Modelling acceptable novelty transitions with emotional habituation: Effects of uncertainty and prediction error on preference changes

Previous mathematical models developed to optimize the degree of novelty in product design have represented novelty by emotional dimensions such as arousal (surprise) and valence (positivity and negativity). Formalizing arousal as Bayesian information gain and valence as a function of arousal based on Berlynes arousal potential theory, the model indicates that novelty becomes acceptable with repeated exposure and changes the preferences of the users. Hence, acceptable novelty transitions are important in the design of long-term product experience. We propose a mathematical model of acceptable novelty transitions with emotional habituation based on the emotional dimension model. We formalized valence as a function of information-theoretic free-energy and expressed free-energy as a function of three parameters: initial prediction error, initial uncertainty, and noise of sensory stimulus. To verify whether the transition speed of acceptable novelty depends on the initial uncertainty in our model, we analysed the responses of participants to historic artworks; we manipulated the uncertainty level by varying the obscurity of the presented pieces and the prediction error by rendering them in different artistic styles. We used the subjective reports of valence in response to the samples as measures of valence levels. The experimental results support our hypothesis.