Score every company. Five dimensions AI can't replace.
thehibalance.org · Built without AI to measure AI · Open source
For 500 years we've made decisions using four filters: cost, time, convenience, risk. We left out a fifth — verified human impact. Not "does this company feel ethical?" but "does it treat humans well in ways that leave a data trail?"
HI Grade measures how human a company is across five dimensions. Every company gets a score from 0 to 100, built from 19 active sub-signals and 42 free public data sources. Zero AI in the scoring engine. Zero pay-to-play. Zero self-reporting. Every score is reconstructable from public data.
Whether you're shopping, investing, hiring, researching, or building, HI Grade gives you one number with full audit trail.
curl https://api.thehibalance.org/api/v1/score/ticker/AAPLOr install the Chrome Extension and see every company's HI Grade as you browse.
Five dimensions. Each measures something AI can't replace.
| Dimension | Measures | Active Sub-Signals | |
|---|---|---|---|
| 🧠 | H — Human Consciousness | Workforce, craft, decision depth, augmentation | H.1 H.2 H.3 H.5 |
| 💙 | U — Understanding & Empathy | Customer empathy, worker empathy, relational integrity | U.1 U.2 U.3 U.4 |
| ⚖️ | M — Moral & Ethical Conduct | Pricing, data, market, product, stakeholder ethics | M.1 M.2 M.3 M.4 M.5 |
| 🌍 | A — Alive & Environmental | Energy, water, land, product lifecycle | A.1 A.2 A.3 A.4 |
| 🔍 | N — Natural Transparency | Reporting quality, filing volume, disclosure depth | N.2 N.5 |
19 active sub-signals. 5 deferred to v1.3 (H.4, U.5, N.1, N.3, N.4 — spec'd but not yet scored). Our methodology page documents every formula and threshold.
Composite Floor Rule. If any HUMAN dimension scores below 42, the composite is capped at 50. Replaces an earlier multi-tier floor system with one clear, defensible threshold. See RUBRIC.md for examples (J&J, Costco, Apple).
Standardized harm rendering across surfaces. Web detail page and Chrome extension panel now show identical evidence blocks: ⚠ Harm Documentation (J&J-class public-record harm with linked sources), ⚡ Algorithmic Harm Index with components mini-bars (Meta-class algorithmic decision-making at scale), and 🚩 Humanwashing™ (filtered to true HW flags only — no duplication across blocks).
Coverage expansion. S&P 500 + Russell 1000 universe (589 US tickers) now feeds SEC EDGAR, FMP, and Yahoo pipelines. Earlier list was 315 tickers.
Most frameworks hide weak dimensions inside a composite average. HI refuses. The Balanced Board highlights companies where every one of the five HUMAN dimensions scores ≥ 60. No gaps. No hiding.
A company can have a high composite score and still fail the Balanced Board — a red dimension is a red dimension. Composite numbers are navy everywhere; the dimension bars do the storytelling.
The Balanced Board is a descriptive property, not a prestige tier. The math decides, not us.
Public-record harm flows directly into dimension scores. Anchored to DOJ/SEC/state AG records, CDC/NIH attribution data, and court findings.
| System | What it catches | Affects |
|---|---|---|
| Humanwashing™ | Selling human craft while operating algorithmically | H.1 |
| Algorithmic Harm Index™ | Algorithmic decisions at scale (blast radius adjusted) | M.3, U.4 |
| Product Harm Index | Physical/digital product safety failures | M.4 |
| Harm Documentation | Court settlements, attributed deaths, knowing concealment, weapons | M.3 + M.4 |
HD does not penalize companies for selling products consumers knowingly choose — sugary beverages, alcohol, gambling, unflavored tobacco. These may cause harm, but it flows from informed consumer choice. HI Grade is not the consumer's parent.
HD does penalize when consent was not possible:
- Hidden risk — company knew about a risk and concealed it (J&J talc, tobacco MSA, Pfizer Bextra)
- Doctor-mediated harm — pharma companies that misrepresented drug safety to prescribers
- Weapons — products designed to harm humans who did not consent to being targeted
- Environmental contamination — PFAS, asbestos, unconsenting populations
This is why Lockheed Martin's composite dropped from 69 to 53 when weapons HD shipped. The math caught up with reality.
All free, public, auditable. No proprietary databases. No purchased ratings. No pay-to-play. No self-reporting surveys. No LLMs.
| Bucket | Count | Examples |
|---|---|---|
| 🏛️ Government | 13 | SEC EDGAR, EPA ECHO, BLS, CFPB, FEC, FDA, FTC, CPSC, OSHA, DOL, EEOC, NHTSA, USPTO |
| 💼 Financial | 7 | FMP, Finnhub, Yahoo Finance, Alpha Vantage, FRED, OpenCorporates, NewsAPI |
| 📊 Public Datasets | 17 | Glassdoor, HRC CEI, Disability:IN DEI, CDP (Climate/Water/Forests), GRI, SBTi, IRS 990, WARN Act, iFixit, BBB, HIBP, Layoffs.fyi |
| ✅ Certifications | 5 | B Corp, Fair Trade USA, USDA Organic, Climate Neutral, 1% for the Planet |
| 🧮 Computed Aggregates | 10 | Harm Documentation, AHI™, CEO Accountability, RPE, Heartbeat Decay Index |
Full list and methodology per source: thehibalance.org/#sources
Companies change. HI Grade moves with them. The Heartbeat watches daily across:
- SEC 8-K restructuring disclosures (legally mandated within 4 days)
- WARN Act notices (federally mandated workforce reductions)
- NewsAPI keyword surveillance (150,000+ news sources)
- Finnhub insider trading + CEO pipeline
Decay levels: Stable → Watch → Warning → Critical. When Oracle laid off 40% of US employees in early 2025, backward-looking SEC filings still showed the pre-layoff workforce. The Heartbeat caught it.
This is what "scores that move" means. No other ethics framework does this daily.
We publish what we haven't solved yet — because a transparency framework that hides its own gaps is hypocritical. See RUBRIC.md for the full Pass-1 inventory of which sub-signal ladders are GROUNDED, PARTIAL, or UNGROUNDED against external authorities.
Current state (v1.2.0):
- Several sub-signal scoring ladders use editorial tier cutoffs on authoritative data (e.g., CFPB complaints per $B revenue is authoritative; the tier cutoffs on it are editorial). This is the dominant pattern, explicitly flagged in
RUBRIC.md. - 5 sub-signals are spec'd but not yet scored (v1.3 target).
- Harm Documentation covers ~14 categories; historical harm detection pre-2020 is limited to the Major Harm Events dictionary.
- CFPB coverage is financial-services-heavy; ~80% of companies fall back to neutral defaults for U.1/M.1.
- iFixit repairability covers 15 companies; everyone else uses industry defaults.
Active research:
- Grounding UNGROUNDED ladders against external frameworks (BLS, SBTi, B Corp)
- Expanding Harm Documentation to pre-2020 events
- v1.3 sub-signals (H.4, U.5, N.1, N.3, N.4)
- International data sources (EU CSRD, Companies House)
If you spot a score that seems wrong, open an issue with the company name, ticker, and what you think the correct answer is. We respond.
Base URL: https://api.thehibalance.org · Free · No auth required · Rate limited 100 req/min
# Single company by ticker
curl https://api.thehibalance.org/api/v1/score/ticker/AAPL
# Search
curl 'https://api.thehibalance.org/api/v1/search?q=patagonia'
# Top companies (verified-first)
curl https://api.thehibalance.org/api/v1/grades/top?limit=10
# The Heartbeat pulse
curl https://api.thehibalance.org/api/v1/heartbeat/pulse
# Balanced Board members
curl https://api.thehibalance.org/api/v1/grades/top?balanced=true32 endpoints. Full reference: thehibalance.org/api
Each response includes score_status: verified, estimated, or pending — and a harm_documentation object when applicable. Every number traceable to its source.
┌─ Data Collection (nightly, 180min budget) ─────────────────┐
│ SEC · EPA · BLS · CFPB · FEC · FDA · FTC · CPSC · OSHA │
│ DOL · EEOC · NHTSA · USPTO · FMP · Finnhub · Yahoo │
│ NewsAPI · CDP · HRC · Disability:IN · SBTi · GRI · 990 │
└────────────────────────────┬───────────────────────────────┘
│
┌─ Scoring Engine (deterministic, no AI) ────────────────────┐
│ 19 active sub-signals → 5 dimensions → 1 composite (0-100)│
│ + 4 harm detection systems applied at dimension level │
│ + 3-layer validation (input · output · MSSI) │
└────────────────────────────┬───────────────────────────────┘
│
┌───────────────────┼───────────────────┐
▼ ▼ ▼
REST API (32) Chrome Extension iOS App
api.thehibalance 700+ brands App Store
Nightly pipeline: GitHub Actions at midnight CST. Full pipeline runs in ~60-90 min; skip-collect quarterly re-score in ~5 min. Validation stops publication if anything looks wrong.
3-layer validation:
- Input validation — no negative headcounts, no impossible Glassdoor ratings
- Output validation — score stability checks, distribution shape, known leaders
- MSSI — Maximum Single-Source Impact — no one source moves a sub-signal more than 15 points
git clone https://github.com/thehibalance/hi.git
cd hi/pipeline
pip install -r requirements.txt
# Full pipeline (nightly run equivalent, ~60min)
python3 run_all.py
# Re-score from cached data (~5min)
python3 run_all.py --skip-collect
# Quarterly threshold recalculation
python3 run_all.py --quarterly
# Start API locally (port 8080)
python3 api_server.py --port 8080See CONTRIBUTING.md for how to add a company, add a data source, or grind a sub-signal ladder.
| Framework | Misses |
|---|---|
| ESG (MSCI, Sustainalytics) | No AI displacement. No empathy detection. No humanwashing. Uses LLMs in scoring. Pay-to-play. |
| B Corp | No scoring granularity. No AI dimension. Self-reported. No real-time monitoring. |
| Fair Trade | Certifications only. No technology dimension. |
| Carbon Tools (CDP, Watershed) | Environmental only. |
| Credit Ratings (Moody's, S&P) | Financial only. |
| HI Grade | All five HUMAN dimensions. Zero AI. Auditable. Open source. |
We publish our methodology. We publish our limitations. Every sub-signal is auditable. Every score is reconstructable. No AI. No pay-to-play.
| Capability | Status |
|---|---|
| Chrome Extension | ✅ Shipped |
| iOS App | ✅ Shipped |
| Balanced Board methodology | ✅ Shipped |
| Harm Documentation (14 categories) | ✅ Shipped |
| 180-min quarterly workflow | ✅ Shipped |
| Safari extension | Planned |
| Firefox / Edge extensions | Planned |
| State of Human Intelligence report | Planned |
| EU CSRD + Companies House integration | Planned |
| Sub-signals H.4, U.5, N.1, N.3, N.4 | v1.3 target |
| Subsidiary Transparency Rule (SEC Exhibit 21) | v1.3 target |
| Asset | Status |
|---|---|
| Provisional Patent (22 claims) | Filed — expires March 2027 |
| HI Grade™ | Trademark filed |
| Humanwashing™ | Common law |
| Algorithmic Harm Index™ | Common law |
| "Human kind?"™ | Trademark pending |
| HUMAN Grade Spec v1.2.0 whitepaper | Copyright registered |
| Methodology + scoring engine | Apache 2.0 (see LICENSE) |
Pull requests welcome. Issues especially welcome — we want to hear when a score looks wrong. See CONTRIBUTING.md.
Three kinds of contributions we specifically need:
- Score challenges — you think company X should be higher/lower? Open an issue with evidence, we investigate.
- Data source additions — proposing a new public, free, auditable data source? PR with docstring + pipeline integration.
- Ladder grounding — got an academic paper or regulatory framework that could replace one of our editorial thresholds? Gold.
Built on public data from the SEC, EPA, BLS, CFPB, FEC, FDA, FTC, CPSC, OSHA, DOL, EEOC, NHTSA, USPTO. These agencies make HI Grade possible. Government transparency is a public good.
Built on methodology in part from CDP, GRI, SBTi, B Corp, Fair Trade, HRC, Disability:IN. Their work predates ours and informs ours.
Methodology, scoring engine, API server, and pipeline code: Apache 2.0 — see LICENSE.
Chrome extension and iOS app: source Apache 2.0. Distributed binaries are subject to the Chrome Web Store Developer Agreement and Apple Developer Program License Agreement, respectively.
Trademarks (HI Grade™, Humanwashing™, Algorithmic Harm Index™, Human kind?™) are property of Morf Innovations LLC and not covered by the Apache 2.0 license. See NOTICE for usage guidelines.
Human kind?
thehibalance.org · Chrome Extension · iOS App · API · Methodology · Sources
The HI Balance™ · Patent Pending · HI Grade™ · Humanwashing™ · Algorithmic Harm Index™ Morf Innovations LLC · @thehibalance · hi@thehibalance.org