Why People Go Wrong
People go wrong through missing evidence, poor tools, distorted perception, social pressure, emotional override, and identity protection.
There Is More Than One Failure Mode
The word "irrational" is too blunt. People can be wrong because they lack information, because they process information badly, because the evidence itself is noisy, because incentives punish accuracy, because fear or hope overrides judgment, or because belonging to a group has become more important than tracking reality.
Credencing treats these as different diagnoses. Different diagnoses imply different repairs.
Deep-Rational Failure
Deep-rational failure is failure in the tools of perception and inference. The agent may ignore base rates, misunderstand a test, confuse correlation with causation, overread anecdotes, neglect alternative explanations, or fail to notice how selection effects shape the evidence they see.
This kind of failure often calls for education, better measurement, better comparison classes, better statistical habits, and more careful exposure to disconfirming information.
Core-Rational Failure
Core-rational failure happens when the agent assigns a credence that departs from their own perceived evidence. They may see the weakness of an argument but keep the belief because it protects identity. They may see the strength of an objection but discount it because accepting it would be costly. They may know that a fear is exaggerated but act as if it were certain.
This kind of failure often calls for courage, emotional regulation, social safety, humility, and practices that reduce the cost of changing one mind.
The Intellect as Bodyguard
One of the most important cases is the biased expert. This person has enough intelligence and training to process evidence well, but uses that capacity defensively. The intellect becomes a bodyguard for a prior commitment. Rather than asking what is true, it asks how the protected belief can survive.
This is why education alone does not guarantee rationality. Better tools can improve tracking, but they can also improve rationalization unless paired with core-rational integrity.