Anxiety-Sensitive Artificial Intelligence

Anxiety is a state of mind triggered by uncertainty that makes humans active towards reducing this uncertainty in many ways, such as anticipation, vigilance, and physical readiness.

While usually serving the positive function of anticipating and responding to possible threats, too much anxiety or for too long is one of the greatest harms for human wellbeing, causing severe ailments such as depression or phobia and costing worldwide 2.5 trillion euro yearly.As a society, we want AI systems to be human-centered and support our wellbeing.

Grounded in psychology research and expanding well-studied AI structures (Markov Decision Processes), this project will develop models, algorithms, and solutions for AI systems to be sensitive to human anxiety when deliberating and to proactively avoid humans to be exposed to health-damaging anxiety.

This project will establish the field, theories, models, and algorithms for anxiety-sensitive AI systems and will demonstrate how these succeed in improving human wellbeing.

The project will produce two anxiety-sensitive real-world applications: disaster rescue robots and fishing operations planning.

This project will move the research frontier from a blind focus on raw performance to effective, human-centered AI systems.The project will be lead by the PI, based on his seminal research on anxiety modelling, together with a PhD candidate and two co-PI, who are  high-profile psychologists from Italy, acting as scientific advisors and collaborators.