What is it about?
The paper focuses on generating adaptive behaviors of virtual agents (or SIAs) in human-agent interaction. The capability of adaptation (i.e. reciprocal adaptation) that is seen in human-human interaction is learned and endowed to the virtual agent via a deep learning model, our ASAP model.
Featured Image
Why is it important?
The modeling of reciprocal adaptation is done by considering both interlocutors of a dyadic setting, learning the interpersonal relationship from multimodal signals. The technique allows the generation of reciprocally adaptive virtual agent behaviors at the frame-level from a small amount of data.
Perspectives
The paper presents the generation of adaptive virtual agents that can be applied in various fields. I hope this paper makes people think about not only the positive practical impacts of virtual agents but also the social implications that they might have that are not always positive such as people avoiding real human interaction and replacing it with virtual agents. We should always keep in mind that virtual agents are here to assist us and not to replace human interaction.
Jieyeon Woo
Read the Original
This page is a summary of: ASAP: Endowing Adaptation Capability to Agent in Human-Agent Interaction, March 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3581641.3584081.
You can read the full text:
Contributors
The following have contributed to this page
