Bridging the Digital Divide with AI: Conversational Analyses of Older Adults’ Social Engagement with Socially Assistive Robots

Objectives: This study explores how older adults in low-resource communities engage with the socially assistive robots (SARs) named Hyodol, focusing on emotional expressions, conversation topics, and their relationship to activity participation—key dimensions of well-being in aging populations. We hypothesize that the interplay between conversation topics and emotional states influences patterns of activity participation.

Methods: We collected log data (content usage and tactile interactions) and conversational voice data (audio recordings) from users via sensors embedded in the Hyodol SAR. Additionally, pre- and post-surveys gathered demographic information, social isolation, depression levels, and health status changes. To assess SAR usage patterns, we analyzed the frequency and duration of interactions with Hyodol over a 90-day period, collected through a web-based monitoring system. The collected voice data was transcribed into text, resulting in 3,212 meaningful conversation segments, which served as a foundation for analyzing the role of SAR-facilitated conversations in promoting various activities among older adults. We systematically examined key elements, including users' emotional expressions (positive, neutral, negative), conversational topics (self, others, Hyodol SAR), and their relationship with activity participation patterns. Applying Levasseur’s classification, activity participation was classified into the following six types: (1) personal, (2) productive, (3) physical, (4) leisure, (5) social, and (6) civic activities. Sentiment analysis was employed to categorize emotional expressions. Data mining techniques identified usage patterns, and advanced AI-driven analyses examined demographic differences, conversational engagement levels, activity participation behaviors, and health-related outcomes. Using GPT-4 for persona construction, we synthesized key traits for each persona, integrating qualitative and quantitative insights.

Results: Among participants, 44.6% engaged in conversations with the SARs, with 30.2% discussing their activity participation. K-Means clustering of 67 participants who engaged in activity participation identified three user segments: Social Butterflies (n = 19, 28.35%) maintained balanced engagement in social and personal activities, with positive emotional exchanges but limited long-term impact on well-being. Lone Wolves (n = 28, 41.79%) had low social engagement yet showed notable improvements in emotional well-being through conversational interactions with the SAR. Emotional Peacocks (n = 20, 29.85%) displayed high emotional and sensory engagement with the SAR, demonstrating the greatest reduction in loneliness among the three groups.

Discussion: This study provides an initial exploration of conversational engagement between older adults and SARs, highlighting the interplay between activity participation, emotion, and verbal expression. Addressing the challenges of the digital divide and limited digital literacy, this study emphasizes that SARs can serve as accessible and interactive AI-driven tools for older adults. By segmenting users based on their interaction styles, this research offers valuable insights into the diverse ways older adults adapt to and benefit from AI technologies, informing the development of SARs tailored to their unique social and emotional needs.