Estimating Human Attitude during Robot-mediated Referential Communication Tasks

Ziming Liu, Parker N. Collier, Robert Bray, Eun Jin Paek, Devin Casenhiser, Wenjun Zhou, Xiaopeng Zhao

Abstract: Referential Communication Task (RCT) is a goal-oriented communicative act, where two partners exchange information through conversation to achieve common understanding of certain given stimulus conditions. Recent advancements in information communication technologies have led to interesting applications in dialogue-based human-robot interaction. Recently, we designed and tested a robot-mediated RCT program using 96 human participants. In this study, we developed a BERT-based machine learning architecture to estimate participants’ negative attitude towards robots during the robot-mediated RCT. The results of prediction demonstrated the success on both machine prediction metrics, and stimulant and time invariant properties perspectives. We proposed that the designed protocol can be applied to measure user’s attitudes towards a robot during a robot-mediated RCT and develop an appropriate feedback plan to maintain user’s acceptance in real-time. In future work, we will explore the applications of robot-mediated RCT for people with cognitive impairment such as Alzheimer’s Disease and Related Dementia (ADRD) since language deficiency is an early symptom associated with ADRD.