Evocatus Ltd, an independent consultancy specialising in collective and command training in the defence sector, and Mission Decisions, a MILSPEC Artificial Intelligence (AI) and Machine Learning (ML) provider, have made a submission for the DSET Serious Games Showcase Europe.

The joint team developed a set of secure Large Languages Models (LLMs) that are able to participate effectively as role-players in Matrix Wargames. In addition, the team have added advanced analytics that are able to measure sentiment of overall responses from both human and AI players during the game.

This first of type capability will be showcased at DSET. The Serious Games Showcase Europe will be running as part of DSET, Bristol from 11-13 June. The development team will be looking for volunteer participants to play with (and against) their new AI strategist, if you are attending the show.

A Matrix Game is a type of role-playing simulation used for educational and analytical purposes, often within military, political, or crisis simulation contexts. A particular benefit of these exercises is the creation of a narrative that incorporates a range of player perspectives and actions, rather than focusing on a more linear progression centred on a single protagonist or binary competing adversaries.

Because these exercises often require upwards of 8 people to play, facilitate, and adjudicate, they are resource-intensive. Although effective in broadening perspectives, they can also struggle to accurately predict potential actions by adversaries and competitors; even the most expert “red” (enemy) and “green” (non-aligned) players are an imperfect reflection of their real-world models.

The use of AI role-players that can be tuned to different behaviours, and act based on reference documentation and intelligence, means that Wargames can be delivered to many more participants at once, whilst reducing cost and improving realism. These AI role players operate within a secure environment, ensuring any sensitive information is protected.

Ant Sharman, Director of Evocatus said, “Early results are promising, with many potential avenues for further research and development. AI personas can be trained on specific literature and defence doctrine to more accurately reflect allies or adversaries. Advanced analytics and sentiment analysis can also be produced by this system, allowing participants to be assessed and compared.”