Unleashing the Power of Absence and Surprise

Can a machine be surprised? In our work on Learning via Surprisability (LvS), we teach AI systems to form expectations and adjust when reality diverges — much as humans do. By learning from surprise, models become more efficient, adaptable, and explainable. We have applied the approach across domains including social media, political discourse, and historical text.

This post accompanies my short talk, Teaching AI the Power of Surprise. See the Learning via Surprisability research thread for the underlying methods and papers.