Interaction with Agents and Robots: Different Embodiments, Common Challenges

August 27, 2017, Stockholm, Sweden

Welcome to IAR 2017, a new workshop aiming to bring together researchers working on virtual embodied agents and researchers in Human-Robot Interaction. The goal is to allow both groups to share information on approaches to their common problems of how to produce believable autonomous interactive agents. Additionally, this workshop will explore the tradeoffs between virtual and physical embodiments and how differing embodiments can be combined to achieve natural and effective interaction with people.

About the IAR Workshop

As hardware and enabling technologies such as speech recognition improve, virtual agents and robots are increasingly able to autonomously take part in “face-to-face” interactions with people. But engaging in these types of interactions successfully requires the complex coordination of verbal and nonverbal behavior, as well as the ability to respond to the verbal and/or nonverbal cues of a human interaction partner.

Researchers in the field of virtual agents and human-robot interactions have been concerned with these problems, and these fields provide many examples of implemented software systems that can give useful insights for researchers of the two communities. However, while the HRI and agents communities share research interests, there are differences in methodology and focus. This workshop seeks to increase communication and knowledge sharing between these communities. What can the agents community learn from HRI about the evaluation of people’s attitudes towards and acceptance of embodied agents? And what can the HRI community learn from the agents community about designing control architectures for embodied conversation?

Research on embodiment in agents suggests that whether an agent is embodied virtually or in hardware can influence how people respond during interaction. But in many real world applications such as assistance in domestic environments, it may be most practical for people to interact with both robots and software agents under different conditions. How to design and realize these systems of hybrid embodiment, as well as how to study people’s impressions of them, are open questions for research. Should differing embodiments have distinct agency/personalities? Or should an agent migrate across them while interacting with a user?

The goal of this workshop is to enable a cross fertilization of ideas and solutions for the issues encountered when attempting to generate behavior coordination between machines and humans, that can be used independently from the embodiment of the agent. Similarly, we aim to better understand the limitations and benefits of physical or virtual embodiment for different interaction scenarios. This approach will enable a discussion about the nature of the differences between software and robotic agents, not only with respect to their construction and scientific use, but also to their fields of application.

Looking forward to see you at IAR 2017!
Mathieu Chollet, Ayan Ghosh, Hagen Lehmann, Yukiko Nakano