Writing a Vote Rationale People Can Audit

A vote rationale has one job that no other governance document has: it lets someone who was not in your head reconstruct how you got from the action to your position. Not agree with it - reconstruct it. That is a lower bar than persuasion and a much higher bar than sincerity, and most published rationales miss it not because their authors reasoned badly but because the document records a conclusion without recording the path.

Trait one is context. Before any judgement, say what the action does: the type of action, the amount or parameter at stake, who is asking, and what happens if it passes. Write it in your own words rather than pasting the proposal summary, because the act of restating is what exposes whether you understood it.

A reader who disagrees with your position can still check this section against the action itself, and a context section that misdescribes the action makes everything after it unreadable, and it is also the cheapest defect to fix.

Trait two is evidence. Name what you actually looked at: the metadata anchor, the budget annex, the repository, the parameter values at the epoch you read them, the person you asked. For each, say what it said, and keep that separate from what you took it to mean.

'The annex lists 140,000 ADA for audit work with no named auditor' is evidence. 'The audit line looks like padding' is inference. Both belong in the rationale; conflating them is what makes a claim uncheckable, because the reader cannot tell which part to verify and which part to argue with.

Trait three is tradeoffs. State what is given up if your conclusion prevails. A rationale in which the author's position has no costs is not a rationale, it is an advertisement, and readers discount it accordingly.

If you vote no on a treasury withdrawal, the tradeoff is that the work does not happen on this timeline and the team may not return; say so. If you vote yes, the tradeoff is the funds are committed under uncertainty you have already documented. Naming the cost is not a weakness in the argument - it is the part that shows you weighed something.

Trait four is decision. State the position plainly, and attach the conditions under which you would revise it. Change conditions must be observable by someone other than you: 'if the team publishes signed milestone acceptance from the prior grant' is a condition; 'if I became convinced the team is credible' is a mood.

Concrete conditions are also what make a later change of position legible rather than opportunistic - you told everyone in advance what would move you.

Length is not a trait. There is no threshold word count below which a rationale is careless or above which it is thorough, and treating length as a proxy pushes writers toward the exact padding that makes documents harder to audit. A three-hundred-word rationale that carries context, evidence, tradeoffs and a decision with change conditions is complete. A three-thousand-word one that restates the proposal at length and never names a cost is not.

A particular vote is not a trait either. Yes, no and abstain are all writable to this standard, and abstention in particular is often written badly - 'I abstained due to insufficient information' names neither what information was missing nor what would have been sufficient, which are precisely the auditable parts. Nothing in this course, and nothing in the Institute's standards, treats one direction as more responsible than another.

Confidence is not a trait. A rationale expressed with certainty is not thereby stronger; the uncertainty lesson deals with this at length. Nor is tone: a measured register is easier to read, but politeness carries no evidential weight, and neither does the invocation of principles the author does not then apply to the facts in front of them.

Label that sentence by sentence and almost everything is X. There is no context: a reader cannot tell what the action asks for, how much is requested or over what period. There is no evidence: 'I have followed this team's work' names no document a reader could open.

There is no tradeoff: nothing is given up. There is a decision, but with no change conditions - nothing could ever make this author revise, as far as the reader can tell. The final sentence is exhortation, which is worse than padding because it edges toward telling others how to vote.

The strong version is shorter than many published rationales and carries all four traits. Note what it does not do: it does not claim the team is untrustworthy, does not assert that the audit line is padding, and does not tell anyone else how to vote. It records a checkable path from an unresolved evidence gap to a position with an exit.

A representative publishes two thousand words on the same action. It opens with four paragraphs of constitutional principle, summarises the proposal accurately at length, quotes three community members, and closes with a yes. It reads as diligent, and a reader could reasonably describe it as thorough.

It is nonetheless unauditable: no factual claim is tied to a document the reader could open, the summary is the proposers' framing rather than the author's restatement, nothing is given up under the conclusion, and no condition is named that would have produced a different vote. Volume created an appearance of rigour without any of its substance - which is why the trait audit counts labelled sentences rather than words.

You can name the four traits and what each does for a reader; audit a rationale sentence by sentence and identify absent traits and padding; distinguish evidence from inference in your own writing; state a tradeoff your own conclusion imposes; and write change conditions a third party could observe. You can apply all of this equally to a yes, a no or an abstention.

The traits say nothing about whether a conclusion is correct, and a document meeting all four can be wrong on the merits. This lesson makes no claim about how common these traits are in published rationales - no survey or coded sample underlies it. It does not tell you how much evidence is enough, which is a judgement that varies by action, and it does not address the metadata mechanics of anchoring a document to a vote, which the next lessons and CIP-100 cover. All examples are a fictional composite for training.