The application of artificial intelligence to solving battlefield challenges runs a risk of making it harder on commanders, warns Peter Schwartz, the lead enterprise systems engineer for the MITRE Corp.

Using machines and algorithms to help feed detailed information on threats and options for action could be a significant new capability for battlefield commanders by quickly doing things faster and more accurately that could be done without artificial intelligence, Schwartz said during a panel discussion at the Army Autonomy and Artificial Intelligence Symposium and Exposition in Detroit hosted by the Association of the U.S. Army’s Institute of Land Warfare.

He cautioned that battlefield commanders already face a significant cognitive burden with the information at hand. Feeding them more information runs the risk of overloading them, especially if they don’t understand the information.

“We need to be careful we don’t create new cognitive burdens,” because using information created through artificial intelligence “can be harder for an end user,” he said, describing the possibility of getting confusing and unexpected results.

This is not an insurmountable problem, he said, but something that needs to be considered as the Army moves forward with autonomous systems, machine learning and other new technology. The key to getting this right is to not test a machine or system in isolation but to do a larger evaluation that looks not just at what information is provided but at how it might be used, Schwartz said.