Confabulation
When people explain why they did something, the explanation can be a confident invention rather than a true account — and they do not know it is invented. The reasons we report are often built after the fact from our everyday theories of cause, not read off from any inner view of what actually drove the behaviour. What began as a word for the way some patients fill memory gaps with believed fabrication is extended here to ordinary, undamaged minds.
The production of a confident, coherent causal account of one's own behaviour that is in fact invented, offered without awareness that it is invented. The founding empirical statement is Nisbett and Wilson (1977), who argued people have little direct introspective access to their higher cognition and report a priori causal theories instead; Gazzaniga's left-hemisphere "interpreter" is the neuroscientific version.
Etymology§
Confabulation enters from clinical neurology, where it named the way amnesic patients fill gaps in memory with fabricated detail they believe. The black-box literature uses it in the extended sense — the same gap-filling running in ordinary, undamaged cognition.
Confabulation is the human black box stated as a finding rather than a metaphor. Nisbett and Wilson assembled experiments in which people confidently explained choices whose real cause the design showed they never noticed, and concluded that verbal reports of mental process are reconstructions built from folk theories of cause. Gazzaniga's split-brain work locates a mechanism: a left-hemisphere interpreter that supplies reasons for actions whose true prompt it cannot see, inventing a plausible account rather than reporting ignorance.
The corpus's predictive-processing material describes the brain as an inference engine that constructs its own states, which several theorists treat as the substrate beneath this kind of after-the-fact report. The machine parallel is unfaithful chain-of-thought: a model that fabricates a rationale for an answer driven by something its rationale never names — the same shape as human post-hoc rationalisation, relocated to a language model.
Discussed in§
- Who's in Charge? Free Will and the Science of the Brain
- The Integrated Mind
- The Emotional Dog and Its Rational Tail: A Social Intuitionist Approach to Moral Judgment
- Telling More Than We Can Know: Verbal Reports on Mental Processes
- Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting