6 ideas
Anyone who argues that reason is unreliable or merely a tool of power must give reasons and expect others to accept them, so the argument relies on the thing it rejects. Pinker therefore treats rationality as a precondition of any debate, not one position within it.
People reason carefully about things they directly experience, such as their groceries, jobs, and neighbors. They hold beliefs about distant matters, such as cosmic origins, remote politics, and history, in a 'mythology zone', where the belief is valued for being uplifting or signaling identity rather than for being true. This explains why otherwise competent people endorse conspiracy theories: the belief sits in a zone where accuracy was never the criterion.
Each individual gains socially by adopting and loudly defending the beliefs of their own political or cultural tribe, while bearing almost no personal cost for being wrong. When everyone does this, a society's shared capacity to find the truth degrades, the way a pasture is overgrazed. Myside bias is therefore individually rational and collectively ruinous, and fixing it requires institutions that reward accuracy over loyalty.
How much a piece of evidence should shift your belief depends on the prior probability of the hypothesis, not only on how accurate the test is. With a rare disease, a 90%-sensitive test with a modest false-positive rate still yields mostly false positives, because healthy people vastly outnumber sick ones. People routinely neglect the base rate and fixate on how vivid or representative the evidence seems.
A game-show contestant picks one of three doors; the host, who knows where the car is, opens a losing door and offers a switch, and switching wins two-thirds of the time. When columnist Marilyn vos Savant gave this answer, thousands of readers, including PhDs, wrote in to insist she was wrong. The case shows that intuition ignores how the host's knowledge constrains which door he opens, and so ignores the information that choice carries.
Any yes/no judgment under noise, whether a diagnosis, a verdict, or an alarm, has two independent parts. Sensitivity is how well the judge distinguishes signal from noise. The response criterion is how much evidence the judge demands before saying yes. Moving the criterion only trades misses for false alarms, so choosing where to set it is a question of values about which error is worse, not a question of accuracy.