PRIVACY & ETHICS — Main Statement (Draft v0.1)
Playful rigor: build → perform → document → iterate. People first, always.
Maintainer: Ben (maker, artist, STEAM educator)
Status: Working draft
Last updated: 2025-09-16
Canonical source:
PRIVACY_ETHICS.md
Last synced: 2025-09-16
0. Plain-Language Summary (TL;DR)
- Dignity over data. I only collect what’s needed to make a thing work, and only with explicit, informed consent.
- Local first, deletion by default. Process on-device when possible. Store as little as possible. “Delete now” is always one click away.
- Open by design, not exploitative. Code and documentation prefer free/libre licenses. Sharing knowledge is not a license to extract people.
- Care for minors and communities. Extra protections for youth. Harm prevention beats novelty.
- Transparent practice. I publish an assumption ledger, data-flows, known risks, and change logs.
- Accountable and reversible. Consent can be revoked, data can be erased, decisions can be appealed.
1. North Star
I build tools, scenes, and learning environments where other people create. My ethics follow that purpose: reduce harm, increase agency, share knowledge, and design systems that support equitable participation. I align with the free software movement—free as in freedom to study, modify, share, and use—and I operate in the spirit of Kent’s 10 rules as a craft ethos of clarity, humility, and responsibility in making.
2. Scope
This statement covers:
- Software, hardware, installations, workshops, and documentation published under this umbrella (e.g., Human-Buffer — Consent-Forward Vision Station (formerly faceTimes), MOARkNOBS-42, curricula, demos, exhibitions).
- All interactions with participants, students, collaborators, and audiences in these contexts.
Note: Specific projects carry addenda with precise data-flows, risks, and micro-copy. Project ethics never weaken any principle here; they may strengthen them.
3. Principles
- Human dignity is inviolable. People are not datasets. No biometric identification or surveillance by default. Detection-only when possible.
- Equity is a system requirement. I design for the margins first and audit for disparate impacts.
- Consent is an interface. Opt-in, affirmative, revocable, time-bounded, and contextual. Consent UX is plain-language and visible.
- Data minimization. If I don’t need it, I don’t collect it. If I must collect it, I keep it short-lived and secure.
- Local first. Prefer on-device processing with no network calls.
- Transparency and legibility. Publish assumption ledgers, data-flow diagrams, and change logs.
- Right to inspect, correct, and erase. Participants can see what exists about them, fix it, or remove it—fast.
- Open knowledge, not open people. I release code, methods, and documentation under free/libre licenses while protecting personal data.
- Accountability. Clear roles, audit trails, and incident response. When harm happens, I repair where possible and learn in public.
- Proportionality & necessity. Fancy isn’t a justification. Risk must be proportionate to purpose.
4. Definitions
- Personal Data: Any information relating to an identifiable person (including images, voiceprints, device identifiers).
- Sensitive Data: Data that could cause harm or discrimination if exposed (e.g., biometrics, health info, precise location, minors’ data).
- Processing: Any operation on data (capture, transform, store, transmit, display).
- On-device: Processing done locally on the user’s hardware without cloud transmission.
- Assumption Ledger: A living document listing design assumptions, uncertainties, mitigations, and evidence.
5. Data Practice
5.1 What I avoid
- Facial identification or recognition; emotion inference; demographic inference.
- Passive, undisclosed logging (no shadow telemetry).
- Cross-context tracking; ad tech; data broker sharing.
5.2 What I may collect (case-by-case)
- Ephemeral sensor data required for live functionality (e.g., audio levels, detection bounding boxes) retained in RAM only.
- Operational logs necessary for stability and safety—anonymized, minimized, and rotated quickly.
- Explicit submissions from participants (e.g., saved images, consent forms) stored only with clear time limits.
5.3 Data lifecycle (default)
| Stage | Where | Default TTL | Who can access | Notes |
|---|---|---|---|---|
| Capture | On-device (RAM) | transient | Participant & operator | Visible indicators (LED/UI) show capture state. |
| Process | On-device | transient | Operator | No cloud calls; no third-party inference. |
| Review | On-device | until decision | Participant | Buttons: Show me my data · Save · Delete now. |
| Save (opt-in) | Encrypted local storage | 30 days (project default) | Maintainer | TTL set in project addendum; user may set shorter. |
| Share (opt-in) | Controlled repository | case-specific | Maintainer & participant | License and purpose specified; revocable when feasible. |
| Erase | Local + backups | immediate | Maintainer | Erasure receipts logged; backups pruned. |
Deletion Controls: Every project includes an obvious “Delete now” control and a post-capture review screen. No auto-save of personal data.
6. Consent Architecture
- Opt-in by default. No implied consent. No dark patterns.
- Micro-copy accompanies every capture: purpose, retention, sharing, and how to revoke.
- Revocation process is simple and honored quickly; erasure confirmations are provided.
- Informed participation includes a plain-language brief, risks, and alternatives.
For youth (minors):
- Require guardian consent and youth assent.
- No public posting of minors’ personal data without explicit, project-specific consent.
- Blur/obfuscate by default in public artifacts; prefer avatars over headshots.
7. Safety, Fairness, and Testing
- Pre-launch risk review for each project (technical, social, legal, and educational risks).
- Bias checks on model behavior where ML is used; document limitations and failure modes.
- Red team exercises proportional to risk (abuse cases, spoofing, misuse scenarios).
- Accessibility is mandatory: readable contrast, captions, keyboard paths, and alt text.
- Environmental safety for hardware: no exposed mains, proper fusing, thermal and RF considerations, and signage.
8. Licensing & Knowledge Sharing
- Software: Prefer AGPL-3.0/GPL-3.0 or CERN-OHL-S/W for hardware where applicable.
- Documentation & Curriculum: Prefer CC BY-SA 4.0.
- Media Assets: Prefer CC BY-NC (or more permissive if participants agree).
- Contributor Terms: Contributions are accepted under the same licenses; contributors affirm they have rights to contribute; no CLA surprise grabs.
- Ethical Use Rider: Licenses are free/libre; I additionally publish norms discouraging use in surveillance, discrimination, or state/corporate harm—even when license text permits. I name those harms plainly.
9. Governance & Accountability
- Stewardship: Named maintainers for each project with contact info.
- Change control: All updates to this statement and project addenda are tracked in CHANGELOG.md with semantic versioning.
- Assumption ledger: Kept in repo; updated with experiments and findings.
- Incident response:
- Acknowledge within 72 hours.
- Contain and assess impact.
- Notify affected participants when relevant.
- Publish a post-mortem with fixes and timelines.
- Appeals & Remedies: Participants can request review of decisions (e.g., takedowns, denials). I prioritize repairing harm over defending reputation.
10. Exhibition & Classroom Norms (Community Covenant)
- Consent signage: Clear, multilingual, with QR to the project addendum and this statement.
- No harassment / no doxxing / no recording without consent.
- Right to be unrecorded. A “no capture” badge or zone is always available.
- Care teams: Identified adults/staff trained in de-escalation and data hygiene.
- Photography protocol: Ask first, explain purpose, provide deletion on request.
11. Project Addendum Template (per-project)
Each project includes a PRIVACY-ADDENDUM.md with:
- Purpose & benefits
- Data categories (exact fields)
- Data-flow diagram (camera ➜ RAM ➜ review ➜ save/erase)
- Risk register (with likelihood × impact × mitigations)
- Consent text (micro-copy as shown to participants)
- Retention policy (TTL & erasure path)
- Accessibility notes
- Contact & appeals
12. Known Limits & Commitments
- I cannot guarantee absolute security, but I commit to minimization, encryption, and rapid response.
- I will sometimes choose less capable technology to reduce risk.
- When constraints force compromise, I will name the trade-offs and invite critique.
13. Contact
- Primary: severns3@gmail.com
- Security/Erasure Requests: same for the time being
14. Change Log (excerpt)
- v0.1 (2025-09-16): Initial draft establishing core principles, consent architecture, and per-project addendum template.
Appendix A: Values Cheat-Sheet (for posters/UI)
- People first.
- Opt-in, informed, reversible.
- Local processing; minimal data.
- Clear purpose; short retention.
- Show • Save • Delete—always visible.
- Open code & docs; closed doors on personal data.
- Document assumptions; publish changes.
- Extra care for youth.
- Accessibility is a feature, not a toggle.
- Repair harm; learn in public.