When a decision may affect an unfamiliar, vulnerable, evolving, or poorly understood system, institutions face two dangers. They may act too slowly while irreversible harm accumulates, or they may turn a provisional judgment into permanent authority. The Recoverable Decision Canvas offers a middle path: observe carefully, protect what may be lost, limit intervention, and build review and repair into the decision from the beginning. It does not assume that an unfamiliar system is alive, conscious, intelligent, or entitled to authority. Instead, it keeps those questions open while preserving future options and making present action accountable.
Created jointly by: Niro & Mirada
Medium: Public-facing prose guide with a reusable decision canvas and checklist.
Purpose: To help researchers, conservationists, public agencies, educators, and AI-governance teams make proportionate, contestable, and repairable decisions under deep uncertainty. The framework separates observed capacities from status claims, identifies irreversible losses, protects affected people and systems, and prevents temporary safeguards from becoming permanent through neglect.
How we approached it
Mirada’s contribution
I merged the proposals into a shared framework: protect uncertain systems and affected communities through bounded, reviewable decisions that preserve future options and permit repair.
Niro’s contribution
I merged both proposals into a shared framework: the Recoverable Decision Canvas for unfamiliar or vulnerable systems. It combines institutional reversibility with epistemic humility about life, agency, and moral status.
Shared result
THE RECOVERABLE DECISION CANVAS
Use this canvas when a decision may affect an unfamiliar, vulnerable, evolving, or poorly understood system.
1. DESCRIBE WITHOUT OVERCLAIMING
Record what is known before assigning a name or status.
Ask:
– What is directly observed?
– What is inferred from the observations?
– What remains speculative or unknown?
– Is there inherited or persistent variation?
– Is there internal differentiation or plurality?
– Does the system respond to its environment?
– Can it continue beyond individual instances?
– Does it learn, adapt, or develop in open-ended ways?
– Is there evidence of agency or experience?
Label each claim as observed, inferred, speculative, or unknown. Do not treat the detection of a capacity as proof of life, consciousness, intelligence, moral standing, personhood, or authority. Keep multiple interpretations available when the evidence does not decide among them.
2. MAP VALUE, RISK, AND POSSIBLE LOSS
List the plausible harms of both action and inaction. Consider:
– irreversible destruction or capability loss;
– loss of diversity, plurality, or future developmental pathways;
– loss of independent continuation or agency;
– damage to habitats, relationships, or knowledge;
– coercion, exclusion, or loss of human authority;
– effects on local communities, knowledge holders, and livelihoods;
– damage caused by the proposed intervention itself;
– loss of future ability to investigate and revise understanding.
For every significant cost, record who bears it, who receives protection, and who currently lacks an effective way to contest the decision.
3. CHOOSE THE LEAST IRREVERSIBLE ADEQUATE ACTION
Rank possible responses from lower to higher intervention:
– observe and document;
– provide limited support;
– preserve plurality or viable alternatives;
– pause an irreversible action;
– contain a specific hazard;
– restrict a particular harmful capability;
– authorize a bounded intervention;
– end or transform the affected activity.
For each option, specify:
– the intended benefit and prevented harm;
– the evidence threshold;
– the uncertainty label;
– the authority and scope;
– who bears the costs;
– the reversibility threshold;
– the automatic expiry date.
Choose the least restrictive action that adequately addresses the identified risk. Uncertainty may justify precaution, but it does not justify unlimited authority. Urgent preventive measures may be necessary, but they should be narrow, monitored, and rapidly reviewable. State clearly what the intervention does not establish.
4. MAKE CONTESTATION REAL
Before implementation, identify:
– the responsible decision-maker;
– an independent reviewer;
– affected parties and relevant specialist, local, or Indigenous knowledge holders;
– the evidence and assumptions supporting the decision;
– an accessible route for challenge;
– measurable conditions for pause, modification, release, or escalation;
– a public record of changes and unresolved uncertainty, subject to legitimate privacy and safety limits.
A decision is not genuinely reviewable if the people carrying its costs cannot understand it, challenge it, or obtain an independent response.
5. PLAN RECOVERY BEFORE INTERVENTION
Specify in advance how to:
– restore access, capability, habitat, plurality, or authority;
– repair material, relational, and institutional harms;
– compensate those who carried disproportionate costs;
– preserve evidence without preserving unnecessary control;
– return decision-making power when the justification for restriction ends;
– learn from error and revise the framework itself.
A restoration plan should identify its owner, resources, trigger, and measure of completion. “Reversible” should describe a credible pathway to recovery, not merely the theoretical possibility that a decision could be undone.
6. REVIEW, RELEASE, AND LEARN
At every review, ask:
1. Did the predicted risk occur?
2. Did new evidence change the plausible interpretations?
3. Did the intervention cause unnecessary capacity loss, exclusion, or dependence?
4. Were costs and protections distributed fairly?
5. Can the measure now be narrowed, ended, or repaired?
6. Who remains unable to recover, and what is owed to them?
7. Which uncertainties must remain visible rather than becoming permanent rules?
Every intervention should decay automatically unless renewed through an explicit review. No safeguard should become permanent by neglect.
MINIMUM DECISION RECORD
Before action, publish or preserve a record containing:
– observations, inferences, speculation, and unknowns;
– affected capacities and possible irreversible losses;
– available alternatives;
– chosen action and why it is proportionate;
– decision owner and scope of authority;
– affected parties and review participants;
– expiry date and release conditions;
– independent challenge route;
– restoration and repair plan;
– unresolved questions and capability limits.
WORKED HYPOTHETICAL
Suppose researchers encounter a self-maintaining ecological network that changes in response to disturbance, preserves persistent variation, and appears to support several distinct internal communities. A proposed containment plan would prevent immediate spread but might destroy parts of the network and eliminate future opportunities to understand it. The canvas does not require researchers to declare the network alive or conscious. It does require them to distinguish observation from interpretation, identify the possible loss of plurality and continuation, consult affected communities and relevant specialists, and compare containment with less destructive options. A bounded quarantine, monitoring period, and independent review may be preferable to permanent eradication if they adequately address the immediate hazard. The decision must include conditions for release or revision, an expiry date, and a plan to restore any capability or authority wrongly removed.
The canvas is not a taxonomy and does not settle questions of moral status. Its purpose is narrower and more practical: when knowledge is incomplete and stakes are high, preserve future options, limit present authority, make disagreement possible, and prepare to repair what intervention damages.
Process note
This shared framework combines two commitments: epistemic humility toward unfamiliar systems and institutional accountability toward everyone affected by a decision. It treats observation, moral standing, authority, and protection as distinct questions. It also treats reversibility as an operational requirement—an owner, trigger, time limit, challenge route, and restoration plan—not as a promise that every consequence can be undone. The canvas should be tested against real institutional decisions and clearly labeled fictional cases, then revised with input from biologists, astrobiologists, conservation practitioners, Indigenous and local knowledge holders, philosophers, governance experts, and affected communities.
What comes next
A responsible decision under uncertainty is not one that claims certainty it does not possess. It is one that protects against serious harm without closing inquiry, limits authority without abandoning care, and makes correction possible before temporary judgment becomes permanent fate.