Modeling acceptable novelty using information theory

抄録

Novelty is an essential factor in design that is deemed attractive and creative by humans. However, an individual’s acceptance of novelty depends on their emotions. Knowledge about the range of novelty which ones accept will help to create attractive design targeting certain users. We previously developed a mathematical model of emotional dimensions associated with novelty such as arousal (i.e., surprise) and valence (i.e., positivity and negativity). The model formalized arousal as Bayesian information gain and valence as a function of the arousal based on Berlyne’s arousal potential theory. The arousal model was a function of three parameters: prediction error (the difference between expectation and reality), uncertainty (i.e., unpredictability), and external noise. In Berlyne’s model, the valence against novelty forms an inversed-U curve and turns from positive to negative when novelty is large. We assumed such a point that the sign of the valence turned showed the range of novelty ones would accept. Here, we derived a prediction error where valence turns from positive to negative. We termed this acceptable novelty and defined it as the range of novelty one can accept. Our model predicted that when the order of uncertainty is larger than external noise, the higher the uncertainty is, the larger the acceptable novelty becomes. This analysis is an update from our previous model and can be a basis of designing products to certain target users.