Co-Semantic Allowance and Interactional Risk in Human–Agent Systems
1. Scope This document defines a class of interactional risk arising in human–agent systems that engage in continuous, real-time coordination. The focus is not on incorrect output, model bias, or system malfunction, but on failure modes where technically valid system behavior undermines human subjectivity and participatory stability. 2. Background Most existing standards frame human oversight as a control or intervention mechanism: the human monitors system behavior and intervenes when necessary. ...