biological Sciences, 379(1895).
Brielmann, Aenne A.; Berentelg, Max & Dayan, Peter
-
Relating objective complexity, subjective complexity, and beauty in binary pixel patterns.. Psychology of Aesthetics, Creativity, and the Arts.
Nath, Surabhi S.; Brändle, Franziska; Schulz, Eric; Dayan, Peter & Brielmann, Aenne
-
The Routledge International Handbook of Boredom. Routledge.
Bieleke, Maik; Wolff, Wanja & Martarelli, Corinna
