Skip to content

Critical Thinking — Best Practise

How to make the five moves automatic, and how to make them work even when the strongest critical thinker on the team isn't in the room.

Daily practice

  • Before accepting any claim in a doc, review, or standup: ask three questions — what evidence would prove it false, what alternative explanation fits the same observation just as well, and who benefits from the claim being true. If you can't answer all three, you don't understand the claim well enough to act on it.
  • Before weighing evidence from two sides: default to direct evidence (measured on your own system) over indirect evidence (a blog post, a colleague's story, a general principle) — someone else's benchmark is a hypothesis about your system, not a fact about it.
  • Before treating agreement as settled: ask "what's the strongest case against this?" out loud. Silence or a strawman answer means the team hasn't actually understood the trade-off.

The pattern that scales past one person: the decision log

A rotating devil's-advocate role and a written decision log turn critical thinking from one person's trait into a team default — a ritual that only works when a specific person shows up isn't a capability, it's a single point of failure.

  • Assign a rotating devil's advocate for any decision above a defined stakes threshold (expensive to reverse, crosses a compliance boundary, affects more than one team). Rotate it so the role stays a structural check, not "that person's known opinion."
  • Log the decision at the moment of commitment, not after — logging retroactively lets the "evidence" field get reverse-engineered to match the outcome. At minimum, record: the options considered (including "do nothing"), the evidence for the deciding factor, the strongest dissent and why it did or didn't change the outcome, and a re-evaluation trigger with an observable condition ("revisit if p99 exceeds 50ms" — not "revisit if it doesn't work out").
  • Audit a sample of past decisions quarterly for a recurring pattern — not to relitigate one decision, but to check whether the same bias (e.g. "do nothing" never seriously considered, the same two people's opinions anchoring every discussion) shows up across several. Report the pattern to the team, not the individual decision.

Level up as scope grows

  • One claim in a doc or PR → isolate it, evaluate the evidence, name the alternative.
  • A team decision → deconstruct the whole argument into premises, mark each fact/assumption/opinion, and find the weakest premise the conclusion actually depends on.
  • An organization-wide decision → install the rotating devil's advocate and the decision log so the check survives any one person being out sick or leaving the team.

Self-check before calling it done

  • Can you write the claim as one sentence, with no evidence mixed in?
  • Can you name the evidence's source, and whether it's direct or indirect?
  • Did you name at least one alternative explanation before accepting the first one?
  • Did you write down what would need to be true to change the conclusion?
  • If this was a team decision, is there a decision log entry with a real re-evaluation trigger?

Signs you're getting fluent

  • You separate a claim from its evidence before anyone asks you to.
  • You reach for "what's the strongest case against this?" instead of taking silence as agreement.
  • You catch a named bias or fallacy in the moment, and name the pattern instead of the person.
  • Your team's decisions get logged and revisited whether or not you're in the room.