Description
A between-groups experiment comparing belief ratings after misleading content with and without a correction, split by prior attitude, to test whether corrections work or backfire.
Participants read content containing a false implication; one group also reads a correction. All then rate agreement with the false claim. Nyhan and Reifler (2010) used a mock news story implying Iraq had weapons of mass destruction; the correction reduced belief among very liberal participants, had no effect among centrists and increased belief among conservatives, the so-called backfire effect. Because the outcome depends on prior attitudes, samples must be profiled and results analysed by subgroup. Later replications suggest backfire is rarer than first reported, so the test is better seen as measuring correction effectiveness across segments. In UX it suits fact-check labels, AI output corrections, and product messaging that retracts earlier claims.
Attributes
- Discipline
- Cognitive Psychology & Perception
- Loss type
- Unmigrated
- Stage
- Design, Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Write a stimulus containing the misleading claim, in the format users will actually meet it.
- Write a correction in the candidate design (label, banner, inline note) and, if useful, several correction variants.
- Measure prior attitudes or relevant identity before exposure so results can be split.
- Randomly assign participants to no-correction and correction conditions, aiming for at least 50 per cell per subgroup.
- Collect agreement ratings with the false claim and with a control claim on the same scale.
- Compare belief by condition within each prior-attitude subgroup, testing the condition-by-attitude interaction.
Sources
- Cognitive Psychology (Goldstein, 2019)
Also known as
- Continued-influence paradigm
- Backfire-effect test
- Correction effectiveness test
Categories
- User Research
- Cognitive Psychology & Perception