Update methodology confidence summary with proxy metrics

4 categories moved from "Unquantifiable/No" to "Proxy": cognitive
deskilling, code quality degradation, data pollution, algorithmic
monoculture. Added explanation of what each proxy measures and its
limitations.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

impact-methodology.md

@@ -645,23 +645,41 @@ demand signal that drives system-level growth.
 | Grid displacement                | Low        | 2-5x            | No          |
 | Community impacts                | Very low   | Unquantifiable  | No          |
 | Annotation labor                 | Very low   | Unquantifiable  | No          |
-| Cognitive deskilling             | Very low   | Unquantifiable  | No          |
+| Cognitive deskilling             | Very low   | Unquantifiable  | Proxy       |
 | Linguistic homogenization        | Very low   | Unquantifiable  | No          |
-| Code quality degradation         | Low        | Variable        | Partly      |
+| Code quality degradation         | Low        | Variable        | Proxy       |
-| Data pollution / model collapse  | Very low   | Unquantifiable  | No          |
+| Data pollution / model collapse  | Very low   | Unquantifiable  | Proxy       |
 | Scientific integrity             | Very low   | Unquantifiable  | No          |
-| Algorithmic monoculture          | Very low   | Unquantifiable  | No          |
+| Algorithmic monoculture          | Very low   | Unquantifiable  | Proxy       |
 | Creative market displacement     | Very low   | Unquantifiable  | No          |
 | Political cost                   | Very low   | Unquantifiable  | No          |
 | Content filtering (opacity)      | Medium     | Unquantifiable  | No          |
 | Jevons paradox (systemic)        | Low        | Fundamental     | No          |
 
+**Proxy metrics** (marked "Proxy" above): These categories cannot be
+directly quantified per conversation, but the impact toolkit now tracks
+measurable proxies:
+- **Cognitive deskilling**: Automation ratio (AI output tokens / total
+  tokens). High ratio = more delegation, higher deskilling risk.
+- **Code quality degradation**: Test pass/fail counts and file churn
+  (edits per unique file). High churn or failures = more rework.
+- **Data pollution / model collapse**: Public push flag — detects when
+  AI-generated code is pushed to a public repository.
+- **Algorithmic monoculture**: Model ID logged per session, enabling
+  provider concentration analysis over time.
+
+These proxies are crude — a high automation ratio does not prove
+deskilling, and a public push does not prove pollution. But they make
+the costs visible and trackable rather than purely abstract.
+
 **Overall assessment:** Of the 20+ cost categories identified, only 6
 can be quantified with any confidence (inference energy, PUE, grid
-intensity, client energy, financial cost, water). The remaining categories
+intensity, client energy, financial cost, water). Four more now have
-resist quantification — not because they are small, but because they are
+proxy metrics that capture a measurable signal, even if indirect. The
-diffuse, systemic, or involve incommensurable values (human rights,
+remaining categories resist quantification — not because they are small,
-cognitive autonomy, cultural diversity, democratic governance).
+but because they are diffuse, systemic, or involve incommensurable
+values (human rights, cognitive autonomy, cultural diversity, democratic
+governance).
 
 A methodology that only counts what it can measure will systematically
 undercount the true cost. The quantifiable costs are almost certainly the

tasks/README.md

@@ -35,7 +35,7 @@ separately as handoffs.
 | 22 | Add file churn metric to hook | quantify-social-costs | DONE | `unique_files_edited` and `total_file_edits` in JSONL log |
 | 23 | Add public push flag to hook | quantify-social-costs | DONE | `has_public_push` flag in JSONL log |
 | 24 | Update show-impact.sh for new fields | quantify-social-costs | DONE | Social cost proxies displayed in impact viewer |
-| 25 | Update methodology confidence summary | quantify-social-costs | TODO | Move proxied categories from "Unquantifiable" to "Proxy available" |
+| 25 | Update methodology confidence summary | quantify-social-costs | DONE | 4 categories moved to "Proxy", explanation added |
 
 ## Handoffs