Abstract
Objectives: Discourse assessments are widely used in clinical settings to identify cognitive and linguistic impairments associated with normal aging and neurogenic communication disorders. Because richer spontaneous speech facilitates more reliable discourse analysis, effective methods for eliciting naturalistic language are needed. As the social roles of robots expand, their potential as tools for language and cognitive assessment has gained increasing attention. This study examined whether a conversational robot elicits greater discourse production and conversational engagement in older adults compared to a tablet and explored the clinical usability of robots as tools for discourse-based assessment. Methods: Fourteen healthy older adults participated in unstructured conversations on three daily topics with both a robot and a tablet. The robot incorporated non-verbal backchannel behaviors, such as nodding and facial expressions, and interacted with older adults using the ChatGPT model, while the tablet offered only verbal communication. Five linguistic measures were analyzed, including turn-taking frequency, total and mean utterances, and total and mean speech duration. Following each condition, participants completed a selfreport survey assessing usability and interaction experience. Results: Results showed that older adults produced significantly more total and mean utterances when interacting with the robot. Additionally, participants reported a greater sense of encouragement to engage with the robot. Conclusion: These findings suggest that the robot’s physical embodiment and interactive behaviors enhance conversational engagement, making it a clinically promising tool for eliciting spontaneous speech and collecting discourse data for language and cognitive assessment in aging populations.
Funding Information
  • National Research Foundation of Korea (RS-2022-NR070151)
  • National Research Foundation of Korea (RS-2024-00461617)
  • Ministry of Science and ICT, South Korea
  • Ewha Global Excellence Program
  • Korea Institute of Science and Technology (2E3384)

This publication has 29 references indexed in Scilit: