- Ethics
- Methodology
BET Index Methodology
The Baldwinson Ethical Technologies (BET) Index evaluates providers using two principles: (1) present-day policies and conduct and (2) an "institutional memory" model that does not forgive unresolved harm.
Scope
What We Rank
The BET Index (Infrastructure Edition) evaluates Infrastructure as a Service (IaaS) and major hosting providers, including hyperscalers and independent infrastructure vendors.
Included examples: AWS, Azure, Google Cloud, IBM Cloud, Oracle Cloud, Hetzner, OVHcloud, DigitalOcean, Vultr, etc.
What We Do Not Rank (In This Edition)
Consumer ecosystem clouds (e.g., iCloud) are not directly comparable to IaaS/PaaS providers. These may be covered in a separate BET Consumer Cloud Index.
Philosophy
Most "inclusion" indices measure policy checkboxes and forget history on schedule. BET is built on two premises:
- Inclusion without worker dignity is performative.
- Institutions must carry moral memory until harm is repaired.
Offenses do not "fade" simply because time passes or leadership changes.
Outputs
For each provider, BET publishes:
- Base Category Scores (0–100) across five pillars
- Ethical Ledger (Event Log) with sources
- Moral Inertia Adjustments (growth while unresolved; decay only after remediation)
- Final Score (0–100) and rank
- Historical Moral Record (Non-Scoring, Permanent) for major historical complicity events
Pillars and Weights
BET uses five pillars. Weights can vary by edition; the recommended default is:
| Pillar | Weight |
|---|---|
| Leadership & Political Integrity | 35% |
| Worker Dignity & Labor Record | 35% |
| Policy & Benefits | 10% |
| ERG Depth & Transparency | 10% |
| External Advocacy | 10% |
Rationale: Leadership and labor reflect what cannot be faked. Policies and ERGs matter, but are secondary.
Pillar Definitions and Scoring
Each pillar is scored 0–100 before Moral Inertia adjustments.
1. Leadership & Political Integrity (35%)
What leaders do and endorse publicly, including political activity, governance posture, and moral direction set at the top.
Scoring Inputs
- Executive public statements and policy positions (CEO/Board/Top leadership)
- Political donations and lobbying patterns (where verifiable)
- Consistency over time (not one-off press releases)
- Governance actions (DEI rollbacks, censorship, retaliation against speech)
Scoring Guide (Illustrative)
- 90–100: Consistent pro-rights alignment; transparency; no major contradictions
- 70–89: Mostly aligned but mixed signals; partial transparency
- 40–69: Repeated contradictions; donations/lobbying misaligned with stated values
- 0–39: Severe contradictions; support for anti-rights agendas; repeated integrity failures
2. Worker Dignity & Labor Record (35%)
How workers are treated in practice, including labor relations, retaliation, safety, pay fairness, and lived experience.
Scoring Inputs
- Safety track record and credible injury reporting
- Anti-retaliation posture; treatment of organizers/whistleblowers
- Contractor and supplier labor standards (where relevant)
- Regulatory actions, lawsuits, settlements (weighted by credibility and severity)
- Accessibility and accommodation practices (including for trans and disabled workers)
Scoring Guide
- 90–100: Strong safety record + worker protections; credible oversight
- 70–89: Generally acceptable with limited concerns
- 40–69: Recurring labor disputes, safety concerns, retaliation allegations
- 0–39: Systemic harm patterns, major controversies, repeated violations
3. Policy & Benefits (10%)
Measures written commitments and formally provided benefits that reduce discrimination and enable equitable participation, independent of enforcement quality.
Scoring Inputs
- Explicit non-discrimination policies covering sexual orientation, gender identity, and expression
- Inclusive healthcare benefits, including partner parity and transgender healthcare where legal
- Family formation and parental benefits evaluated through an access and non-discrimination lens
- Recognition of domestic partners or non-traditional dependents where applicable
- Public clarity and specificity of benefits documentation
This pillar evaluates written commitments and benefits availability only; enforcement and lived experience are assessed elsewhere.
4. ERG Depth & Transparency (10%)
Whether internal representative structures exist (when organizational scale permits) and meaningfully function, including their independence and influence.
Scoring Inputs
- Existence of LGBTQ+ employee resource groups
- Scale, geographic reach, and leadership sponsorship of ERGs
- Structural independence of ERGs from corporate communications or HR
- Evidence that ERGs influence policy, benefits, or workplace protections
ERG-related criteria are evaluated where organizational size and geographic distribution reasonably support representative employee structures (generally organizations with sufficient headcount to sustain independent ERGs).
5. External Advocacy (10%)
Evaluates whether a company’s public actions or material support measurably affect human rights outcomes beyond its own workforce.
Scoring Inputs
- Documented public positions opposing laws or policies that directly restrict LGBTQ+ rights, where such positions are taken
- Financial or in-kind support for human rights organizations, evaluated for transparency and consistency (where applicable)
- Participation in industry coalitions or standards bodies addressing discrimination or civil rights impacts
- For infrastructure providers: policies or actions addressing misuse of technology for mass surveillance, repression, or authoritarian control
Base Score Calculation
Compute the Base Score as:
Example (with default weights):
Ethical Ledger: Events and Evidence
BET maintains a public Ethical Ledger of documented events for each provider.
What Qualifies as a Ledger Event
An event must meet at least one:
- Regulatory action, court filing, settlement, or official investigation
- Multiple credible independent journalistic sources
- Credible NGO reports
- Verified company documents / statements
- Consistent pattern corroborated across sources
Event Record Fields
- Provider name
- Event start date
- Category (leadership, labor, policy, advocacy, etc.)
- Severity level
- Description
- Sources (links)
- Status: ongoing / remediated
- Remediation date (if remediated)
- Notes on verification
Severity Model (Penalty Points)
BET assigns a severity points value (0–100 scale points) per event:
| Severity | Points |
|---|---|
| Minor | 5 |
| Moderate | 15 |
| Severe | 25 |
| Egregious | 40 |
Severity is determined by: scale of harm, intent vs negligence, repetition/pattern, official findings vs allegations, and number of affected people.
Moral Inertia Module
This is the core differentiator: events do not decay until remediated.
Key Principle
Time does not forgive unresolved harm.
Penalties either grow (if unremediated) or decay (only after remediation).
Parameters (Published Each Edition)
- Grace window: 0.25–0.5 years (investigation window before growth begins)
- Annual growth rate (g): typically 10–20% depending on category
- Penalty cap multiplier: typically 3× (can vary by severity)
- Post-remediation half-life: typically 10–12 years (slow forgiveness)
- Mitigation factor (m): 0–1, applied at remediation (partial fixes)
Penalty While Unresolved (Growth)
For an event i:
- severity points = P
- time since start (years) = t
- grace window = w
- annual growth rate = g
- cap multiplier = C
Unremediated penalty:
If t ≤ w: penalty = P
If t > w: penalty = min( P × (1+g)(t−w), P × C )
Meaning: the longer a company delays fixing the issue, the more it hurts.
Penalty After Remediation (Decay)
If remediation occurs at time R, we compute the penalty at remediation, apply mitigation, then decay.
- half-life (years) = H
- λ = ln(2)/H
- time since remediation = u
Remediated penalty:
penalty_at_fix = min( P × (1+g)(R−w), P×C )
post_fix_base = penalty_at_fix × m
penalty_now = post_fix_base × e(−λu)
Meaning: penalties only fade after a verified fix—and even then, slowly.
Final Score Calculation
Final score is bounded to [0, 100].
Remediation Standards
An event is marked remediated only if there is evidence of durable correction.
Acceptable Remediation Evidence
- Independent third-party audit
- Regulator confirms compliance improvements
- Policy change + enforcement + monitoring data
- Worker restitution or structural reform (where applicable)
Non-Remediation Examples (Insufficient)
- PR statements without structural change
- Leadership swap with no policy change
- "We're investigating" without results
- Internal memo without measurable outcomes
Partial remediation is allowed via mitigation factor m (e.g., m=0.7).
Anti-Gaming Rules
To prevent "CEO swap" or PR games:
- No decay without remediation.
- Penalty growth continues until fix is verified.
- Remediation requires external evidence.
- Repeat offenses stack and can escalate severity.
- Pattern multiplier: repeated similar offenses in a rolling window may amplify penalty.
Historical Moral Record
BET publishes a separate permanent record for severe historical complicity events.
These entries:
- Do not affect present numeric scores
- Do not decay
- Exist to prevent historical erasure
- May become scoring-relevant if denial, glorification, or repeat patterns emerge
Transparency and Governance
Right of Reply
Before publication, providers may submit:
- Corrections
- Remediation documentation
- Context and evidence
BET publishes accepted corrections and disputes with rationale for inclusion/exclusion.
Conflicts of Interest
BET discloses funding sources (if any), any relationships with rated companies, and advisory board affiliations.
Versioning
Each release is versioned with: Methodology vX.Y, Ledger snapshot date, and weight changes explicitly documented.