JMIR Publications

ABSTRACT

Background:

The global prevalence of dementia is increasing at a substantial rate, underscoring the urgent need for effective treatment tools. As cognitive interventions become more accessible, non-pharmacological treatments such as digital cognitive training are gaining popularity. Past studies have predominantly focused on cognitive training applications, but their effectiveness can be limited by users' digital literacy. To address this challenge, we incorporate social interaction features, reward systems for app engagement, and AI conversational agents. In this study, we introduce a software robot called Saemi Talk, an AI conversational agent designed to facilitate cognitive training and engagement.

Objective:

We developed a 12-week digital cognitive training program, Care & Cure, which incorporates a chatbot service, Saemi Talk, and a group chat service, Our Town. We conducted a field study to address two research questions. The primary objective aimed to determine whether the Care & Cure program enhances cognitive function in older adults. The secondary objective focused on assessing if Our Town improves social engagement by catalyzing Saemi Talk usage and fostering interaction between the AI agent, a software robot, and users.

Methods:

A total of 133 participants were recruited for this study (age range: 51–83 years; mean age: 64.75 years). None of the participants were diagnosed with dementia. All participants underwent a three-month intervention using the Care & Cure program. The primary outcomes included changes in Korean-Mini Mental State Examination-Version 2 (K-MMSE-2) scores following three months of program participation. The secondary outcomes encompassed participant log data, degrees of social support (Medical Outcomes Study-Social Support Survey; MOS-SSS), depression scores (Short Form Geriatric Depression Scale-Korean Version; SGDS-K), and engagement scores (The Twente Engagement with eHealth Technologies Scale; TWEETS).

Results:

The Care & Cure intervention substantially enhanced cognitive functioning, as evidenced by statistically significant increases in K-MMSE-2 (Hedges’ g = 0.35, p < .001) scores. Active users showed statistically significant improvement in cognitive function compared to non-active users, particularly those with lower baseline K-MMSE-2 scores (Hedges’ g = 0.99, p < .001). Emotional and informational social support showed statistically significant increases (t = -6.509, p < .001), alongside statistically significant increases in participants’ willingness to engage with the program (t = 2.008, p < .05) and decreases in depression levels. Regression analyses demonstrated that engagement in group chat positively influenced cognitive training usage (t = 12.395, p < .001), which in turn improved social support scores (t = 4.102, p < .001) and cognitive function (t = 2.467, p < .05). However, using cognitive training programs alone did not result in a statistically significant improvement in cognitive function. Adherence levels fell below expectations, with participants averaging 0.47 game sessions per day for Saemi Talk and 0.56 responses per day in the chat room for Our Town, indicating the need for strategies to enhance engagement and optimize intervention benefits.

Conclusions:

The Care & Cure intervention effectively enhanced participants' cognitive function, social support, participation, and emotional well-being. Analysis revealed that strengthening social support was essential for achieving cognitive gains. Clinical Trial: N.A.