The Army is using computer analysis that it views as artificial intelligence to wade through the files of enlisted soldiers seeking promotions, checking for eligibility and other basic scores.
Maj. Gen. Hope Rampy, commanding general of Army Human Resource Command ,said enlisted evaluation boards have to review thousands of records, some of which include soldiers that are “not even competitive for promotion, but we still have to give them an order of merit list score.”
But the Army hopes the algorithms it’s developing can screen for soldiers who meet prerequisites for promotion, like certain schooling and job history, so time isn’t wasted on a soldier who is not going to be eligible for promotion.
According to Col. Tom Malejko, chief talent analytics officer at Army Human Resources Command, AI will not replace humans and they will be part of each step to “override whatever the computer’s decision may have been.” Malejko discussed the initiative at the annual Association of the U.S. Army’s annual conference in Washington D.C. this week
The Army has developed its own algorithm that examines soldiers’ records and automatically scores evaluation reports but Malejko said that they’ve also added controls to “mitigate any potential bias” that the AI might have. These checks include making sure that racial or ethnic nature of a soldier’s name isn’t being considered for scoring, as well as not considering individual branches or ranks.
“Some of those things may correlate with different scores,” Malejko said.
Top Stories This Week
The AI screens, he said, are meant to make quicker work of cutting down the field of candidates early in the process. “Can we screen out individuals that are not really competitive for the process upfront, and then help our board members to focus their valuable time and resources on those individuals that are then most competitive for that selection?” Malejko said. “How can we kind of focus our attention on those areas where really only humans can make a good judgment call, and then using computers to screen out based on more discrete criterion.”
Malejko said the Army is beginning to use AI with non-commissioned officer boards as a pilot and then take those lessons and apply it to officer boards. However, he added that they first need additional authorities from Congress before applying it to officer selections.
Rampy also said that the Army has used algorithms for the last four years to screen records and invite them for a command board. Because algorithms are trained on new data, Rampy said, it improved over time.
One year the algorithm excluded military intelligence soldiers from a board invite “because it was the wording that was in those evaluation reports for certain branches of officers,” she said. So the humans involved had to invite more than 30 officers that were missed by the algorithm. Last year, humans had to add a handful officers that were excluded by the original AI selection.
