Integrating Political Tracking Resources with the Wiki

[OAGr]

Context

We have 29 political candidates enriched with positions/sources in data/entities/people.yaml, and 32 political tracking resources cataloged in data/political-tracking-resources.yaml. The wiki already has entity directories, FactBase for structured facts, auto-update for news, source-check for verification, and legislation tracking. How do we deeply integrate these political tracking resources into the system?

Current State

• Politicians are type: person entities with tags: [politics]
• They have positions (ExpertPosition), sources, description, knownFor
• The people directory shows: name, role, affiliation, born year, net worth
• Legislation entities already have keyPoliticians, votes, stakeholders
• Auto-update fetches news via RSS + web search and routes to pages
• FactBase stores temporal structured facts (number, text, ref, range)

---

Tier 1: Quick Wins (existing infrastructure, days not weeks)

1. Auto-generate external links for all politicians

For each politician entity, programmatically generate verified links to tracking sites (GovTrack, Ballotpedia, OpenSecrets, VoteSmart, Congress.gov). Store external IDs (bioguide, CID, etc.) as customFields. Must verify each URL resolves before adding.

2. Politicians filter in People Directory

Add a prominent "Politicians" quick-filter using the politics tag. Add role field to politician entities (e.g., "U.S. Senator").

3. Political Tracking Resources wiki page

Dashboard page rendering data/political-tracking-resources.yaml as a browsable, categorized table with EA-relevance ratings as colored badges.

4. Cross-link politicians to legislation

Bidirectional relatedEntries between politician entities and legislation they authored/co-sponsored.

5. EA stance badges on entity profiles

Color-coded stance badges (green/amber/red/gray) on politician position tables, grouped by EA topic.

---

Tier 2: New Features (some new code, 1-2 week efforts)

6. Dedicated Politicians Directory (/politicians)

New directory page with EA-specific columns:

Name	Office	Party	AI Safety	Biosafety	Animal Welfare	Sci/Tech	Nuclear	Key Legislation	PAC Backing

Keep as type: person with tags: [politics] — create a filtered directory view rather than a new entity type.

7. Race Tracker Pages

Comparison pages for each contested race (NY-12, TX-10, Michigan Senate, etc.) using existing political-race entity type. Shows all candidates side-by-side.

8. Political News Auto-Update Sources

Add political news sources (AI policy, biosecurity, etc.) to data/auto-update/sources.yaml. Auto-routes to politician and legislation pages.

9. Stance Comparison Matrix Component

Reusable React component: candidates × topics with color-coded cells. Embeddable in race pages, topic pages, and the directory.

10. PAC/Funding Tracker

Track key AI-policy PACs (Leading the Future, Think Big, Jobs and Democracy, Public First Action) and their spending per candidate.

11. External ID Registry

Store bioguide, OpenSecrets CID, GovTrack ID, FEC ID as customFields — enables programmatic cross-referencing with all tracking resources.

---

Tier 3: Data Pipeline Integrations (significant infrastructure)

12. GovTrack/Congress.gov API Integration

Scheduled fetcher → voting records → FactBase facts with temporal data. Weekly GitHub Actions cron.

13. Scorecard Aggregator

Pull ratings from LCV, Humane Society, FP4A, Heritage Action → FactBase temporal facts. Display as radar/bar chart on politician profiles.

14. Campaign Finance Integration

FEC/OpenSecrets data → structured records (total raised, top donors, PAC contributions, independent expenditures).

15. EA Political Action Score (Composite)

Weighted 0-100 score across: AI safety (30%), biosecurity (20%), animal welfare (15%), science/tech funding (15%), nuclear risk (10%), competence (5%), bipartisanship (5%).

16. Prediction Market Integration

Polymarket API → election odds on race/candidate pages. Shows which EA-relevant races are competitive.

17. Automated Position Monitoring

Auto-update monitors for changes in politicians' stated positions, flags discrepancies with stored data.

---

Tier 4: Ambitious / Long-term Ideas

18. Interactive "Who Should I Support?" Tool

Users weight EA topics via sliders → see candidates ranked by weighted composite score → click through to profiles.

19. Voting Record Timeseries

FactBase temporal facts → line charts showing how a politician's EA-relevant voting alignment changes over time.

20. Political Influence Map

Network visualization: Politicians ↔ PACs ↔ Legislation ↔ Organizations ↔ AI Companies.

21. Automated Source Verification

Extend source-check to verify political claims against Congress.gov roll calls, LCV scorecard, FEC filings.

22. Election Outcome Tracker

Auto-update win/loss status, compute "EA win rate" across races, archive predictions vs outcomes.

23. Bill Impact Forecasting

Combine positions + bill tracking + prediction markets → probability that specific AI safety bills pass.

---

Key Architectural Decisions

1. Entity type: Keep politicians as type: person with tags: [politics] — avoids migration, leverages existing person infrastructure.
2. Stance data: Use existing positions field with standardized topic names + rating rubric for numeric conversion.
3. External data: FactBase for time-varying data (votes, scores). YAML for static data (IDs, campaign URLs).
4. Update frequency: Weekly auto-update is sufficient — real-time not needed for political data.
5. URL verification: All external links must be verified before adding — never trust generated URL patterns.

Recommended Sequence

Phase 1 (immediate): Tier 1 items — external links, politician filter, resources page, cross-links, stance badges
Phase 2 (next sprint): Politicians directory, stance matrix, political news sources, external IDs
Phase 3 (next month): Race tracker, GovTrack API, scorecard aggregator, PAC tracker
Phase 4 (longer term): Composite score, prediction markets, interactive tools, campaign finance

Notable Gaps in the Ecosystem

• No formal biosecurity congressional scorecard exists — JHU identifies champions but doesn't publish ratings
• No AI governance legislator scorecard exists — bill trackers exist but no one rates legislators on AI positions
• Animal welfare is a blind spot across nearly all candidates — most have zero documented positions

[OAGr]

Refined Architecture: Publications + Resources + Wiki Pages

After exploring the codebase more deeply, here's the cleanest integration path:

Layer 1: Publications (venue metadata)

Add political tracking sites to data/publications.yaml with a new type civic_data:

- id: govtrack
  domains: [govtrack.us]
  name: GovTrack.us
  type: civic_data
  credibility: 4
  peer_reviewed: false
  description: Congressional votes, bill progress, legislator activity since 1789

- id: opensecrets
  domains: [opensecrets.org]
  name: OpenSecrets
  type: civic_data
  credibility: 4
  peer_reviewed: false
  description: Campaign contributions, PAC/Super PAC spending, lobbying, dark money

This immediately gives them:

• Credibility ratings visible in /sources
• Domain matching (any resource URL from govtrack.us auto-links to this publication)
• Publication detail page at /publications/govtrack

Layer 2: Resources (individual politician profiles)

Each politician's GovTrack page, Ballotpedia page, etc. becomes a Resource that references the publication:

Resource: "GovTrack: Valerie Foushee"
  url: https://www.govtrack.us/congress/members/valerie_foushee/456913
  publication_id: govtrack
  related_entity_ids: [valerie-foushee]

These show up on the politician's entity page under "Sources" and on the publication's page under "Resources."

Layer 3: Wiki Pages (for the most important trackers)

For the ~10 most EA-relevant trackers, create full wiki pages (MDX) explaining: what data they provide, API capabilities, EA-specific value, and how we integrate them.

What this means for the 32 trackers:

• ~20 go into publications.yaml as civic_data entries (venues for data)
• ~10 get wiki pages (GovTrack, OpenSecrets, Ballotpedia, CSET, Humane Scorecard, LCV, etc.)
• The YAML reference file stays as a catalog with EA-relevance ratings and API notes
• Individual politician profile URLs get added as Resources linking publication -> politician entity

Key insight

congress.gov is already in publications as type: government. We're extending that pattern — civic data infrastructure gets the same treatment as government agencies and think tanks.

Open question

Should the EA-relevance ratings and API notes from political-tracking-resources.yaml be absorbed into the publications schema (new optional fields), or kept separate? The publications schema is minimal. Probably better to keep the reference YAML separate and link by ID.

[OAGr]

Branch: claude/deploy-tasks-verify-cmd

Closing — overlaps significantly with #3639 (Political Data Roadmap: 6-Month Plan), which is the canonical political data plan. Tier 1 items here should be folded into #3639 Month 1. — Discussion review 2026-04-08.