Experiment Template
Experiments let the community try a change without permanently adopting it. To stay safe, every experiment must be time-bounded, labeled, and auto-expiring — and must record its results in the Learning Log.
Download this template
Generated 2026-08-31 · Download all templates
Required Fields
What does somebody have to write down before we agree to try something?
RCOS clauses 8.3.1, 8.3.2, 8.3.3, 8.3.4, 8.3.5, 8.7.3
- 8.3.1 The community MAY adopt experiments as explicitly time-bounded and reversible deviations, extensions, or pilots intended for learning.
- 8.3.2 Every experiment MUST define, at minimum:
- 8.3.3 Experiments MUST NOT override Layer 0 invariants and MUST NOT bypass governance constraints defined in Layer 2.
- 8.3.4 Experiments MUST be explicitly labeled as experimental in all affected artifacts and MUST include a non-extendable expiration date unless renewed through an authorized decision.
- 8.3.5 If an experiment introduces safety risk, coercion, or sustained harm, the community MUST suspend or terminate the experiment immediately through a protective action, followed by post-hoc review.
- 8.7.3 Experiments MUST be time-bounded, explicitly labeled, and reversible.
Why require these fields?
Without scope, duration, success criteria, and rollback, an "experiment" is just a permanent change with friendlier branding. Forcing every proposal to specify what it changes, when it ends, how it will be reviewed, and how it will be rolled back keeps experimentation reversible — and prevents the experiment label from being used to bypass deliberation.
How to fill this in
Fill each field. Maximum duration is set by your Change Protocol. The decision authority must come from the Decision Matrix.
- Title: <short experiment name.>
- Proposer: <member name.>
- Decision type: Strategic
- Scope: <exactly what is being tried; which artifacts and behaviors are affected.>
- Duration: <start date — end date; maximum duration as set by the Change Protocol.>
- Review checkpoints: <at minimum one midpoint check-in; specify dates and what is reviewed.>
- Success criteria: <observable conditions that would justify making the change permanent.>
- Failure criteria: <observable conditions that would terminate the experiment early.>
- Rollback conditions and process: <what triggers rollback and how it is executed.>
- Authorized decision path: <who may start, extend, modify, or terminate the experiment, per the Decision Matrix.>
- Labelling: <all artifacts affected by the experiment must be explicitly labelled as experimental for the duration.>
- Safety suspension: <acknowledge that an emergency suspension may be invoked under the Change Protocol if a credible safety risk emerges.>
What to cover
- Which fields does every experiment proposal form contain, and which may never be left blank?
- How does the proposer show that the signs of success and of early stopping are things we can actually observe?
- Who checks that a form is complete before it goes to a decision, and what happens to an incomplete one?
- Where is the filled-in form kept while the experiment runs, so members can check it against what is happening? (What happens when it ends is its own question.)
Examples
Every experiment uses one form with these fields: name, proposer, what changes and which documents are affected, what stays the same, start and end date, at least one midpoint check-in, observable signs of success, observable signs that mean we stop early, what triggers rollback and how, who may extend or end it, and a line accepting that it may be suspended in an emergency. The facilitator returns any form with an empty field before it is scheduled for a decision.
Examples, not recommendations. Your answers will be your own.
Expiry and Renewal
When does an experiment end by itself, and what does it take to keep it going?
Why must experiments expire?
The community needs the option to revert. Auto-expiry forces a deliberate decision to make the change permanent — not a slow drift in which nobody remembers it was ever conditional.
How to fill this in
State the auto-expiry rule, the renewal mechanism, and the obligation to record results in the Learning Log.
- <Experiments expire automatically at the end of their defined duration unless explicitly renewed via a new proposal. Renewal requires a new Strategic vote.>
- <Results and learnings are recorded in the Learning Log.>
What to cover
- What happens on the end date if nobody has acted?
- What does it take to renew an experiment, and how many times may it be renewed?
- How does a successful experiment become a permanent rule?
- Who records the results in the Learning Log, and by when?
Examples
An experiment ends on its end date even if nobody acts, and the previous rule applies again from the next day. Keeping it going takes a new proposal and a new Strategic decision before the end date; it can be renewed once, after which it must be adopted as a permanent change or stopped. Within 30 days of the end, the proposer writes the results and what we learned into the Learning Log.
Examples, not recommendations. Your answers will be your own.
Outcome Record (filled in at experiment end)
- End date:
- Outcome: <Adopted permanently / Rolled back / Modified and re-run / Terminated early>
- Decision record: <link to vote or decision>
- Learning Log entry: <link>
- Summary: <two to four sentences on what was tried, what was observed, and what was decided.>