People seeking connection and support during challenging times have long turned to support groups. A new study from the University of Kansas and the University of Southern California explores the behavioral markers of alliance in these settings, particularly in virtual environments. The research, published in the Proceedings of the 27th International Conference on Multimodal Interaction, identifies how nonverbal behaviors indicate whether participants are forming connections.
The study analyzed data from 18 support groups consisting of 96 participants. Researchers measured the dyadic alliance, which refers to the connection between two individuals, by surveying participants about their feelings of connection. Utilizing computational algorithms, the study examined verbal and nonverbal communication features, including language, audio, and visual components.
As the demand for mental health services continues to rise, particularly following the COVID-19 pandemic, this research highlights the potential role of artificial intelligence in mental health settings. According to the researchers, professionals have struggled to meet the increasing demand for support groups and counseling.
Yunwen Wang, an assistant professor of journalism and mass communications at the University of Kansas and one of the study’s authors, noted, “This project came into inception when we were looking at burnout among mental health professionals, especially with the rising need for support groups.” Wang emphasized the lingering effects of the pandemic on mental health and the ethical opportunities presented by AI to enhance access to mental health services.
Participants in the study were college students dealing with general anxiety. Their mental health and emotional states were measured before and after attending a virtual support group session facilitated by a conversational agent with a robot embodiment. This virtual agent, operated by a human, engaged participants in discussions about academic stress and related topics.
To analyze the sessions, researchers transcribed participants’ verbal communication and recorded nonverbal gestures, including head nodding, smiling, and facial expressions. They assessed various communication features, such as the frequency of head nods and variations in pitch, to determine their relationship with dyadic alliance.
The findings showed that certain nonverbal cues, such as frequent head nods and brow raises from listeners, significantly predicted the level of connection felt by speakers. Conversely, speakers who displayed pitch variation and head pose changes also reported stronger feelings of alliance.
The study advocates for incorporating both verbal and nonverbal communication when assessing alliance in support group settings. It builds on previous research that indicates strong alliances can enhance overall group engagement and positive outcomes. The researchers suggest that machine learning could assist in identifying behavioral markers of dyadic alliance, though they caution against unrestricted AI use in mental health contexts.
Wang clarified, “The goal wasn’t to replace humans or to compare human versus AI-assisted mental health support facilitators. We are comparing machine learning fusions of communication features to understand how humans perceive their relationships within the group.”
As discussions around AI in mental health continue, including legislative considerations in states like California, the research team is exploring critical questions about the ethical boundaries of AI applications. Wang’s ongoing work involves assessing the trust participants have in AI agents and the outcomes in contexts where AI plays varying roles.
Ultimately, the study aims to enhance mental health services by identifying how participants in support groups foster genuine connections. Wang concluded, “With many people needing mental health support and a limited number of trained professionals, the goal is to see if AI-assisted systems can be perceived as acceptable by users. The human-to-human dynamic is essential for sharing experiences and providing empathy.”
This research represents a vital step toward understanding AI’s role in mental health, particularly as it seeks to facilitate rather than replace human interaction in support group settings.
