Inferential reasoning ability moderates the influence of mediated exemplars on risk perception

ABSTRACT

Research has shown that vivid exemplars bias risk perceptions, raising concerns about the role of media in shaping public opinion. However, receiver characteristics that moderate the influence of vivid exemplars remain understudied. We propose that insofar as exemplars in media portrayals represent (at times biased) samples of the population of events, inferential reasoning ability (IRA) – the knowledge of and skill at applying inferential and statistical principles – may enable people to avoid drawing invalid inferences from exemplars. In two message-exposure experiments employing different designs and stimulus materials, we find that exemplars conditionally influence perceptions of risk, depending on the receiver’s IRA. Specifically, higher IRA was associated with reduced influence of vivid exemplars. This work highlights the importance of accounting for heterogeneous audiences in risk messaging and offers tools for measuring audience IRA.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Availability and accessibility can be distinguished (e.g., Turving & Pearlstone, 1966; Vaughan, Basak, Hartman, & Verhaeghen, 2008), consistent with an activation-based model of memory; some stored information may be more accessible, meaning it is more readily retrieved and brought into working memory (depending on the type of cue presented). By this terminology, the “availability heuristic” is a function of accessibility. According to more recent literature, availability simply means the information is stored in long term memory, not that it is particularly accessible.

² The study included sample sizes comparable to prior research on the topic. A post hoc power analysis assuming a two-tailed t-test based on estimates for ratings of fatalities in prior exemplification research for the minimum and maximum treatment groups immediately following the experiment (See Gibson & Zillmann, 1994; Table 1) indicates that a sample size of N = 140 per condition (approximate sample size for Study 1) or N = 110 per condition (approximate sample size for Study 2) would yield a power of > .99, based on calculations using the sampsi command in Stata.

³ We use one-tailed tests (α = .05) throughout the analyses as the key hypotheses are directional.