Game Interpretation

Lead Research Organisation: Queen Mary, University of London
Department Name: Sch of Electronic Eng & Computer Science


Academia is abuzz with incredible breakthroughs, but most of them don't directly apply to game developers on the ground. I'm building a bridge between the gritty realities of game development and the theoretical work of brilliant scientists by developing AI that Interprets real world games into formats that theoretical techniques need - like a language interpreter - so that academics and developers don't need to worry about the technicalities of integrating new technologies into their projects.

To achieve this I'm using techniques, such as intrinsically motivated AI, to create agents that can play and understand games without access to code - so that developers don't need to change their codebase to benefit from decades of rich science, and academics don't need to develop "research games" for their research. In the process I'm developing techniques that capture and quantify metrics that aren't traditionally measured quantitatively, or are little understood.

The tools and techniques developed in this research benefit several groups.

Indie Developers, small teams with small budgets. This group has very tight financial constraints which can mean the difference between success and failure. These developers can't afford to hire dedicated data scientists and can benefit from knowledge exchange and tools.
- Quantifying balance efficacy - measuring the effectiveness of patches - can help them to know if design choices are sound earlier.
- Developed tools will reduce their need to hire external play testers by providing AI play testers that provide useful feedback for game designers. This will decrease costs.
Large AAA Developers with high budgets, successful franchises and sophisticated projects. This includes esports which have huge player bases.
- AI play testers can help these companies test large scale changes that would otherwise be dangerous to attempt, and provide confidence in small scale changes before testing on real players.
- Ordinarily, these games can't afford to risk technical investment in bleeding edge machine learning techniques, but may significantly benefit from them due to their large data sets. Tooling that reduces the difficulty of applying machine learning techniques to their data sets can reveal deeper insight into their player bases and potentially improve key metrics.
Game Engine Developers, such as Godot, Unity and Unreal Engine.
- The integration of AI play testers as a built-in tool at the engine level can be a selling point for the game engine, increasing popularity.
- Enhancement of other game engine features due to the integration of new machine learning techniques.
Academics in the Theoretical Stratosphere. Researchers working in Machine Learning and Games Intelligence. This group has a wealth of knowledge that can improve the games industry but may not be in the best position to implement this knowledge due to the aforementioned challenges.
- Allowing researchers to more easily practically demonstrate the impact of their work, improving the confidence of investors and securing funding.
- Reduced friction of integration between academics and game developers can mean improved access to large data sets useful for further machine learning research.


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/S022325/1 01/10/2019 31/03/2028
2275802 Studentship EP/S022325/1 01/10/2019 31/01/2024 Charles Gbadamosi