The Future of Chatbots in Behavior Change

Consumers are increasingly interacting with computers via chatbots. These computer programs use Natural Language Processing (NLP) to quickly process data from emails and texts to spoken words. This is becoming common in daily life for computer functions such as search, spelling and translation apps or digital voice assistants (Alexa) and even organizing email files. Governments, businesses, education, healthcare and other sectors are now incorporating chatbots in their communication strategies. 

Current use of chatbots in behavior change

Healthcare research studies on behavior change indicate that the use of chatbots shows promise. Currently, chatbots used in health improvement primarily focus on nutrition and physical fitness and mental health. This technology is easily accessible, available anytime via smartphone apps, can be scaled up to large population groups and is relatively inexpensive to implement as a method for supporting health interventions. It can serve as a useful tool to integrate data from smartphones on exercise or heart rate and provide reminders to reach goals. Simple questions can prompt a reminder toward action. One example, the Wholistics Chatbot, utilizes this technology to provide “micro-doses” of evidence-based wellness content for employee health programs on the four dimensions of health: diet, exercise, sleep and mood.

For healthcare providers, chatbots can expand support to patients and their caregivers by answering basic questions, provide links to other resources, assist with appointments including reminders (e.g., improve medication adherence) as well as alert healthcare providers on patients requiring immediate attention. This can be especially helpful in rural areas with healthcare shortages.

Over the past few years, the demand for mental health services has increased globally. Chatbots have been developed to provide a communication tool to people with anxiety or depression and utilize specific psychological methods to prompt behavior modification. Some use facial expressions and changes in speech to enhance the Artificial Intelligence (AI) interpretation of the communication and potential diagnosis. The chatbot uses a sequential guide to encourage patients to notice negative thoughts and steer toward more positive ones. More people are accessing these programs because they offer anonymity for the recipients and no judgment. This technology can support mental health programs, but it can’t be used as a substitute for clinical care from a licensed provider.

Limitations of current chatbot technology

Though chatbot technology is emerging in the healthcare arena, the limitations require attention. People are social beings and require a sense of empathy and bonding, especially during times of distress. They have changing needs which requires clinicians to regularly monitor the chatbot – human interactions are still an essential part of the healthcare experience. Chatbots that offer minimal support and narrowly programmed responses limit the effectiveness. People have noted that the AI responses aren’t nuanced and don’t understand the person’s perspective. 

AI linguistics require careful programming and data review to broaden the capability of the communication. The machines must be trained to understand the nuances of human conversation such as the meanings and relationships between words, the evolving use of slang terms, etc. Furthermore, providing this technology in other languages and complying with various regulations by different countries multiplies the challenges. Data security is a big concern. All of this can be a costly time investment. 

Future possibilities

To address these limitations, researchers in health behavior developed the “AI Chatbot Behavior Change Model” including methods to:

  • develop chatbot characteristics framed around the background and habits of the user
  • build conversations
  • integrate behavior change theories into programs
  • evaluate outcomes

Chatbots can be programmed for conversations ranging from pre-scripted interactions that restrict responses, to more open, conversational style. These tools can provide more individual responses based on the human-to-chatbot conversations over time. In this way, the tool can be used to learn individual behavioral patterns and then encourage them toward healthier choices and provide ongoing supportive conversation to continue that behavior. This is referred to as “persuasive technology” that goes beyond basic conversation toward behavior change.

During the design and testing phases, input from representatives of the target population can help determine relevant features and make the chatbot more accessible to all population groups. Other important considerations in this model include informing the users:

  • who created it and why
  • how the chatbot works – data collection, storage and use
  • privacy standards and procedures
  • consent process

Using a systematic model to develop these AI tools can increase the effectiveness of health behavior change strategies.

Using chatbots to support people living with Diabetes

Wholistics Health partnered with the Cabarrus Health Alliance (CHA) and One Charlotte Health Alliance (OCHA), to run a series of pilots to determine if chatbots were useful tools to support diabetes prevention and management, especially in higher risk populations. Users were recruited from the NC Minority Diabetes Prevention Program (NC MDPP) and Mecklenberg County’s Food Pharmacy and Mobile Health Program. Participants were sent daily text messages to their phones with ideas and suggestions on healthy eating, exercise, sleep and mental health. The chats included ad-free links and videos to support their efforts. The chatbot provided support to the MDPP in-person classes and promoted continued participation during the course. Incentives were also provided to encourage completion.

The successful pilot program offered the following highlights:

  • 92% reported that they were eating and drinking healthier foods, had increased their weekly exercise, and improved their mood
  • 67% reported that they were sleeping better
  • 75% found that the daily tips were very useful
  • 78% would recommend this program to others

The pilots showed that frequent “touchpoints” via the chatbot encouraged the participants to make small, incremental changes to their lifestyle and supported the in-person wellness program. It encouraged them to try new things such as yoga and make changes to their diet. Overall, it showed that chatbots can be a useful tool in behavior change efforts towards improved health.

“The chatbot program is a great complement to our NC MDPP classes as it provides meaningful touchpoints with practical information to participants across all dimensions of health to support and promote positive behavior change. We are also excited to roll this out to alumni in order to engage with them in a sustainable manner after graduation,” said Kamaria Mason, NC MDPP Program Lead, NC Office of Minority Health & Health Disparities, NC Department of Health & Human Services.

According to Helen Pak-Harvey, Founder and CEO of Wholistics.Health, “The Wholistics Chatbot Program is designed to make it easier for busy people to make small, achievable changes in their life that can have a lasting impact. If you have diabetes/pre-diabetes or a chronic condition, our chatbot can provide you with a range of evidence-based tips to promote healthy eating, regular exercise, stress and mood management and better sleep.”

Chatbots have a role in healthcare

Well-designed and implemented chatbots like Wholistics can be a useful, accessible, cost-saving AI application for patients, their caregivers, and healthcare providers. Employee assistance programs (EAPs) can employ chatbots to encourage staff to participate in exercise, diet, smoking cessation and other wellness classes. This can boost employee morale and reduce the burden of chronic disease and hospitalization costs. The health insurance industry has started using chatbots to improve customer service, reduce costs, and streamline claims to reduce treatment delays. And importantly, chatbot technology can support individuals to change behavior and improve their overall health and wellbeing over time.