The future of AI lies in enabling people to collaborate with machines to solve complex problems. Like any efficient collaboration, this requires good communication, trust, clarity, and understanding. Explaining to humans how AI reasons are only a part of the problem, we must then be able to design AI systems that understand and collaborate with humans: Hybrid decision-making systems aim at leveraging the strengths of both human and machine agents to overcome the limitations that arise when either agent operates in isolation.

This lecture provides a reasoned introduction to the work of Explainable AI (XAI) to date and then will focus on paradigms in support of synergistic human-machine interaction and collaboration to improve joint performance in high-stake decision-making. Three distinct paradigms, characterized by a different degree of human agency will be discussed: i) human oversight, with a human expert monitoring AI prediction augmented with explanation; ii) Learning to defer,  in which the machine learning model is given the possibility to abstain from making a prediction when it receives an instance where the risk of making a misprediction is too large; iii) collaborative and interactive learning, in which human and AI engage in communication to integrate their distinct knowledge and facilitate the human’s ability to make informed decisions.

This lecture is joint work with Clara Punzi, Mattia Setzu, and Roberto Pellungrini