Confirmation Bias
If you went looking for evidence you were wrong, where would you have to look?
The idea
People seek evidence that supports an existing belief and discount conflict.
In the real world
A founder celebrates praise while ignoring churn.
Going deeper
The damage is done at collection, not at reasoning. A person filtering evidence on the way in can reason impeccably about everything that arrives and still end up confidently wrong, because the sample was selected to agree.
It is also self-reinforcing over time. Supporting cases are remembered as evidence; disconfirming ones are explained away individually and then not retained as a pattern. After a year the belief is backed by a large and entirely one-sided file, which feels exactly like being well-informed. The only reliable counter is to specify in advance what evidence would count against you, and then go and look for that specific thing.
Where it stops applying
Some filtering is necessary — attending equally to all evidence is impossible and treating fringe claims as equal to strong ones is its own error. The bias is in the asymmetry, not in having priors.
Why it matters
It explains how a sincere, careful person accumulates a year of supporting evidence for a belief that was never true.
Try this today
Write what evidence would change your mind.
Test yourself
A founder reads every positive review, quotes them in board updates, and describes churn as a few unhappy edge cases. What is the specific error?
Show the answer
The evidence is being filtered on the way in, not weighed once collected. Supporting cases are sought out and disconfirming ones explained away, so the amount of evidence grows while its ability to correct anything falls to zero.
Learn this in the feed Answering from memory, then again days later, is what makes it stick.