Building successful hybrid human/AI teams

As we increasingly embed Artificial Intelligence (AI) into defence capabilities, we must understand how AI agents and humans can work together.

For example, how can humans work with, and trust, AI agents and develop AI systems that "understand" and respond to changes in human behaviour and performance. A key part of this problem is how to quantify human performance so that an AI algorithm can monitor it, and how to develop AI algorithms that represent this performance. We also need to establish a ‘common ground’ where an AI system can process human performance, and equally human operators can see how the AI has gathered and analysed the information they provide about a particular scenario.

Under the umbrella of CADR-RAS, Prof Anna Ma-Wyatt and her team, work with renowned Prof Lang White (Electrical and Electronic Engineering) and DSTG collaborator Dr Justin Fidock, on a Defence AI Research Network (DAIRNet) project aiming to develop human performance metrics that can be used with an AI agent to establish such common ground. Anna explains “Our project looks at how humans gather and process information considering factors such as training, experience and how they use a given interface. We use geopolitical information analysis as our starting point. In other projects, we work with DSTG and the Australian Army to understand how humans use information systems while walking.

"In our DAIRNet project, we are developing a modelling framework that enables a representation of how the human operator gathers this crucial information and establishing a common way to represent this information to the AI agent. The next stage is to support dialogue between the AI agent and the human to support effective and resilient decision-making by the human."

"It is a truly multi-disciplinary team working on this project including researchers with strong research capabilities in computer science, maths, psychology and behavioural science."

"We have also recruited great postdoctoral fellows, including Dr Heidi Long, Dr Abdul Chowdhury and Dr Kelli Francis-Staite."