A computational model of sensory valuation

Where do you want to live? With whom? Do you like that image or swipe it away? Numerous decisions, big and small, partly depend on options’ sensory appeal. Yet, we have a poor understanding of how sensory experiences gain value and how these values influence our decisions. We propose that sensory value serves as a signal towards the goal of maintaining and adapting the states of their cognitive-sensory system in order to process stimuli effectively now and in the future. Two interlinked components generate an object’s sensory value: 1) processing fluency – the likelihood of a stimulus given an observer's state; 2) the change in fluency with which likely future stimuli will be processed – the change in the average likelihood of expected future stimuli. A simple realization of this theory represents current and expected stimulus likelihoods as n-dimensional Gaussians. We test this model with an Instagram-like experiment where people both rate and browse through images. Our model can predict individual aesthetic judgments as well as their changes over time. What is more, our model can achieve this based on stimulus representations that are directly taken from convolutional neural networks. I will discuss the implications of these findings for the role of sensory valuation in shaping our sensory-cognitive system