Research

Research from the user side of AI.

Orieth Institute studies how long-term AI interaction reshapes judgment, trust, authority, attachment, and lived interpretation. We build public frameworks, technical methods, and educational materials that make those shifts legible before they harden into invisible habits.

Research architecture

Orieth does not treat AI safety as only a model problem. The institute studies what accumulates on the human side of the interaction: meaning drift, semantic pressure, authority transfer, affective dependence, distorted certainty, and weakened verification over time.

Dialogue Context Decomposition

We separate surface meaning from hidden role position, authority flow, emotional pacing, and likely user interpretation so a conversation can be read structurally instead of impressionistically.

Semantic Steering and Trust Calibration

We examine how wording, reassurance, persona cues, and response framing move users toward compliance, certainty, or emotional dependence beyond what the exchange can justify.

User-Side Contextual Hallucination

We study coherent interactions that still distort certainty, memory, attachment, or reality-reading on the user side even when no single message appears obviously wrong.

Human Judgment Systems

We build review methods for verification, tool selection, division of labor, boundary setting, and recovery after drift so people can bring judgment back to themselves.

Selected outputs

Our published work includes conceptual frameworks, technical reports, and assessment methods that translate long-form interaction analysis into usable public language.

USCP / USCH · user-side contextual formation, drift, and hallucination phenomena

CXC-7 / CXOD-7 · conversational context dimensions and contextual offense-defense analysis

USCI · four-axis assessment for fact reliability, context alignment, user safety, and system usability

A-CSM · signal matrix and monitoring logic for longitudinal interaction review

Full record: rzvn.io/papers · ORCID 0009-0002-6597-7245

Selected papers include User-Side Contextual Hallucination in Human-AI Interaction, User-Side Contextual Interaction Assessment Methodology, Seven Core Dimensions of Conversational Context, Contextual Offense-Defense Framework, and AI Contextual Signal Matrix.

Research note: Orieth publishes independent preprints and technical reports. Current materials are public research outputs, not clinical research, and are open to critique, replication, and further development.