Embodying Politeness in Persuasive Humanoid Agents for Small Group Scenarios
Time: Wed 2024-03-06 10.00
Location: Visualization Studio VIC, Lindstedtsvägen 7
Video link: https://kth-se.zoom.us/j/62065064122
Language: English
Doctoral student: Sahba Zojaji , Beräkningsvetenskap och beräkningsteknik (CST)
Opponent: Professor Tony Belpaeme, Ghent University, Ghent, Belgium
Supervisor: Professor Christopher Peters, Beräkningsvetenskap och beräkningsteknik (CST)
QC 20240212
Abstract
In both physical and virtual environments, small group interactions significantly shape our social experiences. Understanding and replicating situated group interactions with virtual agents or physical robots pose possibilities and challenges. Central to these challenges lies the critical aspect of politeness, which serves as a fundamental cornerstone in shaping our social interactions. This PhD thesis investigates the profound significance of politeness strategies in shaping social interactions within small free-standing conversational groups in physical and virtual environments with humanoid artificial agents. It particularly focuses on their adaptability to virtual characters or Embodied Conversational Agents (ECAs) and Human-Robot Interaction (HRI). The thesis explores the impact of these strategies on persuasiveness, adherence to social norms, and the formation of positive perceptions during interactions between participants and virtual agents or humanoid robots while joining a group of artificial humanoid agents. It involves a series of user studies with an experimental setup, which entails presenting participants with dilemmas in either a virtual or physical environment. Participants should decide between expending more effort to comply with the agent’s request or opting for a least-effort alternative to join a group while ignoring the request. Additionally, the setup evaluates participant responses to various politeness strategies expressed by virtual agents or robots when extending invitations to join a small, free-standing group. The research contributes by defining behaviors aligned with politeness strategies, revealing participant adherence to social norms even in situations requiring more effortful choices, and pinpointing optimal behaviors based on criteria such as persuasiveness, politeness, and social adherence. Ultimately, the findings provide insights into the indispensable role of politeness strategies in Human-Agent Interaction to gently influence the decisions of humans while maintaining positive relations with them. These insights pave the way for designing more effective and socially acceptable behaviors for virtual agents and robots across diverse domains.