What characteristics make a person trustworthy? Under what circumstances would a person delegate life or death decisions to artificial intelligence (AI)? Does it matter that AI systems reflect trustworthy humans’ decision-making preferences, morals, and ethics? If so, what characteristics are most important?

These are some of the fundamental questions DARPA researchers are exploring for the In the Moment (ITM) program, which aims to support the development of algorithms that are trusted to independently make decisions in difficult domains, particularly in significant trauma events such as battlefield triage.

DARPA’s research has identified the need for fundamentally different approaches to advance AI technology to a place where we’re willing to trust it and not be foolish to do so. Continuing themes from our mini-series on ELSI – ethical, legal, and societal implications of new technologies and capabilities – we meet with DARPA’s ITM program manager, Dr. Patrick Shafto, and the ITM performers and ELSI advisors, who break down how they’re tackling the fundamental question of alignment in the context of human decision-makers and autonomous decision-making tools.

In case you missed them, check out our previous ELSI series episodes at the following links:

Episode 79: Integrating ELSI

Episode 78: Introducing ELSI

Our special thanks to the following ITM performers and advisors for their contributions to this episode (in order of their appearance):

·        Alice Leung, RTX BBN

·        Joseph Cohn, SoarTech

·        Matthew Molineaux, Parallax Advanced Research

·        Arslan Basharat, Kitware Inc.

·        Jennifer McVay, CACI

·        Dave Cotting, Institute for Defense Analyses (IDA)

·        Sarah Daly, IDA

·        Lauren Diaz, University of Maryland Applied Research Laboratory for Intelligence and Security (ARLIS)

·        Ellie Tyler, ARLIS