USDA’s Measured Approach to Responsible Advanced Analytics


The food and nutrition service focuses on data before scaling AI and ML capabilities.

Through its 15 nutrition assistance programs, the Food and Nutrition Service (FNS) manages America’s nutrition safety net and serves one in four Americans a year – and it collects lots of data. Although FNS has a robust data infrastructure in place, it takes a measured approach to deploying responsible AI instead of forging ahead to apply advanced analytics to all of its datasets.

The SNSF’s Enterprise Data Analytics Platform (EDAP) suite includes data lake portal and dashboard functionality. So far, these capabilities are being rolled out cautiously across the service.

“We’ve really laid the groundwork – at least in terms of technology – we have the ability to use artificial intelligence and machine learning,” Chris Rottler, SNSF’s deputy chief data officer, said at the summit on Tuesday. ATARC on AI. “The question we’re really grappling with now is when and where does it make sense to employ these methods when we have this data, and how do we do it responsibly?”

For Rottler, that means focusing on data.

“It’s all about the data, understanding the context, and exploring the structures that generate the data,” Rottler said. “Inequality is insidious and not always easy to uncover in the data and questions we ask. It is incumbent upon us to ensure that the fuel that powers the models we develop first, does no harm and, second, does not exacerbate inequality.

By bringing together data and policy experts, the service takes a collaborative approach to understanding and improving its investigative practices.

“Twinning our data experts with our policy experts [the Supplemental Nutrition Assistance Program (SNAP)] has ensured that we are really asking the right questions and are able to draw accurate conclusions,” said FNS Senior Technical Advisor Christine Daffan.

FNS envisions data analytics to support not only its external programs, but also its internal operations; including staff diversity, equity and inclusion.

“We have created a barrier analysis dashboard that looks at the landscape of our people – who do we have on board to do this work, provide these services – and where there are opportunities for the agency to increase our diversity, as well as accessibility,” said Tameka Owens, SNSF Assistance Administrator.

Rottler said that while the FNS is just dipping its toes into advanced analytics, he’s excited about all the groundwork they’ve done to prepare the next steps for Responsible AI.

“I know I sound very cautious, but I’m really excited for FNS and USDA as a whole,” Rottler said. “We’re in a good position to start using machine learning more responsibly. And we’ve piloted its use in a very controlled and specific use case. … We’re taking small steps, but I’m excited to see how we can really leverage the investments we’ve made in infrastructure.


Comments are closed.