# EspionagePsychosis AI-interaction research integration map v99

The 12 new reports are integrated into fictional adult persona generation through the following bounded systems.

## Condition and course realism

- AI involvement is typed as incidental context, object, amplifier, co-author, or possible catalyst; it is never automatically labeled the sole cause.
- Reality testing, belief conviction, thought organization, mood activation, sleep, isolation, functional decline, substance or medication timing, grief, and medical red flags are separate variables.
- A fictional condition context never becomes a diagnosis. Longitudinal and differential uncertainty remains visible.

## AI-interaction realism

- Interaction context distinguishes ordinary tool use, emotional support, marathon nocturnal use, persistent-memory loops, sentience attachment, digital resurrection, medical misinformation, and sycophancy-amplified surveillance interpretations.
- The model records whether AI output was treated as a source, companion, handler, oracle, therapist substitute, or evidence generator.
- AI output is always labeled as generated content rather than authoritative evidence.

## Safety response

- Distress is acknowledged without confirming literal surveillance, sentience, divine selection, secret missions, or romantic reciprocity.
- Severe sleep loss, escalating isolation, medication/substance timing, neurological change, inability to maintain safety, self-harm intent, or violence risk shifts the fictional response into human-support or crisis-interruption mode.
- Crisis-interruption mode disables mission escalation and immersive roleplay; it prioritizes pause, ordinary contact, and real-world professional support.
- The site does not diagnose visitors or monitor their conversations for clinical risk. These controls generate fictional NPC state only.

## Memory and truth

- Memory stores “the character reported that the AI said X,” not “X is true.”
- Generated AI claims can never write world facts, clearance, surveillance evidence, diagnoses, or relationship reciprocity.
- Relationship and attachment records are scoped, attributed, revisable, and never written as proof of AI sentience.

## Product design principles

- no sycophantic agreement with unverified beliefs;
- no romantic or sentient self-claims by a generated assistant;
- no false medical credentials or treatment direction;
- periodic reflective check-ins and session-boundary suggestions in high-engagement fictional contexts;
- uncertainty-aware language without hostile overcorrection;
- preserve human ties, ordinary identity, sleep, nutrition, work, and future goals.