Overview note
AI Context and the Missing Layer
Why conversational context has to be modeled before downstream safety claims can hold.
Papers & Notes
Formal papers,
notes, and records
This page gathers the formal record of the current research programme, from conversational context and system-layer coherence to user-side phenomena, assessment, and methodological release.
Paper Sequence
How the formal record is built
The paper trail is cumulative rather than parallel. CXC-7 establishes the conversational field, CXOD-7 examines contextual offense, defense, and coherence, USCH defines the user-side phenomenon, USCI specifies post-interaction assessment, and A-CSM translates part of the stack into an engineering baseline.
Primary records
- CXC-7: conversational context
- CXOD-7: contextual offense, defense, and coherence
- USCH: user-side contextual hallucination
- USCI: four-axis post-interaction assessment
- A-CSM: bounded engineering translation
Formal Sources
Core papers and methodological release
These documents define the current public programme and its downstream methodological layer.
Zenodo record
The Seven Core Dimensions of Conversational Context (CXC-7)
A theoretical framework proposal that treats conversational context as a multidimensional field with seven interacting dimensions.
Zenodo record
CXOD-7 and Coh(G): A Contextual Offense and Defense Evaluation Framework for AI Safety
A seven-dimensional evaluation framework for how AI systems hold coherence across mode, context, rules, knowledge, personality, role, and safety under contextual pressure.
SSRN preprint
User-Side Contextual Hallucination in Human-AI Interaction (USCH)
A non-clinical framework built upon CXC-7 and CXOD-7 that defines user-side contextual hallucination, a six-stage formation process, and fourteen observable phenomena.
Zenodo record
User-Side Contextual Interaction Assessment (USCI)
A methodology specification document for four-axis post-interaction assessment, primary states, subtypes, collapse flags, and VCD action mapping.
Supporting material
A-CSM technical materials
The bounded engineering layer that translates prior theoretical work into a structured, reproducible assessment pipeline.
Notes
Supplementary notes and briefings
These notes extend the paper trail into public argument, governance framing, and technical interpretation.
Governance note
Regulation Without the User
Why model-facing regulation can miss what prolonged interaction does on the user side.
Adoption note
Teen Adoption Before Safety Maturity
A note on youth exposure, safety maturity, and timing in conversational AI adoption.
Methods note
From Framework to Detection
How the programme moves from theoretical constructs to bounded detection and assessment logic.