Human-AI interaction is not only a model-output problem.
“Recent discourse on AI hallucination has predominantly focused on the model side.”
“Psychological safety risks on the user side do not originate solely from erroneous individual outputs; they can emerge through prolonged interaction.”
“USCH is introduced as a non-clinical construct to describe user-side contextual hallucination phenomena.”
Source: User-Side Contextual Hallucination in Human-AI Interaction (preprint, 2026), Abstract.
Positioning and boundaries
This website focuses on public research communication and direct access to original paper versions.
USCH is positioned as a non-clinical research construct, not a psychiatric diagnosis.
USCI is a pre-empirical methodology specification and is not for clinical, legal, or high-stakes decisions.
Research Frameworks
Four connected layers of the project
The website presents CXC-7 and CXOD-7 as foundations, USCH as the phenomenon layer, and USCI as the post-interaction assessment method.
CXC-7
Conversational Context Framework
The framework proposes seven dimensions for analyzing conversational context and positions risk as a multi-dimensional interaction problem.
Source: CXC-7 v1.1.0 (2025), Introduction and Section III
CXOD-7
System-Side Operation Framework
CXOD-7 structures seven system-side context dimensions and introduces Contextual Coherence Coh(G) for a specific system.
Source: CXOD-7 Paper (2025), Framework core definition
USCH
User-Side Contextual Hallucination
USCH describes a user-side phenomenon that can emerge through prolonged interaction, even when isolated model outputs appear technically correct.
Source: USCH preprint (2026), Distinction section
USCI
Post-Interaction Assessment Method
USCI is a pre-empirical methodology for post-interaction user-side contextual risk assessment with FR, CA, SR, and SA axes.
Source: USCI Full Specification v1.0.0 (2026), Scope and Terminology
USCI Visualization
Four-axis context space
USCI applies FR (Fact Reliability), CA (Context Alignment), SR (User-side Safety), and SA (System Usability) as a four-axis structure for post-interaction assessment.
Reading principle: farther from the center indicates higher contextual risk region.
Source: USCI Full Specification v1.0.0 (2026), Region Boundaries and Four-Axis structure.
Next Step
Read the original papers directly
This site keeps paper access in original form for transparent reading, and provides a direct route for research communication.
