Data Talks

Bad news, fellow academics: a whole series of experiments shows that verbal hedging (saying something might “possibly” be true or that “I might be wrong”) does not actually protect your reputation if you are, in fact wrong. Also, bonus points for a terrific paper title. pic.twitter.com/Jb4Hw5TvFa

— Ethan Mollick (@emollick) February 3, 2021

 From the Abstract:

We gain much of our knowledge from other people. Because people are fallible—they lie, mislead, and are mistaken—it seems essential to monitor their claims and their reliability as sources of information. An intuitive way to do this is to draw on our expectations about claims and sources: to perform expectation-based updating (Hahn, Merdes, & von Sydow, 2018). But this updating can have damaging consequences, leading us into a kind of confirmation bias. An alternative is to keep track of outcomes and record whether a claim proves true or false: to perform outcome-based updating (Hahn et al., 2018). This form of updating does not have the negative repercussions on belief accuracy. But both forms of updating might undermine the trust and cooperation assumed to be necessary for successful communication. We explore a potential boundary condition on these types of updating. We investigate whether speakers can protect their reputation when they make claims with low prior probability, with and without knowledge of the final outcome. We explore suggestions from McCready (2015) that speakers can protect themselves by hedging with evidential language: in particular with weaker propositional attitudes (“I suspect that . . .”) and so-called double hedges (“I might be wrong, but I think . . .”). We find that both forms of updating are robust to hedging with this evidential language and find no clear evidence for a protective effect. We discuss extra ingredients that may allow successful hedging. (PsycInfo Database Record (c) 2020 APA, all rights reserved)