ABSTRACT
Refutation texts are designed to facilitate the revision of inaccurate knowledge; however, studies have documented backfire effects wherein respondents become less accurate when exposed to a factual correction compared to another. Here, we explored whether epistemic emotions mediated knowledge revision or backfire processes when reading experimental refutation texts that varied by supporting information. We asked 294 online readers to report their knowledge and attitudes about genetically modified foods before randomly assigning them to one of three refutation text conditions which varied in type of supporting information. We documented relatively low knowledge (19.7%) and attitude (14%) backfire across conditions and found no evidence of backfire effects among types of supporting information. All texts facilitated knowledge revision regardless of the type of supporting information but did not differentially elicit epistemic emotions. For those who revised their inaccurate knowledge, hope and surprise mediated the knowledge revision process. However, those who demonstrated a backfire effect reported significantly more anger and confusion than readers who revised their inaccurate knowledge. Implications for research and practice are discussed.
Disclosure of potential conflicts of interest
No potential conflict of interest was reported by the author(s).
Notes
1. The sample size of approximately 300 (100 participants in each of 3 text) was a rounded estimate based on power analyses for a 1 × 3 ANOVA to detect an effect size of .1 with power of .9 using the “pwr2” package in R (Lu, Liu, & Koestler, Citation2017). This sample is also acceptable for detecting mediation effects (Fritz & MacKinnon, Citation2007).
2. Note that we ran all analyses with the full sample. Although coefficients were slightly different, the significance levels and direction of effects for the main findings were identical.
3. Note that we reran all analyses with positive emotion and negative emotion composite variables and replicated the main findings regarding interactions between prior attitudes and knowledge in the logistic regressions when the outcome was either knowledge or attitude backfire. Similarly, the main findings from the path analyses were replicated. Namely, we found that for the full sample, positive emotions significantly mediated relations between pre-reading and post-reading attitudes, and negative emotions mediated relations between pre-reading attitudes and post-reading knowledge and attitudes. This was also the case for those who experienced knowledge revision, but no significant mediations for those who backfired in knowledge, and identical results when the sample was split by attitude backfire.