The case for keeping the losing argument

Imagine a team deciding whether to try a new tool for tracking its work. The old arrangement loses requests between email and a spreadsheet. The new one is easier to use, and most of the team wants to begin a small pilot.

One engineer remains concerned. In two tests, a dropped connection followed by a retry created duplicate jobs. She wants the fault understood before anyone relies on the tool. The project lead proposes a limited trial, with a daily comparison against the existing record. The team agrees to proceed. The engineer accepts that the decision has been made, but remains unconvinced that the safeguard will hold when work gets busy.

The minutes say: Two-week pilot approved. Daily checks agreed.

That is an accurate record of the decision. It leaves out the particular thing someone thought might go wrong, the evidence she offered, and the reason she doubted the response. A month later, a new colleague can discover what the team authorised without discovering what it still did not know.

I think that distinction deserves more care. A decision gives people a course of action. It does not make the uncertainty behind it disappear.

Someone has to choose. Waiting has costs, and complete agreement may never arrive. A competent leader can hear a serious objection, understand it and still reasonably proceed. Yet the objection may contain information that will matter after the decision has done its immediate work. Keeping it is part of taking responsibility for what happens next.

A minority report, in the sense I mean here, is a record of a serious disagreement and its reasons. It need not be a separate report. A few accurate paragraphs beside the decision may be enough. The purpose is to let a later reader recover the strongest unresolved argument without reconstructing the meeting from memory.

The word losing describes what happened in the room. It says very little about what will happen outside it.

A concern someone can investigate

Summaries necessarily leave things out. A page that gives every remark equal space can make the important point harder to find. The question is which details carry the decision’s uncertainty. In our imagined meeting, everyone knows that the spreadsheet is cumbersome. Only one person has closely examined the duplicate jobs. If the summary keeps the familiar problem and compresses her concern into “technical issues discussed”, it removes the very information the meeting brought together.

There need be no dishonesty in this. The project lead wants an actionable account. The engineer may explain the fault badly. The person taking notes may reasonably assume that agreeing to a safeguard means the concern has been settled. Each contribution can be understandable while the final record says more about agreement than knowledge.

That is why I would give the person raising a material objection a chance to check its wording. A concern can be acknowledged politely and still be misrepresented. “The engineer opposed the pilot” records a position. “The engineer thought daily checks would fail under a heavier workload” preserves a claim someone can investigate.

There is a constructive example in NASA’s account of its X-43A programme. A dissenting opinion questioned the aerodynamic data used in flight-control design, raising the possibility of losing control of the vehicle. The NASA Engineering and Safety Center worked with the project to ensure that the concern was examined through an existing independent Flight Readiness Review.

Its 2004 report describes a committee appointed independently of the project, with relevant technical expertise and several opportunities for review and follow-up. It says that examining dissent can lead either to a changed risk assessment or to a reaffirmed one. NASA Engineering and Safety Center, 2004 Annual Report, p. 18.

What I take from that example is the connection between speaking and receiving an answer. The concern acquired a route through people able to examine it. The value of that route did not depend on promising the dissenter victory. It depended on giving the question a serious response.

Keep it beside the decision

Most decisions do not require an aerospace review. The useful habit can be much smaller. For the imagined pilot, a note in the existing project document could read:

Decision: Begin a two-week pilot with one team, retaining the current record alongside it.

Remaining objection: Two retry tests produced duplicate jobs; the engineer doubts that manual checks will remain reliable under a heavier workload. Test notes are attached.

Why proceed: The project lead judges the limited trial manageable with daily reconciliation; wider adoption is still undecided.

Reopen when: A duplicate appears in the pilot, or a daily check cannot be completed.

Reviewer: The project lead and engineer will examine the record at the first weekly review, or sooner if either condition occurs.

The note is useful because it keeps the disagreement specific. It also makes the chosen response visible. A successor can see that the team accepted a bounded uncertainty for a stated reason. The engineer can see that her concern has survived in a form she recognises. Neither needs to pretend that the discussion ended in shared conviction.

The agreed review matters as much as the words. A concern left in a folder may be available in theory and absent from the next decision. Putting the note beside the pilot results gives it a natural occasion to be read. It can then help the team distinguish several quite different outcomes.

Suppose the pilot runs smoothly. Perhaps the engineer overestimated how often the fault would recur. Perhaps the relevant connection failure never occurred. Perhaps the daily comparison caught a duplicate before it caused confusion. All three can produce a reassuring headline. They teach different lessons about expanding the trial. The record directs attention towards what actually happened to the disputed assumption.

Suppose instead that the problem appears. The note offers a place to begin investigation without requiring the engineer to re-establish that she raised it. Her prediction may have been partly right and her proposed remedy unnecessarily costly. The majority may have understood the fault but overestimated the safeguard. Useful learning is more precise than awarding one side the whole argument.

This applies to the people who made the decision as well. A fair record can show that proceeding was reasonable given what was known, even if the result was disappointing. It can also show when a fortunate result concealed poor reasoning. I want the record to preserve enough of the original circumstances that neither success nor failure gets to rewrite them completely. Keep the original note dated, and add later findings beside it. Otherwise, the record itself can begin to borrow certainty from the outcome.

Leave room for work to continue

There are limits to what deserves this treatment. An organisation cannot give every preference a standing invitation to return. I would reserve it for disagreements that could materially change a decision, expose a consequential assumption, or affect people who bear its costs. A choice of heading does not need the same memory as a commitment that will be expensive to reverse.

The dissenter also owes the record some discipline. What exactly is disputed? Which observation or principle supports it? What would make the concern less persuasive? Listing every imaginable failure creates plenty of opportunities to claim foresight and very little help for the person who must act. A good objection leaves room for its author to learn something too.

Preserving it grants no permanent veto. Once a legitimate decision is made, work can continue within its agreed limits. Reopening it should have a reason: new evidence, a changed assumption or the arrival of a planned review. Otherwise, the record can become a way of prolonging the meeting indefinitely. Immediate concerns still need an appropriate response; writing them down for later cannot make an urgent problem safe to defer.

Confidentiality needs equal care. Some useful reasoning contains personal, commercial or other sensitive information. Its audience and level of detail should fit its purpose. The people responsible for a decision should know where the relevant evidence belongs and who may examine it. A broad summary need not expose the underlying material to everyone, and preserving an argument need not mean preserving every remark about its author.

I would want such a record to make candour easier. If every objection becomes a permanent mark against the person who offered it, people have a reason to keep their doubts private. The account should follow the evidence and response closely enough that a successor can use it without inheriting the personalities of the dispute.

What the next person inherits

That becomes particularly valuable when the original participants move on. “We tried that before” carries a conclusion but leaves the next person dependent on someone else’s memory. A brief explanation of the conditions, the objection and the result gives them something they can judge. They may discover that the old decision still fits. They may discover that the conditions have changed. Either way, they inherit more than an instruction.

I want this care in AI-assisted work too. When I ask an assistant to condense several analyses, the result I want is a clear recommendation with the consequential disagreement still visible. Fluent prose can make a choice feel settled without showing which evidence the alternatives depended on. My test for a useful summary is whether I can still find the strongest reason to choose differently and understand why the recommendation proceeds despite it.

More generated objections would not necessarily help. Several models repeating variations of a concern are not several independent witnesses. The useful contribution is a question tied to something inspectable: a source, a test, a missing fact or an assumption that could change the conclusion. I would rather retain one such objection than pages of artificial debate.

Return to the team at the end of its first week. The review begins with what the pilot revealed, alongside what each side expected. They can continue, change the safeguard or stop. The earlier decision gave them permission to learn through a limited trial. The retained disagreement helps them recognise what they have learned.

That is the record I want to leave: clear enough to act on now, and honest enough that someone can think beyond it later.