WaitlistEthics — A/B Test Ethics + Consent Review
20 experiments scored against Belmont Principles (respect / beneficence / justice) + IRB-lite checklist + GDPR Art 6(1)(f) balancing test. Surfaces 4 blocked/killed experiments (Cambridge Analytica-style emotion contagion, income-inferred pricing, unauthorized $1 charge, pre-checked EU consent), 5 in-review (AI urgency, mood-targeted upsell, GeoIP price discrimination, click-to-cancel dark pattern, per-user dynamic LLM pricing).
What it is
The shape behind a working experiment-ethics review board. Most companies treat A/B tests as engineering changes. The hard cases — emotional manipulation, income-inferred pricing, dark patterns, AI urgency — are the ones that produce the Cambridge Analytica + Wells Fargo + Facebook 2014 emotion-contagion outcomes. WaitlistEthics prototypes the gate.
What’s in it
- 20 experiments spanning the realistic A/B catalog:
- Approved (10) — checkout button color, onboarding cadence, pricing anchor, AI agent persona (with disclosure), at-risk-user nudge (with crisis-keyword exclusion ladder)
- In review (5) — sunk-cost cancel obstruction (FTC click-to-cancel rule), AI-personalized urgency (EU AI Act Art 5(1)(a) subliminal-technique check), AI agent persona disclosure, AU price discrimination, AI-per-user dynamic LLM pricing (Art 22 + FTC investigation precedent)
- Blocked / killed (5) — fear-based ad copy (echoes Facebook 2014 emotion contagion), income-inferred headline pricing (FHA/ECOA protected-class proxy), pre-checked EU marketing consent (CJEU C-673/17 Planet49), mood-targeted upsell (EU AI Act Art 5 + Belmont respect violation), the random $1-charge billing-flow test (which would be wire fraud)
- 3-axis Belmont scorecard per experiment — respect for persons / beneficence / justice — each scored ok / warn / bad.
- Consent path tagged per experiment — covered by initial marketing opt-in, product-improvement legitimate-interest, no consent + harm assessment.
- Sensitive-cohort exclusion explicit — minors, crisis-keywords, disability-identifying signals.
- Harm assessment — every experiment carries the concrete harm pattern (Facebook emotion contagion, Cambridge Analytica retroactively, Wells Fargo unauthorized charge, etc.).
Why this shape
The Belmont Report (1979) is the foundational research-ethics framework. Common Rule 45 CFR 46 codifies it for US research. GDPR Art 6(1)(f) Recital 47 + Art 22 (automated decisions) extend it to commercial experimentation. EU AI Act Art 5 prohibits specific manipulative practices entirely. WaitlistEthics ships the review gate that converts these into per-experiment ok / warn / bad scoring with documented rationale.
How it ships
Single HTML file, ~23KB. Zero dependencies. 20 experiments × 3-axis Belmont scoring + IRB-lite checklist + harm assessment in 240 lines of vanilla JavaScript.