Bayesian Explanation of the Effects of Tactile Feedback on the Affective Evaluation of Switch Sounds

抄録

In our previous study, the effects of tactile feedback on the affective evaluation of switch sounds were examined through auditory and tactile evaluation experiments using 15 switches, and it was shown that these effects could be modeled using multiple regression analyses. However, the reason why the effects can be modeled by multiple regression analyses remains unanswered. In this study, extended experiments were conducted using 25 switches. In the auditory evaluation experiment, using 26 adjective pairs on 7-point category scales, two groups consisting of 22 subjects each participated as operators and listeners, respectively. In the tactile evaluation experiment, using 16 adjective pairs, another group of 52 subjects participated. The results of both experiments were analyzed, separately, using factor analyses, with the three factors of potency, evaluation, and activity being extracted for both experiments. The factor scores of the operators were modeled using multiple regression analyses with the factor scores of listeners and tactile evaluations. Another modeling was conducted based on Bayes’ theorem, wherein the factor scores of listeners and operators were regarded as prior and posterior distributions, respectively, with the tactile factor scores being regarded as likelihood. The fact that these two models show almost the same estimation performance suggests that the effects of tactile feedback on the affective evaluation of switch sounds are modeled well by the multiple regression analysis because Bayesian multisensory integration is the origin of the cross-modal effects. This can be understood by the fact that the estimation equations of both models are essentially the same.