Physiological data for affective computing in HRI with anthropomorphic service robots: the AFFECT-HRI data set

Area of Science:

• Human-Robot Interaction (HRI)
• Affective Computing
• Psychology
• Computer Science

Background:

• Human counterparts influence affective states, but robots lack empathic responses.
• This limits mitigation of non-beneficial affective reactions in human-robot interaction.
• Understanding robot behavior's impact on human affect is crucial for responsible HRI.

Purpose of the Study:

• To investigate the effect of robot behavior on human affect in human-robot interaction.
• To introduce the AFFECT-HRI dataset, featuring physiological data labeled with human affect.
• To enable interdisciplinary research on human-robot interaction and its psychological impact.

Main Methods:

• Conducted an HRI study with 146 participants interacting with an anthropomorphic service robot.
• Collected physiological data, questionnaire ratings (affect, demographics, socio-technical), across five conditions (neutral, transparency, liability, moral, immoral).
• Utilized a realistic retail scenario with three distinct interaction scenes.

Main Results:

• The study generated the AFFECT-HRI dataset, integrating physiological and self-reported human affect data.
• Different robot behaviors across conditions elicited varied affective responses from participants.
• The dataset facilitates analysis of how robot actions influence human emotions and mood.

Conclusions:

• The AFFECT-HRI dataset is a novel resource for studying affect in human-robot interaction.
• Findings underscore the importance of robot behavior design for positive human-robot interaction.
• This research supports the development of more empathetic and responsible service robots.