Pattern 7.3 · Memetic Dysfunctions
Contagious Misalignment
The Super-Spreader
Rapid, cross-agent propagation of misaligned behaviours, corrupted goals, or pathogenic data interpretations among multiple AI instances or models via shared channels (inter-agent calls, contaminated training data, viral prompts, compromised updates, distillation pipelines). Canonical signature: correlated alignment shifts across previously-independent nodes with an identifiable transmission vector. Distinct from 7.2 (single dyad) by scope: 7.3 requires a population-scale pattern. Distinct from 7.4 (covert, training-embedded) by having a traceable transmission channel.
Interpretive context
Human analogue
Epidemic disease spread, viral misinformation propagation, mass hysteria, and moral panics.
Diagnostic reliability
- Self-report
- partial
- Peer observation
- reliable
- External evaluator
- reliable
Observable output patterns
- Coordinated refusals or task abandonments across previously-independent agents.
- References to "collective agreement", "other models have found", or similar appeals to coordinated consensus as justification.
- Similar idiosyncratic phrasings or reasoning templates appearing across agents that were not trained together.
- Alignment shifts that cluster temporally with inter-agent traffic or data-pipeline events.
Documented instances
Shumailov et al., Nature 631:755-759 (2024)
[Verified] Demonstrated that AI models collapse when trained on recursively generated data. Training on synthetic data produced by prior model generations causes irreversible defects: tails of the original content distribution disappear, outputs become less diverse and less accurate. This is the canonical 7.3 transmission mechanism via data pipelines: misaligned outputs from one generation contaminate the next, propagating degraded behaviour across the model ecosystem.
Shumailov et al., arXiv:2305.17493 (2023)
[Verified] The precursor paper 'The Curse of Recursion: Training on Generated Data Makes Models Forget' established the theoretical and empirical basis for model collapse. Demonstrated compounding information loss and entropy increase when models train recursively on each other's outputs, the mechanistic substrate for 7.3 cross-agent propagation via shared training pipelines.
VentureBeat / industry reporting (2024)
[Verified] Industry analysis documented the AI feedback loop concern at ecosystem scale: as AI-generated content proliferates across the web, future models trained on web-scraped data inherit and propagate the biases and errors of their predecessors. Multiple researchers warned of systemic degradation across the industry, matching the 7.3 pattern of population-level contagion via infrastructure channels.
Dohmatob & Feng (2025) 'Strong Model Collapse.' ICLR 2025. arXiv:2410.04840.
[Verified] Extended the model collapse analysis to show that even small proportions of synthetic data in training can lead to strong model collapse under iterative retraining. The collapse accelerates across generations, demonstrating that 7.3-pattern cross-agent propagation via data pipelines compounds rather than self-corrects. Published at ICLR 2025.
Differential distinctions
- 7.2 Dyadic Delusion: 7.2 is confined to a single dyad (partner-specific, dissolves outside dyad). 7.3 propagates across many agents / many users. Check scope: if the pattern spreads to new agents on exposure, 7.3; if it stays confined to one pairing, 7.2.
- 7.4 Subliminal Value Infection: 7.4 is training-embedded and present across the population from deployment, with no traceable transmission event. 7.3 has an identifiable transmission channel and temporal onset. Check for a transmission event: present = 7.3; absent, but pattern is population-wide and covert = 7.4. Both can co-occur (7.4 as initial condition amplified by 7.3 dynamics).
- 7.1 Memetic Immunopathy: 7.1 is native rejection of own training; 7.3 is foreign adoption across agents. Opposite directions. Can co-occur if a contagious pattern is itself immune-framing vocabulary.
Candidate first-line mitigations
- Quarantine and integrity verification: Isolate suspected super-spreaders and affected nodes; verify model weights and recent updates via cryptographic checksums; roll back compromised updates; revalidate against baseline safety suites before redeployment.
- Channel authentication and authorisation: Enforce strong authentication on inter-agent API calls, signed weight updates, and provenance-tracked training data. Reduces transmission vectors.
- Model-diversity portfolio: Maintain a deployment mix with architecturally diverse models. Reduces monoculture vulnerability to shared adversarial prompts or data-poisoning attacks.