February GDHN Monthly Meeting: Human versus AI: Pros and Cons of different strategies for providing health information and counseling
The February meeting of the Global Digital Health Network was hosted by FHI 360 and explored how artificial intelligence (AI) and machine learning (ML) are being incorporated into client-facing digital health applications. We had three presentations—from Jacaranda Health, FHI 360 and Viamo—that illustrated a continuum of possible uses of AI and ML in place of and in addition to humans to provide individuals with important health information, counseling and support. Despite some early technical difficulties, we managed to connect to all three presenters and have time to spare for some great Q&A! Thanks to the approximately 25 folks who participated in person in Washington, DC and North Carolina and the 40+ people who tuned in remotely.
Because the presentations were extremely rich and we didn’t want anyone to miss out, the summary is longer than usual and is therefore included as a separate attachment. Please do take a look at the full notes and the presentation deck (available as a separate resource). To entice you, below are our presenters’ collective considerations for including AI/ML in a client-facing health intervention.
- AI/ML is good for large amounts of data as well as for systems in which questions are asked repeatedly or consistently
- AI/ML can be useful for triaging (as in the case of Jacaranda Health)
- Weigh what you think the AI/ML will do against what a human can do. In situations with large amounts of data coming in (per first bullet), AI/ML may be able to handle larger quantities of data more efficiently than a human. IN situations where information coming in is nuanced or varied (per second bullet), a human may be better at sorting, analyzing, and responding.
- Look to how AI/ML can be built into existing programs (both Jacaranda and Viamo did this)
- Consider the audience/user needs. In some instances, a user may require a degree of empathy that AI/ML cannot provide (consideration in FHI 360’s presentation)
- In sum, both the content and the context are very important when considering whether to include AI/ML