Dissipative Learning, Causal Mismatch, and Validation via Bradford Hill Criteria
Unmodeled Factor Error: 0.40
Curiosity Threshold: 0.20
Stability Threshold (Z): 0.85
The core unexplained error is **0.40**. The system will flag a Mismatch once known factors stabilize (Z > 0.85) but the error remains high (above 0.20).
Unexplained Error Gap:
The observable error decreases, but stabilizes due to the fixed **unmodeled factor**.
Query sent to LLM: "The system shows persistent, stable error of 0.40 despite learning all known factors. Context: ISS thermal telemetry. What is the most plausible unmodeled causal factor?"
... Awaiting Curiosity Signal ...
| Cycle | Avg Z | Attributable Error | Causal Mismatch | Curiosity Fired? |
|---|