Qualitative model for neuro-functional mechanism inducing human error detection to humanlike agents based on cortico-cerebellar function

The uncanny valley is a major phenomenon that occurs during the implementation of facial expressions in a humanlike agent. Recent studies have explained that the negative impression of the agent’s expressions is a “prediction error.” We hypothesized that the prediction error stemmed from the differential output of dual processing, - that is, emotional and rational processing, - and it negatively affected a person’s impression of the facial expression of the agent, which was similar to that when perceiving an eerie sensation. A qualitative computational model was proposed to clarify this mechanism according to cortico-cerebellar function to compute the prediction of the humanlike agent’s facial movement. Thus, rapid emotional processing was indispensable with the thalamus playing an essential role in the processing.