Agencies sharing AI use cases after December executive order

The executive order on trustworthy artificial intelligence issued by President Trump in December has encouraged agencies like the Department of Veterans Affairs to share best practices.

The VA’s National AI Institute is working with other VA components, like the Data Governance Council and Veteran Engagement Board, as well as outside agencies to create an AI use case catalog, said Gil Alterovitz, director of AI at VA.

The trustworthy AI executive order set the process in motion by requiring the retirement of AI applications that didn’t meet a set of minimum standards and setting deadlines for inventorying and sharing agency use cases.

“It’s really enabled agencies to learn from each other,” Alterovitz said during the SNG Live: Enhancing AI in Government event presented by FedScoop. “In interacting with other agencies through different councils we’ve been able to learn about and share different AI use cases.”

VA is further looking to pilot a set of modules that can be added to an internal review board on AI, Alterovitz said.

Before researchers build AI models, they’ll go through a voluntary checklist for planning purposes. The checklist builds on the work of VA’s National Center for Ethics in Health Care and the Food and Drug Administration and will encourage safeguarding research participants and veterans’ data, as well as training data to eliminate bias.

VA developed an initial AI module to assist its hundreds of medical centers nationally with COVID-19 individual risk prediction by analyzing morbidity and mortality data over time. Explainable AI was leveraged to help patients understand their risk of illness.

New, post-hospitalization data is being fed into those statistical models for additional insights to inform treatment decisions at a dozen rural and urban pilot sites, Alterovitz said.