Each year, thousands of research papers and medical studies on human behavior are released. But tracking down the right research—whether on smoking, exercising, or sleeping—can be as challenging as getting people to follow their advice.

Like the brain itself, “we’re only utilizing a fraction’s fraction of what’s out there,” Dr. Emma Norris, a research fellow at University College London’s Centre for Behaviour Change, told Industrious.

And while Norris said her Twitter feed is one of the best places to keep track of new papers, her lab, with the help of a group of natural language processing and AI experts from IBM Research – Ireland, is now harnessing another type of tech to address behavioral research: AI.

Join IBM at HIMSS20 to discover how data and AI can transform health

“Our work requires bringing together very large amounts of data in order to answer questions about what works, for whom, in what settings, for how long and for which behaviors,” Dr. Susan Michie, UCL Centre director, said. “To help answer complex questions and analyze evidence at scale and in timely fashion, technology holds great promise.”

Three years ago, Michie’s team at UCL joined forces with IBM Research – Ireland, Cambridge University and Aberdeen University in Scotland to launch the Human Behavior-Change Project.

The team has already indexed more than 220 million relevant articles.

“An impressive feat,” Michie said.

AI-powered health outcomes

The system will be constantly updated as new research is released. AI will both manage these vast flows, but also analyze them, extract information from them, and interpret them. That will help fill in gaps while refreshing older evidence—an ongoing challenge.

“Even when papers are published, the research and review can take so long, the findings are already kind of out of date,” Norris said. “Computers can keep up in ways we can’t.”

What most excites the team is not the database but what they can do with it. A series of specialized online portals is already in the works. There, public health officials, insurers and doctors can seek guidance on how to change behavior, for example, on how to combat bad habits and form better ones.

And with the power of AI technologies developed by IBM Research, the results can be customized by age, demographics, geography and other inputs. The system will also predict what can work for new population settings—populations that have never been studied.

“There’s a host of different factors that differentiate one study from another, even if they seem similar,” Norris said.

Consider how getting a 17-year-old to stop smoking or overeating differs from getting a 71-year-old to do so. And these individuals, whether in Dublin or Dubai, will respond differently to particular influences. Teens in one location may be more attuned to messages from pop stars, while others might respond to peer pressure, or religious figures.

Changing behaviors with algorithms

AI can also help fill in the gaps where no specific research exists. If the health department in Dubuque, Iowa, wants a smoking program, and there’s no specific data for the area, the project will extrapolate from similar populations.

Norris said that AI is already influencing behavioral health in unexpected ways.

Because computer algorithms benefit from a defined range of ideas and language to draw from, Michie and her partners are developing the Behavioral Change Intervention Ontology, the field’s first standardized set of concepts and relationships between concepts for behavioral interventions, treatments and therapies.

“People are describing interventions in different ways, which makes even a basic understanding challenging,” Norris said. “Having researchers speaking the same language, and knowing what each other is talking about, is hugely beneficial.”

Overall, the team believes that the AI augmented system could help public health officials, insurers and individuals develop more preventative health practices and policies. That means healthier people as well as fewer visits to the doctor and medical bills.

And since massive public funding has already  been invested in past research, increasing access and understanding makes that funding go that much farther.

“Think of how much time and money goes into new studies finding subjects, setting up double-blind trials, getting approvals, and so on,” Dr. Francesca Bonin, a Natural Language Processing specialist at IBM Research – Ireland, said. “While funding new studies is essential, already so much research waiting to be put to use.”