Dr. Aparna Hegde founded India-based non-profit ARMMAN in 2008 after working in an overcrowded Mumbai maternity clinic where she saw new mothers and their babies die of avoidable complications. Her solution is a “tech plus touch model” which provides targeted preventive care information to enrolled women through free voice calls.
But as they scaled to millions of participants, 40% of women stopped engaging midway. “Studies have shown that if women get information at the right time, there is improvement in health outcomes for themselves and their children,” says Hegde. ARMMAN trained health workers to reach out to women that dropped out of the program, but needed help identifying which high risk women to prioritize, so they approached Google for an AI solution. A team led by Google researcher Aparna Taneja trained an AI model on calls from ARMMAN’s databases to learn which women are likely to drop out and would benefit most from personal intervention. The model, tested with a pilot group of around 100,000 women, improved retention rate by about 30%, and increased how likely expectant mothers were to be proactive about following health guidelines—which is correlated with better health outcomes. “It’s the most rewarding experience to do such cutting edge AI research with meaningful social impact,” Taneja says. Plus, she says, “the model is applicable to similar problems, both in the context of health and other domains, because the ideology is to solve resource allocation under budget constraints.”