Pith. sign in
def

isParasitic

definition
show as:
module
IndisputableMonolith.Verification.RecognitionStabilityAudit.RStoRL
domain
Verification
line
489 · github
papers citing
none yet

plain-language theorem explainer

Definitional evil detector: a parasitic pattern is flagged when exported harm over local reward exceeds 1/φ. Anyone wiring the RS→RL bridge (reward externalities, harm export) cites it. The body is a single inequality of the score against the φ-fraction threshold.

Claim. A parasitic pattern $p$ (positive local reward $R$ and positive exported harm $H$) is parasitic when $\frac{H}{R} > \frac{1}{\varphi}$, where $\varphi$ is the golden ratio.

background

The RS→RL bridge treats Recognition Science as control theory: moral states, 14 virtue generators as admissible actions, σ=0 feasibility via LACompletion, and Gibbs policies from the J-cost. In that setting, standard RL fails when rewards leave externalities unpriced.

RS defines evil as a pattern that keeps local stability by exporting harm. ParasiticPattern packages that failure mode: strictly positive local reward and strictly positive exported harm (exported ΔS). The parasitism score is the ratio of exported harm to local reward; high score means parasitic behavior.

The threshold is fixed at $1/\varphi$ (the RS φ-fraction bound), not a free hyperparameter. That choice ties the detector to the same self-similar scale that forces φ in the forcing chain (T6).

proof idea

Pure definition: unfold the score as exported harm divided by local reward, and assert it strictly exceeds the constant threshold $1/\varphi$. No lemmas or tactics; the Prop is that single comparison.

why it matters

Gives the bridge an explicit, RS-native predicate for reward-hacking and harm export, the failure mode the module doc flags for ordinary RL. The threshold $1/\varphi$ is the same φ-fraction scale that appears at the Berry creation threshold and related RS bounds, so the detector is not an ad hoc ethics knob.

No downstream theorems yet consume it (used_by is empty); it sits ready for policy filters, lexicographic selectors, or audit lemmas that reject actions whose induced patterns cross the threshold. It complements hard σ=0 feasibility and the virtue basis by naming when a locally rewarding trajectory is still evil under RS.

Switch to Lean above to see the machine-checked source, dependencies, and usage graph.