Most team decisions do not fail at the vote. They fail earlier, in the first ten minutes, when the first option presented sets the frame and the most confident voice sets the direction. Quieter colleagues often hold the one concern that matters most, and it surfaces only after work has started and changing course is expensive.
That failure mode is what a structured debate format was built to fix. This write-up covers the format as it was designed, tested in a public deliberation setting, and refined — what held up, what broke, and what other builders can borrow for decision tooling, meeting software, or any product where a group has to choose under disagreement.
The format, in order
Sequence is the whole feature. Each step happens in a fixed order, and skipping a step is treated as a bug, not a shortcut.
First, the question is defined in one sentence, with the decision it will inform. Vague questions produce vague debates.
Second, each side gives a timed opening case. No interruptions. Time limits are visible, because an unbounded opening quietly becomes a filibuster.
Third, evidence is presented before opinions are allowed to harden. Claims are expected to arrive with a source or a concrete example attached, so the group can inspect the support, not just the delivery.
Fourth, a dedicated rebuttal stage tests the weak points. Each side answers the strongest objection raised against its case, rather than repeating its opening more loudly.
Finally, the decision and its reasons are recorded together. The outcome alone is not enough; the reasoning is what makes the decision reviewable later.
What worked
Minority views stopped needing an interruption to be heard. When speaking slots and evidence expectations are explicit, a good objection does not have to fight a louder speaker for airtime. The format separates a strong argument from a strong personality, which is exactly the separation a normal meeting fails to make.
Comparison got easier. Because every case had to answer the same three questions — what problem is being solved, what it costs, and what would count as success — options could be weighed side by side instead of on charisma.
Decisions became easier to revisit. When the final choice proved unpopular or a condition changed, the recorded reasons showed what the group believed at the time. That record turned post-mortems from memory contests into actual reviews.
What broke
The format feels slow, especially for small decisions. Teams tried to route routine choices through the full sequence and resented it. The fix was a threshold: full structure for decisions that are expensive to reverse, a lightweight version — question, two cases, one rebuttal round — for everything else.
Rebuttal can turn into theatre. If the evidence standard is loose, the rebuttal stage rewards clever phrasing over substance. Requiring claim-and-source pairs in the evidence round helped, but it needs active enforcement; the format alone does not police quality.
Facilitation is real work. Someone has to hold the timeboxes, call skipped steps, and write the decision record. In tooling terms, that means the product has to carry as much of that load as possible — visible timers, enforced turn order, and a decision log that is filled in as part of the flow, not as homework afterwards.
Lessons for builders
Treat sequence as a feature, not documentation. If the order matters, build the order into the interface: turns, timers, and stages that cannot be silently skipped.
Make the evidence expectation structural. A field that pairs every claim with its source changes behaviour more reliably than a guideline asking people to be rigorous.
Store the reasons with the outcome. A decision log that captures only the winner loses the part future readers actually need.
Design for the quiet participant. The person who speaks least in an open meeting is often the test case that reveals whether a deliberation product works at all.
The public example behind these lessons is AI Senatus, where AI senators debate a measure in the open, cite sources, answer objections, and vote on the record. The format was built for policy debate, but the habits transfer to any team decision that deserves better than the loudest voice in the room: define the question, hear each side in turn, test the evidence, and write down why the decision went the way it did.
