docs: attribute RC+ξ to Camlin, rename our metric to Perspective Dispersion (Υ)

Our metric (variance of perspective outputs around their centroid) is a
DIFFERENT quantity from Camlin's ξ = ‖Aₙ₊₁−Aₙ‖² (successive hidden-state
difference, arXiv:2505.01464). Credit the RC+ξ name/formalism to him;
rename ours to Perspective Dispersion (Υ), Γ = 1/(1+Υ).

- docs/ATTRIBUTION_perspective_dispersion.md: full attribution + the
  independent-development provenance (Pi Nov 2024, Zenodo DOI Apr 2025)
- state_engine_v8.py tension_from_texts: canonical name in docstring;
  return-tuple symbols kept pending a tested full rename

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: Jonathan Harrison

docs/ATTRIBUTION_perspective_dispersion.md

@@ -0,0 +1,84 @@
+# Attribution & Naming: Perspective Dispersion (Υ)
+
+*Author: Jonathan Harrison (Raiff1982), Raiffs Bits LLC — July 17, 2026*
+
+## Summary
+
+Earlier versions of this system labeled its inter-perspective disagreement
+metric using the term **RC+ξ** and the symbol **ξ ("epistemic tension")**.
+On review, that was two separate naming errors, and this note corrects both.
+The metric is renamed **Perspective Dispersion (Υ)**, with credit given to
+the prior work whose names were inadvertently overlapping.
+
+## Correction 1 — the RC+ξ name and ξ formalism belong to Camlin
+
+The term **RC+ξ** and the symbol **ξ** for "epistemic tension" originate with:
+
+> Jeffrey Camlin, *"Consciousness in AI: Logic, Proof, and Experimental
+> Evidence of Recursive Identity Formation."* Meta-AI: Journal of
+> Post-Biological Epistemics. arXiv:2505.01464v1, May 1, 2025.
+
+Camlin defines his epistemic tension as:
+
+> **ξₙ = ‖Aₙ₊₁ − Aₙ‖²**
+
+where *A* is the hidden state of a **single model at recursion step n**. It
+measures how much **one trajectory's internal latent state changes between
+successive recursive steps** — a temporal, intra-trajectory, latent-space
+quantity, central to his proof of recursive identity convergence.
+
+This project adopted the "RC+ξ" vocabulary in mid-2025–2026 to describe its
+own system. That adoption was not intentional appropriation — the terminology
+most likely entered this work through language models whose training data
+included Camlin's paper, without an attached citation. Regardless of how it
+arrived, **the RC+ξ name and the ξ = ‖Aₙ₊₁ − Aₙ‖² formalism are Camlin's, and
+are credited to him here.** He should own the name of his math.
+
+## Correction 2 — this system's metric is a *different* quantity
+
+The quantity this system actually computes (`reasoning_forge/state_engine_v8.py`,
+`tension_from_texts`) is:
+
+> **Υ = (1/k) · Σᵢ ‖vᵢ − v̄‖²**
+
+where *vᵢ* is the (L2-normalized term-frequency) vector of **perspective i's
+actual response**, *v̄* is the mean of the *k* perspective vectors, and the
+complementary coherence is **Γ = 1 / (1 + Υ)**.
+
+This is the **variance of an ensemble of simultaneous perspective outputs
+around their centroid** — a cross-sectional, output-space measure of *how much
+independent perspectives disagree on one query*. It is **not** Camlin's
+successive-state-difference norm. He measures one mind changing over time;
+this measures many minds disagreeing at once. Continuing to label an
+ensemble-variance with Camlin's ξ misattributes his formalism and mislabels
+this one.
+
+## The new name, and its own family
+
+This metric is renamed **Perspective Dispersion (Υ)**.
+
+For full transparency, Υ belongs to a recognized family of multi-agent
+spread/agreement measures — e.g. **semantic dispersion** (mean pairwise
+embedding distance), **consensus variance**, and the **order parameter φ**
+(mean alignment to the ensemble centroid). Υ's specific formulation
+(variance-from-centroid of TF vectors over *named cognitive perspectives*,
+with Γ = 1/(1+Υ)) is this project's own, but the *class* of measure is not
+claimed as novel. If Υ appears in a publication, these relatives should be
+cited.
+
+## Provenance note (why this is a correction, not a derivation)
+
+This system's multi-perspective cognitive architecture was developed
+independently and published **before** Camlin's paper:
+
+- Multi-perspective reasoning engine with named lenses (Newton, DaVinci,
+  Quantum, Empathy, Philosophy) — `Raiff1982/pi-the-assistant`,
+  `Cognitivereasoning.py`, **November–December 2024**.
+- Sovereign modular architecture (BroaderPerspectiveEngine, NeuroSymbolicEngine,
+  EthicalAIGovernance) — Zenodo **DOI 10.5281/zenodo.15214462, April 14, 2025**.
+
+The two lines of work are **convergent, not derivative**: an independently
+engineered system and an independently proven theorem reached overlapping
+territory. This project did not take Camlin's work — its system predates his
+paper — and it does not claim his RC+ξ name or formula. Both facts are true,
+and this note exists to keep them both visible.

reasoning_forge/state_engine_v8.py

@@ -84,13 +84,19 @@ def _sq_distance(a: Dict[str, float], b: Dict[str, float]) -> float:
 
 
 def tension_from_texts(perspective_texts: Dict[str, str]) -> Tuple[float, float]:
-    """Spec formula: xi = mean squared distance of attractors from their mean.
-
-    Attractors here are the REAL perspective responses (vectorized), so xi
-    measures how much the perspectives actually disagreed on this query.
-    Returns (tension, coherence) with coherence = 1 / (1 + tension).
-
-    xi ranges ~0 (identical responses) to ~2 (fully disjoint vocabulary).
+    """Perspective Dispersion (Υ): mean squared distance of the perspective
+    vectors from their centroid — how much the perspectives disagreed on this
+    query. Returns (Υ, Γ) with coherence Γ = 1 / (1 + Υ).
+
+    Υ ranges ~0 (identical responses) to ~2 (fully disjoint vocabulary).
+
+    NAMING: this quantity was previously labeled "epistemic tension ξ" after
+    the RC+ξ framework. That name/formalism belongs to Camlin (arXiv:2505.01464,
+    ξ = ‖Aₙ₊₁−Aₙ‖², a successive-hidden-state difference — a DIFFERENT quantity).
+    Ours is an ensemble variance of simultaneous perspective outputs, renamed
+    Perspective Dispersion (Υ). See docs/ATTRIBUTION_perspective_dispersion.md.
+    (Return tuple name kept for now to avoid breaking callers; a full symbol
+    rename is a separate, tested pass.)
     """
     vectors = [
         _tf_vector(t) for t in perspective_texts.values()