A summary report through home experiment on a dialogue system for older adults

Abstract

Purpose Dialogue systems for older adults become more important because some of the older adults living alone have fewer chances to do that because they are alone; because of the COVID-19 situation, they have decreased the opportunity of conversation in the world. However, for academic studies, dialogue systems have been evaluated at laboratories. This is because the experimental conditions can be standardized. Moreover, on-site support is easier for the experimental staff than home-based experiments because some older adults are not familiar with ICT technologies such as tablets, robots, etc. Some researchers tackle developing a dialogue system specializing for older adults by Inoue et al. (2020); however, there is no information on what kind of robot interaction is effective for the cognitive function to the best of the authors' knowledge of the elderly. Hence home-based dialogue system experiments have become much more important. Method In a total, 11 groups consisted of a maximum of 8 persons per group, and the experiments were conducted with 1 group sequentially. The experiment is conducted along with a randomized controlled experiment with two sub-groups. An intervention group participant listens to the story and asks questions to the robots (Bono-06 reported in Tokunaga et al. ( The other is a control group of participants who only listen to stories and speak a short greeting to the robots, which indicates they participate in the session within 1 minute. The experiment was designed to examine the effect of difference for cognitive training by listening to the stories and asking questions or just listening to the stories. In the experiment, participants listened to a total of 12 stories over a period of 2-3 weeks with our specializing system (DBSPS), as shown in Figure Participants participated in 2 sessions per day at their home, except for the first day, where we had an exceptional practice session and the other two sessions. We remotely checked the progress of experiment status of the participants that we monitored with Slack bots (See Center of Figure The essential of our system is reported in Tokunaga et al. (2021). Results and Discussion We have completed our experiments with 81 participants. The participants include those who have not familiar with smartphones or tablets. There were no lost to follow up. We found that the number of utterances is varied for each topic (Figure The average number of questions seems to vary because some topics were easy to ask questions, others were not. We have confirmed through the interviews that many participants have enjoyed our experiment because due to COVID-19, older adults have less schedules than usual. Hence the participation to the experiment becomes pleasurable events in daily lives. We have received remarkable comments that some participants said, "I was glad to join the experiment so much." We have also received negative comments, "I felt thinking about the question was difficult" and "I felt nervous in the experiment." Those comments would be helpful to be designed for a home-based experiment.