Gaming requires players to make increasingly complex sequential decisions.
While AI has been used to play games at or above the level of humans, this has often relied on inordinate investment in heavy hardware, assumes that games are fixed, have only two players and that game states are fully observable. We are interested in much more complex game scenarios involving partial information, complex rules, potentially multiple players and structural game changes driven by events within the game. The goal of this component is the development of AI assistance for supporting human players in more complex game situations, where the role of AI is to help players understand and rapidly test strategies and evaluate their relative strengths and weaknesses.
There are no studies to date that have examined what strategic game players need from AI assistance; the work aims to be the first to obtain a coherent understanding of the role of Autonomous Analyst instances for human-agent teaming for playing complex games, using methods including instantiation of test implementations to evaluate as well as demonstrate the new knowledge.
The first goal of the work is to identify effective automated techniques for assisting humans in strategic game play, and the second is to implement and experimentally evaluate the use of computational techniques for supporting strategic game playing.
The Commonwealth of Australia (represented by the Defence Science and Technology Group) through the Operations Research Network supported this research through a Defence Science Partnerships agreement.
- Melissa Rogerson, Lecturer
- Tim Miller, Professor
- Ronal Singh, Associate Lecturer
- Joshua Newn, Research Assistant in Explainable AI