Cognitive Biases · Belief
AI hallucination
Also called: Confabulation, Fabricated output
An AI model states something false or made up, such as a fake citation or number, in the same confident tone as its accurate answers.
Evidence grade: Strong. How we grade evidence
For example
An analyst asks an AI tool for studies on volunteer retention. It returns five citations with authors and journals, in a confident tone. Two of the studies do not exist.
A caution
A property of AI models, not a human bias. Rates vary by model and task. The human risk is trusting a fluent answer without checking it.
Spot it in your week
Three quick questions. Your answers stay on this page.
Your DecisionDNA™ can tell you which biases you're most exposed to.
Find your DecisionDNA™Sources
Related entries
- Cognitive BiasesAlgorithmic biasA model repeats or widens unfair patterns in its data, often because it predicts a...
- Cognitive BiasesBase Rate fallacyIgnoring how common something is overall when judging how likely it is in a specific case.
- Cognitive BiasesComposition fallacyAssuming that what is true for a part is also true for the whole.
Bias of the week
One decision trap a week, in plain English. Unsubscribe any time.
We send one confirmation email first. What we keep.
Get the countermeasure
Get the countermeasure and the decisions where this bites, free in the Volley app.
