Development of RehaBot-A Conversational Agent for Promoting Rewarding Activities in Users With Traumatic Brain Injury.

Objective: In this article, we describe the development and preliminary testing of RehaBot-a chatbot that users communicate with via text messaging designed to augment behavioral activation (BA) treatment of reducing depression and increasing participation in individuals with moderate to severe traumatic brain injury (TBI).

Setting: Outpatient brain injury rehabilitation facility.

Participants: Outpatient brain injury clinicians and individuals with moderate to severe TBI.

Design: Focus groups, software demonstration trials, and single-case experiments with an A-B-A design.

Main Measures: System Usability Scale (SUS) and self-reported completed target activities.

Results: Focus group feedback guided the development of a flexible system to be used in conjunction with face-to-face therapy, designed to provide reminders, encouragement, and supportive feedback. Two of 3 participants completed all 6 RehaBot tasks independently. One completed 4 tasks independently and a fifth with assistance. Average SUS score in demonstration trials was 76.77 (SD = 21.19). In 4 single-case experiments, participants completed their highest proportion of planned activities in the RehaBot phase and the lowest proportion of planned activities in the reversal phase. They all interacted with RehaBot daily, exchanging an average of 225 messages over the 1-week period, and average SUS score was 95 (SD = 4.74). Their open-ended feedback revealed that participants found RehaBot enjoyable and easy to use. They felt it was a helpful memory aid and promoted better adherence to planned activities by providing accountability and positive reinforcement.

Conclusions: This article presents a chatbot development process heavily involving consumer input, which may serve as a model for future development efforts. Our findings provide preliminary evidence suggesting that RehaBot is usable and may promote better adherence to planned target activities. However, future research is needed to establish usability and efficacy of RehaBot and to explore applications of chatbots to other domains of TBI rehabilitation.