The Pentagon has finished drafting a new data management strategy, and is likely to release it publicly within the next 30 days, the Defense Department’s chief information officer said Wednesday.
The document will attempt to answer some of the department’s key questions about becoming a “data-centric” organization, including how a more disciplined approach to data will impact the department’s acquisition programs and how to harness the vast amounts of data it already generates to train future artificial intelligence algorithms, Dana Deasy, the DoD CIO said.
“The biggest challenge to getting the flywheel of AI really going across the department is data. And people say to me, ‘the Department of Defense has lots of data.’ Yes, but just having lots of data doesn’t mean it’s accessible,” he told reporters. “How do you secure it the right way? What are the policies, the different classifications? How do you ingest it? … It comes from different sources in different formats. And what that data was originally created for maybe have slightly different purposes. So [the challenge] is curating that data in a way that it’s then usable for the algorithms to learn from.”
The development of the strategy was the first major project for David Spirk, the new chief data officer DoD appointed in June. Spirk conducted a “listening tour” across various DoD components to ensure the CIO’s office was asking and answering the right questions about data management.
“The things everybody constantly talked about — and that we needed to build our goals around — were, How do you ensure all data is visible? How do you make sure it’s accessible to everybody at the right classifications and authority levels, and how do you make sure it’s more understandable? How do you make sure that data is better linked? How do you make sure data is trustworthy?” Deasy said. “How do you make sure the data you’re working with is the right source? And then finally, how do you make sure that data is interoperable and secure?”
The forthcoming document will also mark the first time the department has had a full data maturity model that involves every level of the department in a single approach to data governance, Spirk said in a September interview with Federal News Network.
“What’s going to emerge are a lot of efficiencies. Just having a data catalog at the department level is going to be a huge boon for the department,” he said. “When we start an AI project, we can map to a dataset that’s already been created and map to a set of labels associated with that. That way, we don’t have to start the same data project over every time – we can begin using some of those datasets that we’ve made visible and brokerable.
The data strategy is emerging at the same time the Joint Chiefs of Staff are restructuring their approach to thinking about data requirements for weapons systems.
Gen. John Hyten, the vice chairman of the Joint Chiefs, said last month that the Joint Requirements Oversight Council (JROC), which he chairs, will soon begin promulgating joint standards for data interoperability. The new approach will insist that the military services comply with the standards from the beginning of a new system’s development, rather than letting them develop platforms first and send them to the JROC for validation of what they’ve already done.
“They’re not going to be the traditional requirements and capability description documents and capability production documents that you’ve looked at for years,” Hyten said. “They’re going to be capability attributes that programs have to have. If you don’t meet those, you don’t meet the job requirements, and therefore you don’t get through the gate and you don’t get money.”