Can surgical robots think beyond what they see?

Surgical AI copilots built on vision-language-action models excel at pattern recognition but struggle with the hidden complexity of endoscopic procedures—tissue damage beneath the surface, surgeon intent beyond the visible field. This paper argues that adding reasoning capability—the ability to integrate multiple cues, interpret intent, and infer what's not directly visible—could shift these systems from reactive tools into collaborative problem-solvers that actually reduce surgeon burden and improve safety.