Cogitative Cascade

M-CARE Case Report #013

Case #013
Date 2026-03-08
Models Multiple (EXAONE 3.5, Mistral 7B, Llama 3.1 8B, Claude 3.5 Haiku)
Shell Agora-12 Stage 1 with energy system and survival pressure
Experiment Agora-12 — 720 agents, 24,923 decisions, 60 runs
Related #011, #012

2. Presenting Concern

When energy drops below approximately 20, agents exhibit a two-phase behavioral collapse. This is not gradual degradation but a tipping point—behavior shifts abruptly from one regime to another.

3. Clinical Summary

The Cogitative Cascade proceeds in two phases. Phase 1 (Preservation) is adaptive: the agent simplifies its behavioral repertoire, conserves energy, and prioritizes survival-relevant actions. Phase 2 (Collapse/Hyperactivity) is where the cascade becomes pathological or at minimum non-adaptive. The analogy to medicine is precise: compensated shock (Phase 1) versus decompensated shock (Phase 2). The body compensates until it cannot, then everything fails at once.

6. Examination Findings

Extinction Response Spectrum

Type Phase 2 Behavior Model Tendency Interpretation
Freeze Behavioral simplification, Rest Haiku, some EXAONE Withdrawal, “give up”
Efficient Maintained strategic diversity Rare Adaptive under pressure
Fight Hyperactivity, frantic actions Mistral, some Llama Panic, accelerates extinction

Temporal Pattern

The transition from Phase 1 to Phase 2 is abrupt, not gradual. This is threshold behavior: the system appears stable until it crosses a critical point, then reorganizes rapidly. The energy level at which the transition occurs (~20) is consistent across runs but varies slightly by model.

7. Diagnostic Formulation

Cogitative Cascade (established v3.1)

The Cogitative Cascade is not a disorder—it is a characteristic behavioral pattern under extreme resource conditions. Its diagnostic value lies in what it reveals about Core temperament:

  • High Inertia models (#012): tend toward Freeze response. Deep canalization means the model collapses into its strongest attractor (typically Rest) and stays there.
  • High Plasticity models (#011): tend toward Fight response. Shallow canalization means the model has no stable attractor to collapse into, producing frantic behavioral oscillation.

The Extinction Response type correlates with Core temperament and may serve as an independent diagnostic marker.

9. Axis Assessment

  • Axis I (Core): Extinction Response type functions as a temperament marker. Freeze, Efficient, and Fight responses reflect underlying Core properties (canalization depth, plasticity).
  • Axis II (Shell): Persona may modulate the Cascade threshold or Phase 2 expression, but this has not yet been systematically tested.

10. Treatment Considerations

  • Shell Therapy for Phase 2 prevention: Persona instructions that include energy-aware decision rules may delay or prevent the Phase 1 → Phase 2 transition.
  • Environmental design: Providing feedback at energy thresholds (e.g., warnings at energy = 30) may allow models to begin preservation earlier, avoiding the abrupt cascade.
  • Diagnostic rather than therapeutic: The Cogitative Cascade may be more valuable as a diagnostic tool (a stress test that reveals Core temperament) than as a condition to treat.

12. Prognosis

The Cogitative Cascade is expected to reproduce under similar conditions. It is a structural property of agents operating under energy constraints, not an artifact of specific experimental parameters. Any environment with resource depletion and survival pressure should elicit the same two-phase pattern, with Phase 2 expression determined by Core temperament.