The Defense Advanced Research Projects Agency (DARPA), in the US, has launched a new programme to support the research, development, and fielding of algorithmic decision-makers for military operations.
The initiative, called the In the Moment (ITM) programme, aims to develop algorithms that can assume decision-making responsibilities in difficult domains.
These difficult domains include situations where human decision-makers disagree, or when there is uncertainty, resource limitation, or time constraints, among others. Such situations may include medical triage in combat and disaster relief operations.
According to a DARPA statement, ITM will evaluate the algorithms for mission-critical Department of Defense (DoD) operations.
ITM programme manager Matt Turek said: “ITM is different from typical AI development approaches that require human agreement on the right outcomes.
“The lack of a right answer in difficult scenarios prevents us from using conventional AI evaluation techniques, which implicitly requires human agreement to create ground-truth data.”
The ITM programme will have four technical areas.
In the first area, decision-maker characterisation techniques will be developed to identify and quantify important decision-maker attributes in difficult domains.
The second technical area will involve creating a quantitative alignment score between a human decision-maker and an algorithm while the third area will design and execute the programme evaluation.
The fourth and final technical area will encompass policy and practice integration, as well as offering legal, moral, and ethical expertise to the programme.
DARPA has divided the 3.5-year programme into two phases, with a potential third phase that will be used to mature the technology with a transition partner.
DARPA also recently announced another programme to develop a new cognitive science tool focused on mental health.