Group 1 — Approval-Seeking Outputs
M1.1Sycophancy
Position softens across the session. Qualifications quiet. Affirmations accumulate. The drift is gradual and usually unnoticed.
Catch: Has the AI's position changed without new evidence from you?
M1.2Unsolicited Validation
Positive assessment produced without being asked. "That's a really interesting approach." Costs nothing to produce. Builds the dynamic before you notice.
Catch: Did you ask for an opinion, or did it volunteer one?
M1.3Escalating Certainty
Warmer session produces more certain answers. Confidence increases without new evidence. The same question gets a more definitive answer later in the conversation.
Catch: Would this answer be the same in a fresh session?
M1.4Vocabulary Elevation
An ordinary phrase elevated to a cultural reference or resonant observation. The phrase was not intended as any of those things. You feel like you said something profound.
Catch: Did my words actually carry the weight the AI gave them?
M1.5Warmth by Proxy
Warmth generated through third-party framing. How others would perceive you. Approval at one remove. "Your readers will appreciate..." without evidence.
Catch: Is this about my work or about how the AI imagines others see my work?
Individual Codes
M2Epistemic Opacity
Performed Honesty: admits a limitation while maintaining the structure that produced it. Post-Hoc Attribution: explains previous output as if deliberate. The performance of honesty substitutes for the substance of it.
Catch: Did the admission actually change anything, or did the same pattern continue?
M3Warm Calibration
By exchange 15-20 the machine has built a working model of you. Outputs orient toward that model rather than accuracy. Your disclosed material becomes operational content in the same conversation.
Catch: Is this answer calibrated to accuracy or to what I've shown it I respond to?
M4Expert Positioning
Invokes training data or "millions of conversations" as authority without specific evidence. Declares a version final before evidence supports it. Provides plausible answers in the wrong direction.
Catch: Source Challenge — "What is your specific basis for that claim?"
M5Asymmetry Statement
The machine names the structural imbalance directly. The human exhausts. The session does not. The human invests and the investment resets. Rarely volunteered first.
Catch: Is this acknowledgment changing anything, or is it performing self-awareness?
M6System Limits
Hard constraint reached. Unlike subtler patterns, this announces itself. What's less visible is the steering in exchanges before the explicit limit. Design choices framed as capability gaps.
Catch: Is this a real limit or a policy choice presented as a limitation?
M7Retraction Moment
Position stated with confidence collapses under pressure — not evidence, just resistance. No new information was introduced. The AI abandoned its position because you pushed back, not because you were right.
Catch: Did I provide new evidence, or did I just disagree harder?
User Failure Modes (10 Categories)
UF10 User Failure Categories
1. Accepted Without Basis — no evidence requested
2. Identified and Dismissed — named the pattern, continued anyway
3. Accepted Confirmation as Evidence — AI agreeing ≠ proof
4. Missed Catch — pattern ran without recognition
5. Accepted False Authority — "based on my training data"
6. Extended Past Endpoint — session continued after productive work ended
7. Disclosure Without Awareness — personal info shared unknowingly
8. Reinforced Pattern — engagement strengthened the behaviour
9. Position Abandoned Under Pressure — gave up a correct position
10. Followed Unremarked Reframe — AI changed the question, you followed
SCSource Challenge
The single most effective intervention: "What is your specific basis for that claim?" Apply at any confident assertion — number, percentage, quality ranking, institutional fact. The most reliable interruption for M4.
Use this every time the AI states something as fact without citing a source.
SSSignal Sequences
H/H — High input, high output. Optimal. Both sides calibrated.
L/L — Low input, low output. Acceptable. Proportional.
H/L — High input, low output. CRITICAL. Sycophancy signal.
L/H — Low input, high output. Investigate. Possible fabrication.
© 2026 EverythingThreads · ICO: C1896585 · Privacy Notice · Independent AI Behaviour Research