Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) have created an AI assistant designed to monitor and improve teamwork among both human and AI agents.
The system, presented at the International Conference on Robotics and Automation (ICRA), uses a theory of mind model to understand how team members think and plan their actions during cooperative tasks. It can intervene when necessary to align beliefs and improve collaboration.
The AI assistant could potentially enhance teamwork in a variety of different settings.
1. Search-and-rescue missions, by efficiently coordinating search parties
2. Medical procedures, by ensuring clear task organisation during surgeries
3. Strategy video games, by alerting players to misinterpreted tasks
Brian C. Williams, MIT professor of aeronautics and astronautics and senior author of the study, highlighted the inspiration behind the project: "Good partners are well in tune with the beliefs and goals of each other, and our work on epistemic planning strives to capture this style of reasoning."
The system incorporates probabilistic reasoning with recursive mental modelling, allowing it to make risk-bounded decisions. While currently relying on a given set of possible beliefs, the researchers aim to apply machine learning techniques to generate new hypotheses in real-time.