Our desire to exercise, even just to get in a few extra steps each day, often meets the reality of our motivation. Hearing the right words can often be the spark that gets us back into our sneakers, and those words might not have to come from another human. 

Jingwen Zhang, a professor of communication in the College of Letters and Science at UC Davis, has been developing and testing artificial intelligence (AI) chatbots designed to motivate and persuade us to get in those extra steps for health. For this work, Zhang has received the 2023 K. Everett M. Rogers Award from the American Public Health Association.  

“Given what we have with today’s technological capacities, our goal is to design AI-based interventions to reach needed populations better with information to increase the efficacy of health interventions that we know work,” said Zhang.

Putting the human in artificial intelligence 

An artificial intelligence (AI) model is a computer program that analyzes a set of data to respond to prompts without a human’s direct guidance. The most familiar AI model is probably ChatGPT, a chatbot that can answer questions based on everything it has accessed and “understood” on the Internet. 

But AI chatbots have the potential to be much more than a personal assistant, web research shortcut or just fun to chat with. Zhang is developing an AI model that improves our health with more than just reminders to go out and move.  

Zhang’s AI chatbot crafts personalized, persuasive messages that motivate us to act. Zhang began this line of research in 2019 with the question of how a chatbot should be if it is to succeed in changing people’s decisions on whether to exercise. It came down to two characteristics: persuasiveness and an ability to relate on a personal level. 

Chatbots simulate the experience of interacting with another human being so well that earlier this year there were debates on whether or not AIs were conscious. This ability of an AI model makes it ideal for combining the benefits of mass communication and the meaningfulness of an individual human interaction.  

“A chatbot is different than mass media communication campaigns because the conversation is at a very interpersonal level,” said Zhang. “You know consciously it’s a computer, but you still use interpersonal communication norms." 

Chatbots motivating better choices through personalization 

Zhang’s work with AI chatbots has grown tremendously since 2019. She and her team now draw from multiple research disciplines, including data science, linguistics, public health and her own field of communication.  

Her first field experiment used an AI chatbot that delivered a single session of 10-minute conversation to the study’s participants. It increased the intention to exercise by 37%. According to Zhang, adding to the duration and dynamic nature of the interactions could have a much stronger effect. 

Zhang and her team are currently testing a one-week intervention through a mobile app that hosts an AI chatbot. The app tracks daily steps as well as how the AI chatbot conversations unfold. This experiment compares two different types of chatbots. One only has the ability to persuade. The second adds more interpersonal communications, like humor and relationship building. 

“When you talk with an AI model that has more of this interpersonal understanding and uses common interpersonal human techniques like humor, empathy and rapport, it increases this positive perception of the AI model,” said Zhang.  

Imagining a future of public health that includes AI 

The American Public Health Association’s K. Everett M. Rogers Award honors contributions to research and practice in public health communication. Rogers was a prominent sociologist and communication scholar whose work focused on health promotion and education. He studied how innovations spread throughout society and lead to social change. 

For Zhang, the award is particularly fitting because of its namesake. She has admired and built upon Rogers’ work since she began her own research career. There was no AI in Rogers’ time, but his research has served as part of a strong foundation for how Zhang thinks about how AI models can change our lives for the better.

“We are going to increasingly interact with AI models in the future,” said Zhang. “We have to really think about the ideal scenario of how humans interact with them and how AI models are designed to benefit humans.”