Explainable Distributed Constraint Optimization

The distributed constraint optimization problem (DCOP) model is an abstract way of modeling problems where multiple entities (humans, robots, computer pro- grams) need to coordinate together to perform some task. Examples include disaster rescue scenarios, where various first r esponders need toc oordinate with e ach other to save lives. However, the existing model and techniques are unsuitable to scenarios in which humans are required to participate in the process. In this research, we are proposing a new framework that will generate explanations to decisions and actions performed by the agents during the process of seeking for a solution, which will al- low humans to participate and contribute to the solving process and in executing its outcome. In order for this framework to be useful it must respect the privacy of the humans involved, therefore we will require that the framework produces explanations that justify decisions made without revealing the identity of the humans involved. Thus, through the proposed research, we will build the foundations for a general ex- plainable DCOP framework, which will spur deployment of DCOP algorithms iii the real world.