Tool 13 of 32 · about an hour
Diagnosis Bias Scan
Runs on whoever is doing the diagnosing, including us, before any conclusion is acted on.
In Pattern, before findings are written up or any conclusion is acted on, including when you are diagnosing your own company.
What it is for
To catch errors in the diagnosis itself, by checking the conclusion against the common distortions before anyone acts on it.
The failure it catches
The diagnosis is assembled by humans replaying their own lived experience, and experience cuts both ways: deciding the answer early and collecting evidence for it, pattern-matching this company to a previous one it merely resembles, anchoring on the first strong finding, and selecting evidence that fits the story already forming. A biased diagnosis poisons every tool that follows while looking rigorous.
Why it works
Naming the distortion is cheap. Running the check that would catch it if it were real is the active ingredient. A conclusion that has never been attacked is just a mood with paperwork.
How to run it
- Before acting on any conclusion, answer in writing. Which of the seven distortions is most plausibly pulling on this diagnosis: sunk cost, urgency, ego preservation, narrative, recency, social pressure, ambiguity aversion.
- Answer the three diagnosis-specific questions. Which finding did I anchor on first. Which other company is this one reminding me of. What evidence would change my named root, and did I actually look for it.
- For each answer, name the check that would catch the distortion if it were real, usually a Convergence pair or a re-run of the tool that produced the finding, and run that check before write-up.
- An unanswered what evidence would change my mind blocks the write-up.
How to read the result
Any yes means go back to the records and try to break your own conclusion before you act on it. A diagnosis that survives an honest attempt to kill it is worth acting on. The scan is complete when the anchor finding, the reminded-of company, and the disconfirming search are all named in writing and the named check has run. Never present the scan as proof of objectivity. It is hygiene, not a credential. A second version of this scan runs live during depth sessions, and that one lives inside the Room Bias Scan.
How to run it well
- The active ingredient is running the named check, not naming the bias. Performed as a checkbox it is theatre.
- Include yourself. This tool runs on whoever is doing the diagnosing.
Where it breaks
- Performed as a checkbox it is theatre. Naming a bias and then skipping the check that would catch it is how a diagnosis stays a mood with paperwork.
- Presenting the scan to anyone as proof of objectivity turns hygiene into a credential.
What would prove it wrong
If the scan never changes a finding across many runs, either the diagnoses are cleaner than claimed or the scan is toothless. Audit which by re-running one diagnosis cold.
What we watch
Whether the named disconfirming check ever revises a root before anyone acts on it, or whether the scan only ever confirms.
Ask yourself
- Which conclusion did you reach first, and did you then look for evidence that would overturn it?
- Is this company reminding you of a previous one, and what would make this case different?
- Which distortion is most likely pulling on you today: sunk cost, urgency, protecting a story, recency, social pressure, or discomfort with ambiguity?
- What evidence would change your named root, and did you actually look for it?
- Is the conclusion suspiciously close to the one that makes the diagnoser most valuable?
Fed by: Room Bias Scan
Feeds: Convergence Test