ABSTRACT
When people encounter potential hazards, their expectations and behaviours can be shaped by a variety of factors including other people's expressions of verbal likelihood (e.g., unlikely to harm). What is the impact of such expressions when a person also has numeric likelihood estimates from the same source(s)? Two studies used a new task involving an abstract virtual environment in which people learned about and reacted to novel hazards. Verbal expressions attributed to peers influenced participants’ behaviour toward hazards even when numeric estimates were also available. Namely, verbal expressions suggesting that the likelihood of harm from a hazard is low (vs. higher) yielded more risk taking with respect to said hazard. There were also inverse collateral effects, whereby participants’ behaviour and estimates regarding another hazard in the same context were affected in the opposite direction. These effects may be based on directionality and relativity cues inferred from verbal likelihood expressions.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 At the start of data collection (Fall, 2010), there were two additional conditions included in our design that tested hypotheses quite different from those discussed here. Neither included any verbal information from peers. In one, participants received peers’ numeric estimates about both hazards rather than just one. In the other, participants were given peers’ numeric estimates about one hazard, but they were asked to guess the numeric attack rate before receiving it from peers. After collecting data from 91 participants for the three conditions reported here (and doing preliminary analyses), we decided to split the design into two separate studies, and we collected data from another 117 to arrive at our full sample of 208 in this study. This process was not optimal, but it should have little bearing on the interpretation of the study's results. The final sample size was based on a target of 200 participants (shaped by informal power estimates) coupled with variation due to lab logistics (time of semester, sign-up rates, availability of experimenters).
2 The precise expected value for the risky option is difficult to compute because it depends, in part, on how quickly participants implement their decisions (going to the safe zone early rather than late during the 4-second buffer sacrifices more point-gaining opportunities), how quickly they proceed out of the safe zone when released, and how generally efficient they are at chasing targets and gaining points. It is even possible that going to the safe zone occasionally serves as a rest, which might make people more efficient in immediately subsequent trials in which they stay out of the safe zone. Informal simulations suggest that a person who chases targets with moderate intensity would earn 50–60 points on most trails within 6 seconds, which is the length of time that one is precluded from catching targets when in the safe zone. The penalty of being attacked is 100 points. Given all this information, we could tentatively estimate that the expected value of the risky option is greater than 0 points for any hazard that attacks less than 55% of the time. The expected value of going the safe zone is 0 points.
3 When block is included for a 2 (numeric estimate) × 3 (verbal estimate) × 3 (block) ANOVA on decisions from nonfocal trials, the block factor is involved in a significant two-way interaction with verbal estimate (p = .02) and a significant three-way interaction (p = .04). However, these unexpected interactions do not seem especially important, and they do not replicate in Study 2.
4 We do not claim that the numeric information from peers was not salient. As reported earlier, the manipulation of this information had a significant effect on participants’ own numeric estimates. Also, exit questionnaire data revealed that 83.6% of the participants were able to accurately recall the peers’ numeric estimate.
5 The correlation between scores for objective and subjective numeracy was r = .39 (p < .001), between objective numeracy and need for cognition was r = .16 (p < .01), and between subjective numeracy and need for cognition was r = .37 (p < 001).
6 A challenge for this speculation is that numeracy did not significantly interact with the verbal estimate factor to influence participants’ numeric or verbal estimates.