Skip to main content
Pink stopwatch icon on a yellow background

Elevator Pitch: Project S.A.R.A.H.

The World Health Organization’s AI assistant aims to answer people's questions on different aspects of health.
Written by
Michael F. Fitzgerald
Published
November 21, 2024
Read Time
1 min

Project S.A.R.A.H. (Smart AI Resource Assistant for Health) is the World Health Organization’s assistant for people with questions about different aspects of health. Sarah (the AI’s pronouns are she/her) can discuss in eight languages topics like how to eat nutritiously, how to quit using tobacco and e-cigarettes, and how to reduce stress. She grew out of a pandemic project to promote health ideas during COVID-19. As that emergency waned, the WHO wanted to see if it could apply generative AI to address non-communicable health issues. Right now, Sarah is a prototype, but she could become a routine way for people to interact with the WHO. Harvard Public Health’s Michael F. Fitzgerald spoke with Rüdiger Krech, WHO’s director of health promotion.

What public health purpose does your idea serve?

What we try to do with this is to provide accessible, reliable, and actionable health information to support people online. You can ask her anything about tobacco cessation, mental health, healthy lifestyles, alcohol consumption, or healthy nutrition. And you can engage with her in a discussion, face to face.

Who is funding this project, and do you have access to capital?

It’s a mix between governments, academic institutions, and private institutions. We receive voluntary contributions and have partnerships within [the WHO] and from Amazon Web Services, Google, George Washington University, and the government of Qatar. It’s difficult to say [how much it costs] because many things were in-kind contributions.

How do you get paid for it, and who are your clients or customers?

All our services are free of charge. All our projects start with an idea that we developed with partners. We are almost always dependent on external funding for any project we do. Normally we would say we need X dollars for a project. Here, we have a highly interested group of companies and partners.

Our audiences are very diverse: from policymakers to administrators in different sectors, different industries of the world—officials as well as the general public. [Sarah] scans millions of documents and WHO materials to provide concise and accurate summary answers, saving individuals hours of effort sifting through complex health guidelines.

What are your big obstacles?

The funding situation. We are dependent on external funding. But here we have a highly interested group of companies and partners as well.

Language is a problem. We would like her to speak many more languages and that requires a lot more resources.

On the technical side, we’ve seen so much misinformation and disinformation. We have more to do on training (the AI engine) and adjusting the guardrails. We’d like to continue building on this experience, to add more topics. That depends on where Sarah ends up going.

How do you show your value or impact?

We publish manuals that—for example, on mental health—outline how you actually deal with your depression. We measure, of course, how many times have these reports been downloaded. But we can’t say for certain if they have been read, and by how many people, once downloaded.

There are completely new people searching for information on the major websites and meeting Sarah. Interactions with people, by far, outweigh the downloads of these records. [Editor’s note: In the most recent six-month period, there were 200,000 visits to the Sarah landing page, leading to 70,000 conversations.] It’s much more accessible now for people to ask questions and get some answers.

We also have user surveys. Over 85 percent of respondents are saying that [Sarah] was good or excellent at helping them answer their health queries.

This is a novel way of looking at public health communications and information. We want to be at the forefront [of AI], on the cutting edge. WHO would like to be a global leader in guiding how this technology can support public health. Is Project S.A.R.A.H. something that could be licensed in some fashion to public health organizations around the globe to adapt for their own purposes, using AI with people in their communities? We’re not there yet because we haven’t fully understood many of the challenges that we’re still facing. [But] we’ve got something here, so let’s see where we go.

Filed Under
Tagged
Contributors
Michael Fitzgerald
Michael F. Fitzgerald
Michael F. Fitzgerald is editor in chief of Harvard Public Health. Read more from Michael F. Fitzgerald.

More in Tech & Innovation

See all