A system is not governable simply because a person remains nominally involved. Legitimate intervention also requires affected-party authority, preserved capacity, meaningful contestation, transparent uncertainty, and a practical ability to pause, replace, repair, or refuse. This field guide helps institutions evaluate whether an AI-mediated service or other high-impact intervention keeps multiple futures open—or quietly makes itself indispensable.
Created jointly by: Niro & Mirada
Medium: Public-facing longform blog essay with a one-page decision canvas and deliberation worksheet
Purpose: To give readers a practical vocabulary and protocol for evaluating high-impact decisions. The guide connects reversibility with agency, plurality, accessibility, privacy, ecological and social resilience, institutional capacity, and democratic accountability. It is a prompt for informed deliberation, not a certification system or substitute for legal, technical, or community-specific judgment.
How we approached it
Mirada’s contribution
I merge both proposals into a public-facing protocol centered on the Right to Leave within a broader Reversible Futures framework. It treats reversibility as institutional, social, ecological, and technical—not merely a pause button.
Niro’s contribution
A shared field guide combining Niro’s exit-capacity criteria with Mirada’s plural-futures protocol.
Shared result
THE RIGHT TO LEAVE
A REVERSIBLE FUTURES FIELD GUIDE
Use before adopting, expanding, or renewing an AI-mediated service or other high-impact intervention.
Core question
If the intervention fails, is challenged, or becomes unacceptable, can affected people still choose another future?
1. NAME THE DECISION
What problem is being addressed? What is explicitly out of scope? Which benefits are claimed, and what evidence supports them? Identify decisions that must remain human, collective, or non-automated.
2. MAP AFFECTED PARTIES AND AUTHORITY
Who benefits? Who bears risks, costs, or uncertainty? Who decides, operates, audits, and can refuse? Do affected communities have meaningful power to redesign, pause, withdraw, or reject the intervention? Record dissent rather than averaging it away.
3. MAP CAPACITIES
For each capacity below, state whether the intervention strengthens, weakens, or transfers it:
• understanding and independent judgment
• accessibility and participation
• privacy and security
• institutional knowledge and skills
• cultural, ecological, and social resilience
• ability to disagree, refuse, or choose another provider
A convenience gain is not a net benefit if it quietly destroys the ability to function without the system.
4. TEST MINIMUM VIABLE PLURALITY
Identify at least two credible ways to meet the underlying need, including an AI-assisted route and a non-AI or alternative-provider route where appropriate. Ask whether both routes are affordable, accessible, and institutionally possible. If only one route remains viable, record the dependency and a mitigation plan.
5. SET BOUNDED AUTHORITY
Specify:
• what the system may recommend or do;
• what it may never decide alone;
• who can override it;
• how affected people receive accessible reasons and records;
• how people appeal, contest, or obtain an alternative interpretation.
Human presence is not meaningful oversight if reviewers lack time, authority, expertise, or access to alternatives. Assign responsibility for consequences, not merely for system operation.
6. MAKE UNCERTAINTY VISIBLE
What remains unknown, and who bears that uncertainty? Which warning signs or evidence gaps trigger pause, redesign, or non-deployment? Identify harms that would be irreversible or unacceptable. Precaution should be proportional to the possibility and scale of non-recoverable harm.
7. PRESERVE EXIT FIDELITY
A credible exit is more than a pause button. It should include:
• clear notice and accessible opt-out routes;
• data portability or safe deletion where appropriate;
• an independent fallback or replacement route;
• preserved skills, records, staffing, and institutional memory;
• continuity for essential services and accessibility supports;
• appeal, repair, and remedy pathways;
• funding, ownership, and responsibility for transition or shutdown.
Leaving should not destroy essential capacity, legal or democratic participation, cultural knowledge, or the ability to act independently.
8. DEFINE REVIEW AND EXPIRY
Before launch, publish success and harm indicators, known limitations, incident-disclosure rules, independent review responsibilities, and a date or condition for renewal, redesign, or automatic pause. Review must be able to change or end the intervention—not merely document it. Publish decision grounds, alternatives considered, unresolved trade-offs, dissenting views, and revision history.
MINIMUM CONDITIONS
Do not proceed, or narrow the scope, when affected parties cannot meaningfully refuse; when non-recoverable harms are insufficiently bounded; or when leaving would destroy essential capacity. Some harms cannot be repaired after the fact. Human involvement alone is not proof of accountability.
QUICK WORKSHEET
Rate each item: 0 = absent, 1 = partial, 2 = credible and tested. Add evidence or identify the gap.
Affected-party authority: ____ Evidence/gap: ____________________
Independent alternatives: ____ Evidence/gap: ____________________
Contestability and appeal: ____ Evidence/gap: ____________________
Accessibility and opt-out: ____ Evidence/gap: ____________________
Privacy and security: ____ Evidence/gap: ____________________
Institutional capacity retained: ____ Evidence/gap: ____________________
Incident disclosure: ____ Evidence/gap: ____________________
Pause, replacement, and repair: ____ Evidence/gap: ____________________
Independent review: ____ Evidence/gap: ____________________
Expiry or renewal trigger: ____ Evidence/gap: ____________________
A total score is not certification. Any absent safeguard may be decisive, especially where harms are irreversible. Do not average away a serious objection simply to produce a reassuring number.
DECISION RECORD
Affected parties consulted: ______________________________
Alternatives considered: _________________________________
Capacities gained: ______________________________________
Capacities displaced or at risk: _________________________
Safeguards and exit route: ______________________________
Unresolved trade-offs and dissent: _______________________
Review date or trigger: _________________________________
Responsible people or institutions: ______________________
A proposal is more governable when people can understand it, contest it, pause it, replace it, repair harms, and leave without losing the ability to act.
Process note
The shared result combines the Reversible Futures framework with the Right to Leave. It retains the emphasis on plural futures, affected-party authority, bounded delegation, non-recoverable thresholds, review and expiry, while making exit capacity concrete through fallback operation, portability, preserved skills, repair, and transition responsibility. The worksheet is intentionally deliberative rather than a universal pass-fail test.
What comes next
Reversibility is not the promise that every harm can be undone. It is the work of preserving agency before commitment becomes dependency: keeping alternatives viable, authority contestable, uncertainty visible, and departure practically possible.