1 of 16

Email: mohamed.abdelazez@forces.gc.ca

Defence Research Development Canada – Centre for Operational Research and Analysis

Decision Support for Commanders: An AI Approach to Wargaming

Dr. Mohamed Abdelazez

Defence Scientist

Defence Research Development Canada

Centre of Operational Research and Analysis

2 of 16

Introduction – Wargaming

“Analytic games that simulate aspects of warfare at the tactical, operational, or strategic level”
An example of wargaming being leveraged at the strategic end of the spectrum was US General Donn Starry’s application of wargaming to develop what would become AirLand Battle – the principal US warfighting doctrine of the latter stages of the Cold War
“A peculiar result of not properly wargaming force structures has been the sub-optimization of force structures to combat fixed assumed enemy COAs. By not properly wargaming such force structures, the proponents of the structural changes succeed in beating their favourite enemy COAs but fail at challenging their own ideas.”

"Wargaming," [Online]. Available: https://www.rand.org/topics/wargaming.html

P. Norwood and B. Jensen, "Wargaming the Third Offset Strategy," Joint Forces Quarterly, 1 October 2016.

P.-I. Evensen, S. E. Martinussen, M. Halsør and D. H. Bentsen, "Wargaming Evolved: Methodology and Best Practices for Simulation-Supported Wargaming," in Interservice/Industry Training, Simulation, and Education Conference (I/ITSEC), Orlando, FL, 2019.

Image copied from: https://www.usni.org/magazines/proceedings/2020/january/more-war-game

2

3 of 16

Introduction – Crowdsourcing

Developing new COAs is a creative process that benefits from involving many participants with varying backgrounds, as realized by the Office of Naval Research (ONR)
This form of wargame relies on crowdsourcing, defined as ‘harnessing the diversity of perspectives in large populations to enhance decision making.’
A successor to crowdsourcing is AI

https://freerangestock.com/photos/136854/silhouettes-of-business-people.html

W. D. Jr., "Crowdsourced collaboration: ONR seeks players for futuristic online wargame," Military News, 16 March 2017. [Online]. Image copied from: https://www.usni.org/magazines/proceedings/2020/january/more-war-game

P. Norwood and B. Jensen, "Wargaming the Third Offset Strategy," Joint Forces Quarterly, 1 October 2016.

3

4 of 16

Introduction – Artificial Intelligence

Image by: Seema Singh (Cousins of Artificial Intelligence | by Seema Singh | Towards Data Science)

4

5 of 16

AI in Wargaming – Potential Benefits

Developing scenarios for a wargame could be a time-consuming, complicated task that requires ingesting information on the current operational environment and developing scenarios to simulate future/unseen warfare or ideation of innovative COAs
LLMs have shown their ability to generate text mimicking human writers
The text of the scenario could be supported with videos, images, and voice and sensor data that are fully AI-generated to provide an immersive environment for wargamers

GENAI ANNOUNCES MOU TO SELL 85 MILLION AI COMPUTE HOURS OVER THE NEXT FIVE YEARS (yahoo.com)

5

6 of 16

AI in Wargaming – Potential Benefits

Acting as the blue or red teams requires in-depth knowledge of the forces and ideas they represent
The gathering of knowledge and trend analysis could be extended to form agents representing the forces and ideas competing within a game, also known as agent-based modelling
Unlike humans, AI produces repeatable actions under the same circumstances, so using agents in wargames could improve repeatability and action verification through future testing

Artificial intelligence - Free technology icons (flaticon.com)

6

7 of 16

AI in Wargaming – Potential Benefits

AI agents can be used to provide decision support through the exploration of known and unknown COAs
Lee Sedol (Go’s world championship) reported on AlphaGo, ' Its style was different, and it was such an unusual experience that it took time for me to adjust.’
AI has the potential to iterate over thousands of potential COAs, analyze their outcomes, and recommend the most feasible ones

Reinforcement Learning, Part 1: A Brief Introduction | by dan lee | AI³ | Theory, Practice, Business | Medium

E. Willingham, "AI’s Victories in Go Inspire Better Human Game Playing," Scientific American, 13 March 2023

7

8 of 16

AI in Wargaming – Potential Challenges

Unlawful COAs could be a challenge of AIs used in wargames
As AI can generate thousands of COAs, sifting through these COAs could slow down the wargame instead of speeding it up
Large-size models are difficult to interpret, which reduces their explainability, traceability, and reliability
A biased AI may prefer one action over the other regardless of the impact of that action

Bad AI Comes From Bad Data | HuffPost Contributor

8

9 of 16

Current AI in Wargaming – NATO

Task Group SAS-139
NATO Wargaming Community of Interest – the NATO Audacious Wargaming initiative
US / UK / Australia team paper titled, Opposed Artificial Intelligence: Developing Robustness to Adversarial Attacks in Attacker-Defender Games via AI-based Strategic Game-Playing

NATO - Wikipedia

9

NATO Analytical War Gaming – Innovative Approaches for Data Capture, Analysis and Exploitation, documents the findings of the Task Group SAS-139 (System Analysis and Studies) regarding best practices and innovative approaches for wargaming, including the NATO Modelling and Simulation Group (NMSG)
The group also established a NATO Wargaming Community of Interest – the NATO Audacious Wargaming initiative.
A significant force multiplier for NATO’s research and development activities is the multi-national nature of the alliance itself. In the context of this paper, an example of this is a 2021 paper written by a US / UK / Australia team titled, Opposed Artificial Intelligence: Developing Robustness to Adversarial Attacks in Attacker-Defender Games via AI-based Strategic Game-Playing. The paper leverages the research capacities of the UK’s Defence Science and Technology Laboratory (Dstl), Australia’s Defence Science and Technology Group (DST) and the U.S. Naval Research Laboratory and Military Academy (West Point). The paper's premise is that, as AI has emerged as the significant enabling technology behind automated systems used in military applications, computerized systems continuously interact with other entities in their environment, including humans, smart devices, computers and other AI.

10 of 16

Current AI in Wargaming – US

DARPA – Gamebreaker, Artificial Social Intelligence for Successful Teams (ASIST) program, Constructive Machine-learning Battles with Adversary Tactics (COMBAT)
Naval Postgraduate School – Game Theory and Prescriptive Analytics for Naval Wargaming Battle Management Aids
Office of Naval Research Global (ONRG) Experimentation and Analysis (E&A) – Tactical Artificial Intelligence (AI) for the Marine Corps
Air Force Research Lab – Stratagem Wargame

Flag of the United States - Wikipedia

10

“In future conflicts, DoD investment is designed to maximize imbalance to create an advantage or to seek equilibrium when an adversary seeks an advantage. New AI Algorithms inspired by Gamebreaker could help develop winning warfighting strategies when the adversary’s objectives – i.e., the “rules of the game” – are not clearly known. By exploiting game balance, Gamebreaker addresses a gap in AI and data analytics research as applied to current wargaming and simulation.”
ASIST researching an AI that models the world and generates problem-solving scenarios the same way humans do, has proven beneficial for further research into AI-enabled wargaming in work underway at the Naval Postgraduate School
(COMBAT) project identified an interest in industry developing AI algorithms to generate unexpected behaviours by synthetic enemies
Game Theory and Prescriptive Analytics for Naval Wargaming Battle Management Aids, provides a solid example of the ongoing wargaming research underway, with concrete applicability to naval warfighters, ‘The research approach revealed three different applications for automated wargaming: (1) to support table-top wargames as an automated white cell for adjudication or as a red team cognitive agent, (2) to support operational mission planners as a non-real-time course of action (COA) engine, and (3) to support the tactical warfighter as a real-time COA engine that considers second, third, and nth order effects as it evaluates and recommends possible tactical COAs.’
A pertinent example of ONR AI-enabled wargaming is the theme of Tactical Artificial Intelligence (AI) for the Marine Corps, which addresses advanced research into modelling and simulation, emphasizing developing tactical decision support for Marine Corps small units.
By the conclusion of the Stratagem program, the objective is "to have an AI capability that can compete at, or above, the level of capability of seasoned veteran planners in realistic wargame scenarios (that have experience using the wargame tool)."

11 of 16

Current AI in Wargaming – US

United States Marine Corps (USMC) – leveraging artificial Intelligence to learn, optimize, and win (LAILOW)
Army Research Laboratory (ARL) – Games and Simulators as a Platform for Developing Artificial Intelligence for Command and Control

Flag of the United States - Wikipedia

11

12 of 16

Current AI in Wargaming – UK

Defence Science and Technology Laboratory (Dstl) – Intelligent Wargaming Approach to Increase Course of Action Effectiveness in Military Operations
Defence Wargaming Centre

Union Jack - Wikipedia

12

13 of 16

Current AI in Wargaming – Australia

Defence Science and Technology Group (DST) – Modelling Complex Warfighting (MCW) Research Agenda

Flag of Australia - Wikipedia

13

14 of 16

Current AI in Wargaming – China

China Institute of Command and Control – Artificial Intelligence and War-gaming Forum
Prophet 1.0
Intelligentization

Flag of the People's Republic of China - Flag of China - Wikipedia

14

Kania has reported on the China Institute of Command and Control cosponsored Artificial Intelligence and War-gaming Forum, held at the National Defense University’s Joint Operations Academy in September 2017.
A Chinese Academy of Sciences Institute of Automation-developed AI system called Prophet 1.0 debuted at this exercise - competing against human teams, which were defeated “seven to one”. This exercise provided useful data on what vulnerabilities in human decision-making were exploited by AI.
This strategic initiative includes the development of AI for military applications and leveraging wargaming platforms to advance technological experimentation. The progress to date includes PLA contests and competitions that have concentrated on developing AI systems for wargaming in complex scenarios

15 of 16

Conclusion

An important function of C2 is making the right decision at the appropriate time
AI is a powerful tool with multiple capabilities, including generating text, sound, and video, ingesting and analyzing large amounts of data, and producing COAs that are novel to humans
There is the potential for AI to improve wargaming efforts by speeding game setup, improving the objectivity and repeatability of the games, producing better models of elements in the games, and suggesting innovative COAs to the players

Hyperparameter Tuning using MLOps (linkedin.com)

Free Shield SVG, PNG Icon, Symbol. Download Image. (iconpacks.net)

15

An important function of C2 is making the right decision at the appropriate time, which can be supported by having validated COAs available for the commander. Wargaming could be used to develop and validate COAs, but it could be slow to develop, subjective, and unrepeatable. AI is a powerful tool with multiple capabilities, including generating text, sound, and video, ingesting and analyzing large amounts of data, and producing COAs that are novel to humans. There is the potential for AI to improve wargaming efforts by speeding game setup, improving the objectivity and repeatability of the games, producing better models of elements in the games, and suggesting innovative COAs to the players. Multiple nations are already researching AI for wargaming, and it is possible that soon AI produced COAs will be used in operations. However, there are challenges to adopting AI such as the generation of unlawful COAs, the lack of explainability, traceability, and reliability, and the potential propagation of bias. These challenges must be addressed or communicated to wargamers before the potential of AI can be realized.

16 of 16

Email: mohamed.abdelazez@forces.gc.ca

Defence Research Development Canada – Centre for Operational Research and Analysis

Decision Support for Commanders: An AI Approach to Wargaming

Dr. Mohamed Abdelazez

Defence Scientist

Defence Research Development Canada

Centre of Operational Research and Analysis