Better Understanding of Humans for Cooperative AI through Clustering
The aim of this research project is to explore new ways in which artificial intelligence can work with humans to accomplish tasks using games as a testbed.
Though there has been substantial progress for the optimising of performance for pure common-interest problems, there has been little for the advancement of cooperative AI. With AI systems becoming more complex and larger in scope, problems relating to their ability to collaborate with other models, systems and humans become more distinct as well as necessary to address.
This project focuses on addressing this through developing AI that are more capable of understanding humans through constructing an artificial representation of them through play-data and clustering. As a result, the AI should be better able to cooperate with users and their unique behaviours.