Assessing the Quality of Decision-making by Autonomous Systems

Test and evaluation of autonomous systems will depend on the ability to assess the quality of their decision-making capability, both in interpreting their environments and in selecting courses of action. IDA argues that observed system performance will not be sufficient to evaluate autonomous decision-making in the ways necessary for successful deployment, especially for systems designed to team with humans. Instead, novel instrumentation approaches will be required to support diagnosis and assessment of the algorithms, training data and operational concepts that support teaming.