DAFIA leveraging artificial intelligence to transform inspections Published Oct. 20, 2021 By John Cochran 377th Air Base Wing Public Affairs KIRTLAND AIR FORCE BASE, N.M. -- The Department of the Air Force Inspection Agency at Kirtland AFB, N.M., is using artificial intelligence, in the form of machine learning, to identify emerging issues of concern to Airmen and enable the Air Force to isolate, respond to, and resolve them as early as possible. Based on lessons learned from previous inspections, DAFIA’s Enterprise Support Division petitioned AFWERX, a program office at the Air Force Research Laboratory that connects innovators across government, industry and academia, to solicit vendors who could tackle this problem. Lt. Col. Kent Moore, enterprise data manager at DAFIA’s Analysis & Assessments Division, explained the initial driver propelling the growing effort and outlined the process so far. “A couple of years ago, DAFIA led a short-notice inspection across 10 installations to assess military privatized housing issues identified to the Air Force by Congress. Afterward, DAFIA analysts began questioning why it took Congressional direction for the Air Force to respond to the concerns of Airmen and their families residing in military housing. There were indicators in other studies and surveys that the problem was prominent, but the volume of the information sources made synthesizing the data a challenge. The tool being prototyped for DAFIA is going to do it very quickly. Our team submitted a formal request to AFWERX and gave them some parameters. Three companies were interested, and each received $50,000 in AFWERX Phase I Small Business Innovation funding to develop concepts to address DAFIA’s issue. Versionista LLC produced the most promising results and, in August 2021, received an AFWERX Phase II award of $750,000 under a contract for prototype development and evaluation. If we like it at the end of this 15-month phase, then we’ll explore a Phase III contract to apply the capability within the IG enterprise.” AFWERX, according to its website, “through innovation and collaboration with the nation’s top subject-matter experts, harnesses the power of ingenuity by expanding technology, talent, and transition partnerships for rapid and affordable commercial and military capability. AFWERX transitions agile, affordable, and accelerated capabilities by teaming innovative technology developers with Airmen and Guardian talent. AFVentures, the commercial investment arm of the Air Force, creates simple pathways for commercial innovators and private capital investment to help the Air Force solve problems.” Moore talked about what data DAFIA initially supplied to the contractor, and the somewhat surprising results the agency has received in response. “We gave them several years’ worth of Unit Effectiveness Inspection data, along with the 2019 Military Career Decisions survey data. After analyzing free-form text, they came up with several word clouds. An unexpected word that popped out was ‘lactation.’ It was discovered that a significant quality-of-life factor among female Airmen was a lack of clean, dedicated and accessible facilities for nursing mothers. Versionista used their model – machine-learning – and it picked up on that. It shows you the utility of this tool we’re building. We’re getting more data for Versionista to analyze through the Air Force Survey Office.” He said his team’s goal with the project is to be able to scan official surveys, blogs, websites, social media, traditional media, and other data sources, to capture issues of broad concern to Airmen and get those issues in front of decision-makers as soon as possible, so they can be addressed. Peter Bray, CEO of Versionista, the Oregon-based software company working with DAFIA, said his company is building a tool that leverages artificial intelligence to identify emerging topics of concern. “These issues, such as military privatized housing and others, are hard to detect through conventional means, and can have severe impact on Airmen morale and readiness. Versionista automates machine comprehension of vast freeform text from surveys and other sources. We use AI to surface distinct, emerging topics. Facing tens of thousands of pages of input is beyond the capability of human readers and many current analysis techniques. As an early warning system, Versionista correlates emerging topics with related programs, locations, and other aspects to better understand root causes,” Bray said.