Skip to content

DARPA Launches AI-Based Decision-Making Program

What’s happening?

The Defense Advanced Research Projects Agency (DARPA) is launching a new program to support AI-based decisions for Department of Defense (DoD) operations.

The program, called In the Moment (ITM), aims to research and develop algorithmic decision-making capabilities for difficult domains. DARPA defines these domains as situations where there is no easy right answer and human decision-makers are constrained by time, uncertainty, rapid developments, and conflicting values. This includes situations like combat, disaster relief, and medical triage.

According to the program’s manager, Matt Turek, the goal of ITM is to build “rigorous, quantifiable, and scalable approaches to evaluating and building algorithmic systems for difficult decision-making where objective ground truth is unavailable.”

This approach differs from other AI developments in that it relies less on human analysis and agreement to make decisions.

The program will occur over 3.5 years with support from private corporations, and covers four technical areas:

  1. Identify and quantify key decision-maker attributes for use in difficult domains.
  2. Create a scoring system to evaluate human versus algorithm-based decision-making.
  3. Design and execute program evaluation.
  4. Integrate legal, moral, and ethical expertise into the program and develop program policies and processes.

ITM will also focus on decision-making first for triage in small unit injuries (like targeted combat fire) and later for mass casualty events (like public bombings).

What’s the impact?

Such AI capabilities can be highly valuable for operations where even experts may struggle to make timely decisions and triage effectively. 

According to Turek, algorithms aim to mimic highly-trusted human expertise while incorporating vast data sets for more accurate decision-making. For example, data surrounding supply inventory, staff volumes, and drug availability would support algorithms responding to medical triage.

In a Washington Post article, Sohrab Dalal—colonel and head of the medical branch for NATO’s Supreme Allied Command Transformation—says current medical triaging is based on a 200-year-old process of one-on-one patient evaluation. Algorithms could provide a much-needed refresh to this practice, expedite decision-making in urgent situations, and potentially save human life.

However, the program is raising red flags for ethicists, who are concerned about the potential for algorithms to generate biased decisions. There’s also the possibility of relying on algorithms even when they contradict common-sense scenarios, and assigning responsibility for negative outcomes related to algorithms.

How Can OSINT Help?

Decision-making algorithms, particularly those for small unit injuries and mass casualty events, will likely rely on proprietary data like medical information. There’s also potential for this emerging technology to benefit from open-source information provided by social media, news, and other publicly-available outlets.

For example, military operations can benefit from open-source social media content, such as:

  • Posts documenting military operations from specific locations
  • Picture or video evidence of casualties or infrastructure damage
  • Real-time updates for natural disasters, combat, bombings, and other critical events

For difficult domains as defined by DARPA, open-source data complements classified information sources and can be used to develop decision-making algorithms. 

Echosec Systems has developed machine learning capabilities using open-source web data and offers an API for users developing advanced data applications like AI. Book a demo to learn more about OSINT and its role in emerging AI technologies.

Subscribe to get our weekly bulletins delivered straight to your inbox