What you are fighting
Confirmation bias is not "having an opinion." It is the habit of hunting for support, skipping counterexamples and reading ambiguous news as proof. Francis Bacon described the pattern in 1620; Nickerson's 1998 review still supplies the working definition used in psychology classrooms.
Wason's 2-4-8 task shows the trap in miniature: people propose triples that fit their guessed rule and rarely propose triples that would kill the guess. The same shape shows up in comment threads, hiring debriefs and literature reviews that only cite friendly papers.
For neighboring concepts see hindsight bias versus confirmation bias and belief perseverance. This page is only the avoidance checklist.
Five-step checklist (each step has a pass/fail test)
Run the steps in order. Skipping to "I asked a friend who agrees with me" is a fail on step 3.
| Step | Action | Pass | Fail (common) |
|---|---|---|---|
| 1. Name the claim | Write one sentence you currently favor | A falsifiable statement on paper | A vibe ("they are shady") with no test |
| 2. Seek disconfirmation | List evidence that would count against you; search for it first | You can point to sources that hurt the claim | Only bookmarks that agree; UC Merced calls the fix "focus on falsification" |
| 3. Outside view | Ask someone who already disagrees, or assign a devil's advocate | A real opposing case you can restate fairly | Asking a teammate to "play skeptic" for 30 seconds then ignoring them |
| 4. Separate want from test | Write what outcome you hope for, then what data would settle it | Hope and criterion are different sentences | Moving the goalposts after seeing the result (related) |
| 5. Pre-commit the update | Before new data arrives, write what would make you change your mind | A dated note you can reopen later | "I will know it when I see it" with no threshold |
Why "consider the opposite" beats more of the same reading
Lord, Ross and Lepper's classic work on biased assimilation showed that the same mixed evidence can polarize people further when each side mines it for support. The repair used in later decision research is not "read harder." It is to generate the strongest alternative explanation and ask what would favor that alternative.
UC Merced's short HR note compresses the same idea into three workplace moves: hunt for disconfirming evidence, leave the echo chamber, and talk with an outside party who is not invested in your story. None of those steps require a statistics degree.
In research settings, Lakens groups stronger versions under organized skepticism: preregistration, fixed error rates, independent replication, devil's advocates and adversarial collaborations. You can steal the spirit for everyday decisions without running a clinical trial: write the test before you see the outcome.
Where the checklist still fails
Identity threats. If changing your mind costs status in your group, steps 2 and 3 get performed as theater. Log the social cost honestly; sometimes the fix is a private note, not a public recant.
Low-quality dissent. A devil's advocate who only repeats slogans does not count as step 3. Nemeth's work suggests authentic minority views beat ceremonial skeptics. Prefer someone who actually holds the other view.
Blind spot about the blind spot. People who finish this checklist can still rate themselves as less biased than peers. Pair it with Pronin's self-versus-other asymmetry research if the argument is about who is the rational one in the room.