# Design Ethics And Anti Dark Patterns > Use when auditing or redesigning consent, urgency, pricing, subscription, checkout, or retention flows for deception and coercion. Unlike product-design-audit, this is the focused ethics gate; conversion optimisation remains with its surface skill. - Skill: `peterbamuhigire/design-ethics-and-anti-dark-patterns` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add peterbamuhigire/design-ethics-and-anti-dark-patterns` - Raw SKILL.md: https://api.skillmd.com/api/skills/peterbamuhigire/design-ethics-and-anti-dark-patterns/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Web & Frontend - Author: peterbamuhigire (https://skillmd.com/u/peterbamuhigire) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/peterbamuhigire/design-ethics-and-anti-dark-patterns --- # Design Ethics & Anti-Dark-Patterns Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com. > A dark pattern (the regulators now say "deceptive pattern") is an interface built to make the user > do something they would not have chosen with full information and a clear head. It is the exact > inverse of the doctrine's Mission: where authored design is the signature of a skilled human acting > *for* the user, a dark pattern is design turned *against* them. It works once and is then resented — > and increasingly it is illegal. This skill is the cross-cutting ethics gate: the catalogue of the > coercive patterns, the honest swap for each, and the consent/default rules that keep a flow on the > right side of the line. Every conversion skill in the engine defers here. ## Use When - You are running the **ethics / dark-pattern pass over a whole flow or product** — onboarding, sign-up, checkout, cancellation, a permissions/consent dialog, a paywall — and need the named pattern families to detect against and the honest alternative for each. - A stakeholder (or your own instinct) is **reaching for a coercive lever** — a resetting countdown, a pre-ticked add-on, a "are you sure you want to miss out?" decline button, a cancel flow designed to be hard — and you need to **refuse it and offer the honest equivalent** that still converts. - You are designing **consent UX, defaults, or an opt-in/opt-out** and need the rule for what an ethical default and a genuine (freely-given, specific, informed, unambiguous) consent looks like. - A conversion/retail/trust skill has **deferred its cross-cutting ethics question here** (e.g. `ecommerce-and-checkout-ux`, `trust-credibility-and-social-proof`) and you need the general catalogue rather than the surface-specific rule. - You want the **"would you be comfortable if the user saw this?" test** applied to a design decision, with the light legal-exposure note (GDPR, CCPA/CPRA, EU/UK consumer law, FTC) attached. ## Do Not Use When - You are designing the **specific retail buying journey** (PLP → PDP → cart → checkout → payment → post-purchase) — use `06-sector-and-domain-ux/ecommerce-and-checkout-ux`. That skill *applies* this catalogue to the retail surfaces (sneak-into-basket, drip pricing, forced continuity in a cart); this one is the general doctrine it defers to. - You are designing the **trust / social-proof section** and the question is fabricated vs. authentic proof — use `14-conversion-and-web-page-patterns/trust-credibility-and-social-proof`. It forbids proof-specific fakery (anonymous "★★★★★", borrowed logos, fake review counts) and defers the cross-cutting scarcity rule here. - You need a **full scored, skill-routed product audit** across surfaces and platforms — use `00-cross-cutting-ops-qa-a11y/product-design-audit`; it runs this ethics lens as one dimension. - The behavioural detection you need is a **usability anti-pattern, not an ethics one** (mystery-meat nav, idiot-box modals, dead ends) — that is `doctrine/references/interaction-anti-patterns.md` directly. Some entries overlap (B1 Forced Sign-Up, C4 Trap); this skill adds the *intent-to-deceive* framing and the honest swap. ## Required Inputs | Input | Supplied by | Required? | Why | |---|---|---|---| | Complete user flow and all states | Product team | yes | Reveals hidden and delayed coercion | | Commercial objective and user promise | Owner or brief | yes | Exposes incentive conflicts | | Applicable jurisdiction and policy | Legal/compliance owner | conditional | Frames legal escalation, not legal advice | - **The flow and its goal, stated honestly.** What is the business outcome (a sale, a sign-up, a retained subscriber) *and* what is the user actually trying to do? A dark pattern is exactly the gap between those two being closed by coercion instead of value. Name both. - **The decisions the user makes in the flow** — every opt-in, default, consent, charge, and exit. Each is a place a dark pattern can hide; you cannot audit what you have not enumerated. - **The real facts behind any urgency, scarcity, price, or social signal** — is the stock count true and live? Is the deadline genuine? Is the "47 people viewing" measured or invented? Honesty is judged against the facts, so the facts are an input. - **The cancellation / opt-out / data-deletion path** — symmetric exit is the single most-tested guardrail (and the most-regulated). If you cannot describe how the user *leaves*, that is the first finding. ## The deceptive-pattern families (detect → why-harmful → honest swap) Full catalogue, with the regulator names and legal angle, in `references/dark-pattern-catalog.md`. The seven families to walk a flow against: 1. **Sneaking** — slipping something past the user. *Sneak-into-basket* (an item/warranty/sample added unasked), *hidden costs / drip pricing* (fees, tax, shipping sprung only at the final step), *hidden subscription* (a one-off purchase that is quietly recurring). **Honest swap:** the basket holds only what was explicitly chosen; the **all-in total is shown before the commit point**; recurring terms are stated at opt-in. (Ref `interaction-anti-patterns.md`; doctrine "Hidden costs".) 2. **Urgency** — manufacturing time pressure. *Fake scarcity* ("only 2 left" that resets on reload), *fake / resetting countdowns*, invented "high demand." **Honest swap:** show urgency **only when it is true and live** — real stock, a real deadline. Real scarcity is a legitimate, helpful signal; invented scarcity is the banned one. 3. **Misdirection** — steering the eye/instinct away from the user's interest. *Confirmshaming* (a decline worded to shame: "No thanks, I like paying full price"), *visual interference* (the wanted choice de-emphasised, the upsell styled as the only button), *trick wording / double negatives*. **Honest swap:** the user's likely choice is **at least as easy and prominent** as the business's preferred one; decline copy is **neutral** ("No thanks"); no double negatives. (Ref B3, D6.) 4. **Forced action** — demanding something unrelated to proceed. *Forced continuity* (a free trial that auto-charges with a buried cancel), *roach motel* (easy in, deliberately hard out), *forced sign-up* (a registration wall before any value). **Honest swap:** **symmetric exit** — cancel is as easy as sign-up; trials announce the charge up front and remind before billing; value before auth, guest mode offered. (Ref B1, C4; doctrine "Forced continuity / Roach motel".) 5. **Nagging** — repeated, persistent interruption to wear the user down. The notification that re-asks every session, the "rate us" that won't take no, the cookie banner that re-prompts. **Honest swap:** ask **once**, honour the answer, make "no" stick; an interval before any re-ask, and a real "don't ask again." 6. **Obstruction** — making a path artificially hard to discourage it. A cancel buried behind a phone call, a maze of confirmations to delete an account, comparison/price info hidden. **Honest swap:** the friction to leave/decline/delete is **no greater** than the friction to join/accept/buy. 7. **Social-proof & interface deception** — *disguised ads* (sponsored styled as organic), *fake activity/testimonials*, *bait-and-switch*. **Honest swap:** label sponsored content; use only real, attributed proof (defer to `trust-credibility-and-social-proof`); the control does exactly and only what its label says. ## Workflow 1. **State the flow's honest goal and enumerate the user's decision points.** Write the business outcome and the user's actual job side by side; list every opt-in, default, consent, charge, and exit in the flow (Required Inputs). This is the surface the audit walks. 2. **Walk each decision point against the seven families.** For every opt-in/default/charge/exit, ask which family could hide there and check the facts (`references/dark-pattern-catalog.md`). Any match is a finding: record the pattern name, where it sits, the user harm, and the legal exposure. 3. **Apply the consent & default rules.** Every default must be the one an informed user would *choose*, not the one that maximises revenue by inertia. Consent must be **freely given, specific, informed, unambiguous, and as easy to withdraw as to give** (the GDPR test, which is also just good design). Pre-ticked boxes are never valid consent. See `references/dark-pattern-catalog.md` §Consent. 4. **Apply the comfort test to anything ambiguous.** For each questionable choice, ask: *"Would I be comfortable if the user could see exactly why this was designed this way — would I explain it to their face?"* If the honest explanation is "to trick you into not noticing / not leaving," it is a dark pattern. Reframe it as the honest swap. 5. **Swap every finding for its honest equivalent.** Replace each detected pattern with the swap from the catalogue (transparency, symmetric exit, neutral decline copy, true scarcity, opt-in defaults). The honest version is the deliverable — never ship the coercive one and "flag it." 6. **Attach the legal note where relevant — lightly.** Where a finding maps to a named rule (GDPR consent / dark-pattern guidance, CCPA/CPRA opt-out and the "symmetry" rule, the EU Digital Services Act and UCPD, UK CMA/DMCCA, FTC "click-to-cancel" / negative-option), note it as *additional* exposure, not the primary reason — the primary reason is that it is wrong. See the catalogue's §Legal angle. Do not give definitive legal advice; flag and recommend counsel for live launches. 7. **Produce the ethics audit.** Write the findings → honest-redesign report (shape: `examples/ethics-audit-filled.md`): each pattern found, its family, evidence, harm, legal note, and the honest swap, ending with the consent/default verdict and the comfort-test summary. ## Anti-Patterns - **Shipping the coercive version with a comment.** Detecting a dark pattern and leaving it in "because the team wants the numbers" — the swap *is* the job. If a stakeholder insists, refuse and state the honest alternative that still converts. - **"Asymmetry of effort" in disguise.** A flow where joining is one tap and leaving is ten — even if no single screen is egregious, the *asymmetry itself* is the obstruction/roach-motel pattern. - **Treating real scarcity as banned.** True, live "only 2 left" or a genuine sale deadline is honest and helpful — the ban is on the *invented* and *resetting* kind. Don't strip truthful urgency. - **Consent theatre.** A cookie/permission dialog where "Accept all" is one button and "reject" is buried three clicks deep — that is misdirection + obstruction, not consent, regardless of the legal box it appears to tick. - **Confusing this with the usability checklist.** `interaction-anti-patterns.md` detects *incompetent* UX; this detects *deceptive* UX (intent to mislead). Route accordingly; some entries pair. - **Citing the law as the only reason.** "We can't because GDPR" trains teams to do the minimum legal. Lead with the user harm and the Mission; the legal note is reinforcement, not the argument. ## Outputs | Output | Consumer | Evidence / acceptance | |---|---|---| | Ethics finding register | Product and legal owners | Pattern, location, harm, evidence, and severity | | Honest-swap specification | Designer and engineer | Equivalent non-coercive flow and copy | | Ethics verdict | Release owner | PASS, CONDITIONAL, or BLOCKED with owner | - An **ethics / dark-pattern audit** of the flow: every detected pattern by family, with location, evidence, user harm, the light legal note, and the **honest swap** — in the shape of `examples/ethics-audit-filled.md`. - A **consent & default verdict** confirming every default is the informed-user choice and every consent is freely-given/specific/informed/unambiguous/withdrawable. - A **completed comfort-test summary** — the questionable decisions, each resolved to its honest form. ## Examples - Use the worked example in `examples/` for the pattern-to-harm-to-honest-swap trace. ## Quality Standards - Evaluate the complete journey, including cancellation, refusal, renewal, and post-purchase states. - Preserve informed, freely given, specific, and reversible user choice. - Stop release for hidden costs, fabricated urgency, non-equivalent refusal, or obstructed cancellation. ## Decision Rules | Condition | Decision | Wrong-choice failure | |---|---|---| | Claim or urgency is objectively verifiable and expires truthfully | Permit with evidence | Legitimate information is suppressed | | Decline is less visible or harder than accept | Equalise prominence and effort | Consent is coerced | | Optional data is required without necessity | Remove requirement or state dependency | Forced action captures excess data | | Legal interpretation is needed | Escalate to qualified counsel | Design guidance is misrepresented as legal advice | ## Capability Contract Read and complete-flow inspection are required. Audit defaults to read-only; editing needs remediation authority. Do not approve legal compliance, manipulate production data, or publish consent changes without accountable owner review. ## Degraded Mode Without all states or policy context, issue a conditional ethics review, identify missing screens and claims, and withhold a pass. Recover with recordings, state maps, terms, and legal-owner input. - `examples/ethics-audit-filled.md` — a fully worked ethics audit of a real-ish flow (the sign-up → trial → cancellation journey of a habit-tracking SaaS, "Streaky"), finding the dark patterns across all seven families (drip-priced "pro" fee, pre-ticked annual-plan upsell, confirmshaming decline, forced-continuity trial with a roach-motel cancel, nagging notification opt-in, a disguised "recommended" ad), then the honest redesign of each, the consent/default verdict, the comfort-test pass, and the light legal note (GDPR consent, FTC click-to-cancel, EU UCPD). Concrete, no lorem. ## References - `references/dark-pattern-catalog.md` — the full catalogue: each deceptive pattern (by the seven families) → why it harms the user → its honest swap, plus the consent-UX rules, the default-ethics rule, the "would you be comfortable if the user saw this?" test, and the §Legal angle (GDPR, CCPA/ CPRA, EU DSA/UCPD, UK CMA/DMCCA, FTC negative-option) kept deliberately light. - `doctrine/design-doctrine.md` — the Mission (§0): authored design acts *for* the user; a dark pattern is design turned against them and reads as exactly that once noticed. - `doctrine/references/interaction-anti-patterns.md` — the behavioural detection checklist this skill shares a border with: **B1 Forced Sign-Up**, **B3 Vague/competing CTAs**, **C3 Dead End**, **C4 Trap / Forced Action**, **D6 Oceans of Buttons** — the usability half; this skill adds the intent-to- deceive half and the honest swap. - Sibling skills: `06-sector-and-domain-ux/ecommerce-and-checkout-ux` and `14-conversion-and-web-page-patterns/trust-credibility-and-social-proof` (both **pair** with this skill and name it as the cross-cutting ethics catalog they defer to); `00-cross-cutting-ops-qa-a11y/product-design-audit` (runs this ethics lens as one scored dimension of a full product review).