On the Identifiability of User Adaptation in Co-Adaptive Neural Interfaces

The paper presents an analysis of identifiability in co-adaptive neural interfaces, revealing that closed-loop encoder estimates fail to uniquely identify user adaptation, as they instead reflect characteristics of the entire human-machine system. It discusses the implications for understanding behavioral adaptation and proposes specific conditions necessary for achieving identification. This work is significant for practitioners as it highlights challenges in accurately interpreting user adaptation in interactive AI systems, which is crucial for improving user experience and system performance.