A Theory of Proactivity: a Step Toward Human-Centered AI

Most of today´s AI systems and robots are merely capable of reactive behavior: responding to explicit user requests or external events.

With the increasing maturity and pervasiveness of AI, there is now a growing interest in making those systems proactive, that is, able to act in an anticipatory way and on their own initiative.

While a few specific attempts exist in this direction, we do not have today a general theory of what it means for an artificial system to be proactive, nor the computational tools to realize it. The aim of this project is to provide such a theory.

In particular, our ambition is to model a form of proactivity that is human-centered, that is, where proactive decisions take into account the preferences of humans, respect human autonomy, and can be explained to humans.

The theory that we shall develop will encompass the ability to reason about the human world, to predict its possible evolutions, and to proactively decide and execute actions in consideration of the goals, intentions, and preferences of humans.

Our theoretical study will be complemented by the development of computational methods to be used in AI systems and robots, by the investigation of the formal properties of our theory and methods, and by empirical validations with physical robot systems.

The project will last four years and will be carried out by a team of a full professor (main applicant) supported by an associate professor and a postdoc.