How do older and younger adults learn health information with voice interfaces?

Description

As conversational agents (CAs) become more available and affordable, it is essential to discuss their impact on various application domains, including learning. For this purpose, we designed a conversational agent, “Health Buddy,” capable of delivering learning instructions using three self-regulated learning (SRL) learning strategies: monologue, dialogue-based scaffolding building, and conceptual diagramming. In this study, we investigate the (1) feasibility of using commercially available smart speakers for self-regulated learning (SRL) in informal learning environments and (2) the effects of perceived sociability and social presence on the adoption of smart speakers using Computers are Social Actors (CASA) framework. We conducted a mixed-factorial design experiment with 26 younger adults and 25 older adults. For (1), we find older and younger adults have equivalent learning outcomes; however, younger adults benefit more from scaffolding building and conceptual diagramming than younger adults. Interaction fluency, operationalized as the number of conversational breakdowns, also impacts learning performance. For (2), we use two types of interactions: human-human interaction (human experimenter collects measures from participants post-learning) and human-agent interaction (conversational agent collects measures from participants). Older adults perceived the agent as more favorable when the conversational agents interviewed them. Furthermore, we also find associations between pleasant social experiences and technology adoption. Finally, we discuss the design implications of (1) and (2) to create more accessible conversational agents.