Modelling individual aesthetic judgements over time.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Individual differences in the value of sensory experiences are significant and variable.
• Existing theories suggest these values are learned through optimizing sensory-cognitive systems for current and future input.

Purpose of the Study:

• To test a theory that idiosyncratic values of sensory experiences arise from tuning the sensory-cognitive system.
• To develop and validate a computational model predicting subjective evaluations.

Main Methods:

• Utilized a dataset of 55 dog images generated via the NeuralCrossbreed algorithm.
• Employed deep neural networks (e.g., VGG-16) to extract image features.
• Collected liking ratings from 59 participants for each image.

Main Results:

• The computational model, using deep neural network features, accurately predicted trial-by-trial liking judgments (median r = 0.65).
• Model predictions significantly outperformed population average predictions (median r = 0.01).
• The model's learning component explained sequence-dependent rating changes, capturing 17% more variance than random trial orders.

Conclusions:

• The study validates a theory linking individual value judgments to sensory-cognitive system tuning.
• This work represents a foundational step towards a comprehensive understanding of individual differences in evaluation.
• The findings highlight the potential of computational models in explaining subjective aesthetic experiences.