情報利得を用いた操作主体感の数理モデリング（応答遅れ時間を用いた検証）

抄録

The Sense of Agency (SoA) is the subjective awareness of controlling external world through one’s action. This paper proposes a mathematical model to explain the degree of SoA using information gain, representing the difference between expected and actual sensory responses to an operational action. We hypothesize that the degree of SoA is proportional to the inverse of information gain as a function of three parameters: prediction error, uncertainty and external noise. The model predicted that the prediction error decreases SoA, but uncertainty attenuates the decreases by prediction error. SoA in certain expectation is higher when prediction error is small, whereas SoA in uncertain expectation is high when prediction error is large. To validate the model prediction, we conducted three experiments with participants using varied operational delays as different level of prediction errors. Participants were adapted to expectation of low/high uncertainty and rated SoA at each delay. When prediction error was from 50 ms to 250 ms, the participants who were adapted to expectation of low uncertainty tended to rate lower SoA than the participants who ware adapted to high uncertainty (experiment 1). This tendency was also observed in event related potential N100 (experiment 2). These experimental results supported our model predictions.