Intelligent Decision Support and Meaningful Human Control in Multi-Domain Operations

Future military operations take place across multiple domains (land, sea, air, cyber, space) and include multiple layers in time and distance. Planning involves making sense of information across different domains and developing creative solutions that produce converging effects in various dimensions (e.g. physical, cognitive, virtual). Multi-domain operations (MDO) require large and heterogeneous streams of data to be monitored, appreciated, and managed to effectively plan an operation. Due to this, decision makers are at risk of being overloaded with information. Many emphasize the use of artificial intelligence (AI) to support the planning of MDOs. AI-based systems may offer a wide variety of support to the decision maker, like collecting and analyzing evidence; detecting familiar patterns in the data; suggesting courses of action (COA); comparing different COAs by performing wargaming in the background; and signalling potential cognitive bias. The prevailing view is that AI may be of help, but that its deployment is not without risks. AI may potentially harm a human’s control over the decision making process in several ways, e.g., AI presents improper analysis due to lack of relevant training data; human fails to comprehend the AI’s analysis and recommendations; and AI induces automation bias and tunnel vision in the human. Careful consideration is therefore necessary when integrating AI-support into the process of MDO-planning. It is essential that the human maintains meaningful human control over decision making, and that AI-assisted decision-making achieves complementary performance. The design guidelines discussed in this paper emphasize critical capabilities of both human and AI, and underline that human and AI should be supported to jointly and iteratively refine their understanding of the situation and of each other.