Better than humans: a method for inferring consumer shopping intentions by reading facial expressions

摘要

How to make robots more human-like has been a hot topic of interest for researchers in the field of robot interaction. Empowering robots in retail services with empathy to read minds is still a major challenge. To address this problem, we proposed a deep learning-based method for programs to infer consumers' shopping interest levels, which we confirmed effectiveness has met or exceeded that of human judgement. The experimental results demonstrated that the proposed method achieved an accuracy of 65% in predicting consumer shopping interest levels, compared to the accuracy of 46% in predicting consumer shopping interest levels for manual assessment (n = 40), respectively. We anticipate that the results of this study will have considerable implications for research on human-robot interaction in service robots.