Blog.

Technical framework for detecting and classifying cognitive drift signals in AI conversation contexts. Defines signal taxonomies, detection algorithms, and criteria for monitoring user-side contextual stability. doi.org/10.5281/zenodo.19097267

Four-axis assessment methodology (FR, CA, SR, SA) for evaluating contextual risk in human-AI interactions, with structured criteria for identifying deviation and alert states. doi.org/10.5281/zenodo.18678458

Introduces User-Side Contextual Hallucination: cognitive distortions, behavioral deviations, and emotional dependency arising through sustained AI interaction, even when individual responses are accurate. Fourteen distinct phenomena across three clusters. doi.org/10.2139/ssrn.6135732

Extends CXC-7 with an offense-defense matrix, modeling how contextual signals in AI conversation create vulnerability or protection for users. Provides a mechanism for assessing systemic contextual risk. doi.org/10.5281/zenodo.17403793

Introduces a taxonomy of seven core dimensions present in AI conversation contexts: intent, tone, role, knowledge state, emotional state, goal, and history. Structural foundation for analyzing user-AI interaction patterns. doi.org/10.5281/zenodo.18615646