Credit: Evelyn Verdin / Unsplash
What you’ll learn from this new MIT Sloan study:
- Digital tools built around generative artificial intelligence can help people reduce their body mass index at a lower cost than with medical interventions such as weight loss drugs.
- Generative AI is less effective at fostering community among weight loss participants.
- Generative AI can help reduce health inequality. The study participants who benefited the most from AI tools were less educated and had less nutritional knowledge compared with other users.
It’s a common refrain among many of us at the start of a new year: I need to lose some weight.
Nearly three-quarters of Americans are overweight, and obesity is increasing in the rest of the world as well. Demand for expensive medical treatments like surgery or GLP-1 medications is soaring, at a high cost to health care systems.
A new working paper shows that generative artificial intelligence can help people lower their body mass index, and at a low cost to boot. However, generative AI does not replace the benefits of supportive communities like Weight Watchers, where people can connect with others and openly discuss the physical and psychological struggles of being obese.
MIT Sloan professor of marketing and Singapore Management University’s Linyi Li studied 416 men and women of various ages over three weeks in late 2024.
Their research found that implementing a generative AI tool that provides real-time, personalized dietary recommendations increased the number of participants who were no longer classified as obese from six people to 17 (around 4% of the total). The number might seem small, but it is significant, Tucker said.
“Weight loss is such a big challenge. If it were easy for us all to lose weight, we’d just lose weight,” Tucker said. “The fact that a digital tool such as AI can have any effect is wonderful because interventions such as surgery or injectables are expensive. This is evidence of the cost efficacy of a very small intervention in terms of changing behavior.”
While the paper, “Building an Ecosystem or Prioritizing Personalization With AI? Evidence From a Field Experiment,” specifically studied weight loss, its findings have broader implications for how humans interact with AI agents, especially as it relates to community.
“GenAI works and is inexpensive, but it doesn’t provide an effective sense of community,” Tucker said. “The reason we build ecosystems is because people crave genuine connections. Just because it’s cheaper to produce content doesn’t mean it’s going to effectively serve the purpose of community building.”
A visual AI assist for some
To conduct the research, the authors partnered with an Asia-based Fortune 500 company that operates an online weight loss boot camp that incorporates healthy diet and exercise tips. The weight loss program includes a group chat element via the WeChat app. WeChat allows the participants to interact, share experiences, and support one another in their weight loss efforts.
Over the course of three weeks, participants were divided into three groups so that the researchers could assess the impact of a generative AI tool that evaluated participants’ meals. The tool evaluated the nutritional value of each meal component and provided personalized recommendations, such as “add more vegetables” or “reduce fat intake by choosing lean meats.”
The study comprised:
- Group 1, a control group, in which participants were given healthy-diet tips and encouraged to participate in a group chat but did not have access to the generative AI food-analysis tool.
- Group 2, a private analysis group, in which people could send pictures of their meals to an administrator via private chat to receive a personalized nutrition report created by the generative AI tool.
- Group 3, a public analysis group, in which participants were encouraged to use the AI food-analysis tool within the group chat. When a meal photo was shared publicly, all participants in the group could view the photo and the corresponding nutritional analysis report.
Here’s what Tucker and Li learned.
Finding 1: Generative AI was effective at helping people lose weight.
Compared with the control group, the two groups exposed to the AI food-analysis tool demonstrated:
- Higher participation in the chat, in addition to frequent use of the AI tool available to them.
- Increased weight loss.
- Greater reductions in individuals’ body mass indexes.
Those in Group 1 lost 0.966 kg, users in Group 2 lost 1.426 kg, and those in Group 3 lost 1.358 kg, on average.
Tucker said that the results show that generative AI is effective “for personalizing individual experiences” because it can offer input on dietary choices, provide helpful knowledge, and guide users in the right direction.
Finding 2: Public analysis dampens individual participation.
Users in Group 2 used the food analysis tool privately, while those in Group 3 used it publicly.
Being offered private access to the food analysis tool significantly increased the number of users who participated in the three-week experiment from start to finish. The authors found that making users’ responses public in the group chat discouraged other users from participating.
Of all groups, Group 3 had the highest dropout rate. Tucker said that some users may have felt overwhelmed by the high level of engagement of top-performing participants in that group, leading them to drop out of the program.
“Dropout is the big enemy of weight loss,” Tucker said. “A likely explanation [for dropouts in Group 3] is that staying in the group introduced pressure [when] consistently reporting less-favorable statistics compared to others.”
The results indicate that making AI suggestions public leads to a decline in the number of active users, perhaps because it alienates some individuals.
Community-based programs like Weight Watchers have done well for years because “there’s a set of people there to support you through good or bad weeks,” Tucker said. “I think what we are demonstrating is that if you make it too easy to post success stories, then you lose some of that [shared] vulnerability within the community.”
Generative AI Business Sprint
Attend Online
REGISTER NOW
Finding 3: Generative AI can help reduce health inequalities.
The researchers found that those who benefited the most from the generative AI tool were less educated and had less nutritional knowledge compared with other users. Such individuals often struggle to understand weight loss education and would greatly benefit from generative AI’s detailed, personalized diet recommendations, the authors write.
The takeaway: AI tools can be good for individuals but are less effective at building community
In terms of the research’s implications for firms, the authors suggest using generative AI for its personalization capabilities rather than for digital community building or large-scale ecosystem development.
Although the research took place in China, Tucker said the results are generalizable.
“I think what our research shows is that in the generative AI age, technology can certainly assist with information retrieval, reminders, prompts, all those good things, but we can’t really use it to replace that sense of community,” Tucker said.
Catherine Tucker is a professor of marketing at MIT Sloan, faculty director of the school’s Executive MBA program, and a co-founder of the MIT Cryptoeconomics Lab. She studies the interface between marketing, the economics of technology, and law. Tucker has particular expertise in online advertising, digital health, social media, and electronic privacy. Her recent work explores the pros and cons of using blockchain for marketing and ways that social platforms can protect children from hateful speech.
Linyi Li is an assistant professor of marketing at Singapore Management University’s Lee Kong Chian School of Business. His research focuses on e-commerce and marketplace analytics, marketing strategy, and retail operations and strategy.
