RSS feed Add to your Facebook page Watch us on Youtube

Activity title

Human in the loop considerations for Artificial Intelligence

Activity Reference

IST-157 (AI2S)

Panel

Information Systems Technology

Security Classification

NATO UNCLASSIFIED

Status

Active

Activity type

RTG

Start date

2017

End date

2020

Keywords

architecture, Artificial Intelligence, Autonomy, Command and Control, operations, Standard

Background

Work conducted by the Multinational Capability Development Campaign has concluded that “The prevalence of systems employing… [Artificial Intelligence (AI)]… is growing in the military sphere; such systems will likely become a permanent feature of military operations. The implications for military operations and capabilities are broad and significant. The role of [AI] in military systems is one of the most important considerations for defence policy makers in the near future. An important consideration for the design and operation of systems with autonomous capability is the level of human control in the system.” In the military context, it is important that systems that employ AI remain under appropriate human control. The implementation of AI will differ between Nations. In a coalition environment, such systems potentially deploy in parallel during an operation, which requires that NATO commanders understand the subsequent effect on planning and C2. Creating a common standard for describing the role of the human operator and the role of the machine in systems that use AI will help commanders incorporate such systems in their planning processes

Objectives

a. Identify the doctrinal implications for the use of AI in coalition environments; b. Identify potential C2 arrangements (oversight, decision rights, evaluability of information, etc.); c. Develop a taxonomy for describing human in the loop architectures for systems that employ AI

Topics

Decision rights allocation – which decisions are potentially made by AI; Does the use of AI in information systems create any special considerations in a military environment ? What do commanders need to know when AI systems are part of operational planning ? How can human in the loop architectures be described so that a lay person clearly understands? Which key decisions are being made by the person and which are made by the machine ? How future Commanders should be educated and trained to make a conscious and effective use of these technologies?

Contact Panel Office