U.S. Air Force: Converting Data to Actionable Insight

The United States Air Force is in the midst of a fundamental transition in the way its teams are organized—a shift made possible by the proliferation of geospatial intelligence data. Instead of assembling teams around specific data sources, a new “environment of abundance” of geospatial insight has led to “a problem-centric, sensor-agnostic approach” that relies on cross-functional teams, according to Brigadier General Max Pearson, director of intelligence, surveillance, and reconnaissance operations, within the Air Force’s Office of the Deputy Chief of Staff for Intelligence, Surveillance, Reconnaissance, and Cyber Effects Operations.

Brigadier General Pearson’s remarks came during the GEOINT 2022 symposium as part of a panel discussion of the Air Force’s use of geospatial intelligence. Yet the firehose of data also has presented challenges, prompting the Air Force’s leaders to investigate the data formats and specific insights that are most relevant to forces making critical decisions.

That’s the point of the Air Force’s emphasis on object-based production, which combines numerous sources of intelligence into a single “object” from which an analyst can derive actionable insight without being overwhelmed by a deluge of data.

“We’ve done the exercise where we’ve taken every sensor collected and pushed it through the system. Our battle managers, our [command-and-control] element, chokes on it—it’s just too much data,” said Brigadier General Steven Gorski, director of intelligence for the Air Force’s Air Combat Command, at the GEOINT 2022 symposium. “We have to be focused on pushing the right data: What are the threats that can actually impact the mission? Which high-value target presented itself and needs to be actioned immediately?”

The Air Force faces a similar challenge related to data collected over time. It collects such a vast amount of sensor information, for example, that it’s impractical to keep a complete historical record of every data point. Yet some of those past normal-state readings are worth holding onto because they can be used to train artificial-intelligence applications that, in turn, learn to recognize a deviation from the norm. So the challenge becomes recognizing which data is worth storing, and which is only worth collecting in case it provides useful intelligence regarding a real-time event.

“We are going to need [artificial intelligence and machine learning]…to help us go back and look in historical repositories of data to make sense of what just happened,” said Brigadier General Pearson.

That remark resonated with a fellow GEOINT 2022 panelist, Carl Brenner, technical advisor to the Geospatial and Signatures Intelligence Group at the National Air and Space Intelligence Center. Brenner said that historical data “helps us over time to characterize the performance expectations of an adversary weapon system, and then in more real time to be able to identify anomalies to those expectations that then enables the analysts to spend their time super wisely. They can look at the anomalies—those things that fell outside the expected performance envelope of the system—and go after that, instead of going after the thousands of data points that fell within the expectations.”

The notion that geospatial data is more useful when translated in to an intuitive model became a theme of the Air Force panel discussion. Brenner said geospatial intelligence is often more useful to Air Force decision-makers when presented as part of an interactive model, rather than the traditional format: a series of flat images with text overlays.

Air Force tacticians “want something they can load into their own trainer, into their own mission-planning system…and then run a simulation a thousand times, make changes and run a thousand more,” Brenner said.