Smoking-cessation expert Dr. Jonathan Bricker, a public health researcher with Fred Hutchinson Cancer Research Center, just received a $3.65 million grant from the National Cancer Institute for what may be his most innovative behavior modification trial yet. The five-year study will test a “conversational agent” to help participants stop smoking.
What on earth is a conversational agent? Bricker described the new tool, which he terms a “quitbot,” as part therapist and part chatbot.
“The bot is a virtual therapist who helps you stay motivated,” he said. “It works with you to set goals, offers encouragement and support when you need it, teaches you skills to cope with cravings and helps you recover from lapses.”
A behavioral psychologist who specializes in smoking-cessation research, Bricker said the foundation of a solid therapeutic relationship is the sense that your therapist understands you, that you have the same goals, that you agree on the tasks for changing behavior and that you feel a bond.
“We took the idea of ‘creating the therapeutic alliance’ and brought it into a new technology platform,” he said. “That’s what this conversational agent is all about.”
His new creation, which he’s dubbed Ellen, is much savvier than ubiquitous chatbots like Siri or Alexa.
“Those have a very utilitarian purpose,” he said. “You ask it a question and it gives you an answer. ‘What’s the weather like today? Call my friend Joe.’ There’s no relationship. There’s just question and answer.”
This conversational agent, on the hand, will utilize natural language processing, or NLP, cloud computing technology and the tenets of effective therapy to create a bond with its users. The quitbot will check in a couple of times a day. If you don’t want to interact, it will back off. If you do, it will offer praise and encouragement, share charts that show your progress and, like any good coach, teach you skills to stay motivated or deal with cravings.
It even has a bit of a personality, Bricker said.
“It’s all via text — there’s no voice — but Ellen will ask questions, such as, ‘Does this make sense? Is this helpful?’ She’ll use a little humor, some wordplay and puns, or sometimes even talk about herself,” he said. “It’ll be like you’re texting with someone. It’s designed to create the impression that you’re working with someone who’s thinking about you, who never forgets anything about you.”
Early participants who tried out a prototype “reeeeally liked” the new technology, Bricker said. As with much of his habit-busting research (and resultant apps) Bricker incorporated cutting-edge behavioral modification principles into Ellen’s design. Think “smart” therapy plus NLP.
The result? People felt like they were texting back and forth with a real person.
“They wanted more of her. They wanted her to be more real. People felt very connected to her,” he said. “It’s just a program but I’ve found that even I’m personifying her. Calling her a ‘her.’ People said things like, ‘She was always there when I needed her’ or ‘She held me accountable’ or ‘She was kind and nonjudgmental.’”
Bricker said that kind of personification isn’t unusual.
“Everybody knows it’s a machine, but humans want to connect,” he said. “It’s naturally what we do. Even though we know we’re interacting with a computer, when we get these kinds of verbal interactions, these conversations, we create a bond with a machine.”
Another feature of the quitbot: It was purposely designed to be accessible to everyone.
“The idea was to have it be reachable to a really broad spectrum of people,” Bricker said. “It’s a very low-tech option so people with low tech literacy will be able to use it without any special skills or ability.”
Cigarette smoking accounts for 480,000 premature deaths and one-third of all cancer deaths annually in the U.S. There is enormous need for high-impact, cost-effective, population-level interventions for smoking cessation.
Bricker’s new quitbot may be just that. The National Cancer Institute grant is designed to find out.
The new randomized, controlled trial will pit the quitbot Ellen against another texting program known as SmokefreeTXT, a text messaging service offered by the NCI. Each study arm will recruit 760 smokers and follow them for a year to determine whether Ellen is better at getting people to quit and is more cost-effective than SmokefreeTXT. The researchers will also assess the theoretical mechanisms and the effectiveness of conversational agents for health behavior change in general.
The trial will begin in April of next year. Results will not be available for some time.
Bricker said he was grateful to the now-defunct Hutch Data Commonwealth for its design help and now to Microsoft’s AI for Health group which, through a previous Hutch research arrangement, is helping his team to fully customize a conversational agent framework as the basis for the Hutch quitbot program.
He also expressed appreciation to the NCI.
“This is a blend of NLP and cloud computing technology and state-of-the-art smoking-cessation treatment and the tenets of effective therapy,” he said. “It’s bringing them all together in one platform, which is really cool. I’m delighted that the NCI is supportive of this. It’s helping us stay on the cutting edge of behavior change technology with the potential to make an impact on public health.”
Bricker also stressed the importance of kicking a habit that’s recently become even more dangerous, thanks to the severe acute respiratory virus, SARS-CoV-2.
“A recent NEJM report suggested that smokers are at a much higher risk of developing COVID-19,” he said. “They were more likely to get it and die from it. It’s not just about cancer. There’s a much more acute and immediate consequence.”
Diane Mapes is a staff writer at Fred Hutchinson Cancer Center. She has written extensively about health issues for NBC News, TODAY, CNN, MSN, Seattle Magazine and other publications. A breast cancer survivor, she blogs at doublewhammied.com and tweets @double_whammied. Email her at email@example.com. Just diagnosed and need information and resources? Check out our patient treatment and support page.