Peoples Bank – 7-month CD – thru 7-1-24

NIMH: Chatbot encourages people with eating disorders to seek care

In Community, Education by OC Monitor Staff

BETHESDA, Md. — Eating disorders are serious and often fatal illnesses associated with severe disturbances in people’s eating behaviors, thoughts, and emotions. Research has found that delaying treatment results in poorer outcomes for people with eating disorders. Despite this, less than 20% of people with such disorders ever receive treatment. Tools that encourage and assist people with eating disorders to engage with mental health services are critical to helping them get the care they need.

In a new National Institute of Mental Health-funded study, Ellen Fitzsimmons-Craft, Ph.D., an associate professor of psychiatry at Washington University School of Medicine, and colleagues developed a chatbot to encourage people with eating disorders to connect with care. Chatbots are computer programs designed to simulate human conversation. Research suggests that people respond to chatbots in the same way they respond to humans and that chatbots can be an effective way to reach a wide range of people who need assistance.

In this study, the researchers designed a chatbot named “Alex.” This study is the first in a planned three-part series, which includes a preparation phase, an optimization phase, and an evaluation phase. In this “preparation” phase, researchers developed Alex to include four theoretically informed components:

  • Psychoeducation: This component helped refute stereotypes about eating disorders, emphasized the seriousness of this mental disorder, and provided information specific to the eating disorder the user indicated they were experiencing.
  • Motivational interviewing: This component highlighted differences between users’ health goals and their current behavior by encouraging them to evaluate how important it was to address their eating disorder behaviors and their confidence in making changes.
  • Personalized recommendations: This component provided personalized recommendations for seeking treatment.
  • Repeated check-ins: This component included up to three check-ins in the weeks after interaction with the chatbot, which reminded users of available resources for treatment and promoted reflection on overcoming barriers to care.

The psychoeducation, motivational interviewing, and personalized recommendation components were designed to take a total of 15 minutes to complete. The repeated check-ins each took about 3 minutes to complete.

Testing the Alex prototype

The researchers held two in-person testing sessions to get feedback on a prototype version of Alex. After incorporating user feedback, the researchers tested the chatbot again in two remote-testing sessions. Participants in the testing sessions were people who had screened positive for an eating disorder but were not currently in treatment. After engaging with the chatbot, participants rated the chatbot on usability, usefulness, ease of use, ease of learning, and satisfaction. The researchers also interviewed participants to learn more about their experiences.

In the first two testing sessions, participants rated the chatbot with an average of 83.0 and 77.0 out of a possible score of 100 on usability, indicating they had an above-average user experience with Alex. The participants liked how human-like the chatbot was, with some noting that knowing they were speaking to a chatbot allowed them to open up more than they might have if they were speaking with someone face-to-face.

Participants generally offered positive feedback, while also suggesting ways to improve Alex’s components. For example, in the Motivational interview component, participants indicated that they liked how the chatbot helped them think about their ability to enact change, but they found completing several quantitative scales related to this reflection confusing. As another example, participants liked the pressure-free nature of the personalized recommendations for treatment that they received; however, they wanted the ability to select an option that allowed them to receive information on multiple types of care.

The researchers updated Alex in response to this feedback. For instance, they altered the flow of the chatbot-user conversation, improved the reflective scales, and supplied users with resources for various types of care, including individual treatment in person or via telehealth and online self-help resources.

Testing Alex “2.0”

After incorporating the user feedback, the researchers tested Alex again in two remote testing sessions. Usability ratings in the remote testing sessions, which were 75.0 and 85.8, showed a slight overall increase over ratings from the two in-person testing sessions. Only a modest increase was expected due to the high ratings in the initial testing sessions.

Participants in the remote testing sessions also received check-ins in the 2 weeks following interaction with the chatbot. These check-ins reminded participants of the available treatment resources and encouraged them to seek help-. Participants generally found that the reminders reinforced help-seeking behaviors but thought that it would be helpful to be able to schedule these check-ins. This insight offered important feedback for future chatbot iterations.

Overall, participants were open to the chatbot and able to successfully use it, suggesting its potential as a highly scalable tool to improve motivation and help-seeking behaviors among individuals with eating disorders. The researchers note that future studies should be done to determine how effective the chatbot is at improving help-seeking behaviors immediately and longer term. It will also be vital to understand which specific aspects of chatbot interaction help motivate help-seeking behaviors. In the future, the chatbot could be adapted and tested for encouraging services use among people who screen positive for other mental disorders.