Mathematical modeling of intrinsic motivation in reversal theory

- Promoting exploration for AI agents

抄録

The function and performance of cloud connected products such as AI speakers are continuously updated over time. Such updates are based on the user’s exploration of unknown functions. Apter’s reversal theory proposed a mental condition termed the paratelic mode in which one acts to explore the purpose and enjoy certain actions in itself. We assume that the paratelic mode motivates users to explore continuously updated functions of cloud connected products which enables them to make full use of them. In this study, we aim to create a mathematical model that can explain the paratelic mode. We propose a model that explains the condition of the paratelic mode by integrating two principal motivation theories: Apter’s reversal theory and Berlyne’s optimal arousal level (OAL). We mathematically formulate the model by applying the Bayesian information gain as an index of arousal. By analyzing the model, we predict two hypotheses: a) when OAL is low, the lower the uncertainty, and the more likely it is that the paratelic mode is achieved, and b) when OAL is high, the higher the uncertainty, the more likely it is that the paratelic mode is achieved. The experimental result of our previous study using an AI speaker supported the former hypothesis. In this study, we verify the latter hypothesis by conducting an experiment using two AI speakers with different uncertainties. The results showed that when OAL is assumed to be high, users are more likely to be in the paratelic mode for an AI speaker which was subjectively evaluated to have higher uncertainty.