LAUSR

Objectives Uncertainty is an integral part of science, statistics, and measurement and yet researchers, clinicians and journalists often worry that attempting to communicate uncertainty about scientific facts and numbers will only serve to decrease trust and undermine credibility. However, the effects of the communication of uncertainty – even epistemic uncertainty, around past and present facts and numbers – remains little studied. We are currently undertaking a set of experimental studies to test under what conditions communicating uncertainty about numbers and facts does, and does not, influence comprehension and trust. Method We are currently conducting online experiments comparing different forms of uncertainty communication (no uncertainty, a verbal statement that there is some uncertainty around the estimate, and a numerical range), on different topics (in our pilot study: UK unemployment, the number of tigers in India, and an estimate of global warming). We are measuring people’s reaction to this communication both in terms of their feeling of uncertainty around the number, their trust in the number, and their trust in the source of the number. We also collect demographic information about the participants including their numeracy and educational background. Results As might be expected, the results of our first pilot experiment showed that across all topics people perceived estimates about which uncertainty was communicated as more uncertain than when no uncertainty was communicated. This effect was stronger for verbal uncertainty communication than when a numerical range was presented. In addition, people perceived estimates about which uncertainty was communicated as less reliable and trustworthy, but mostly so for verbal uncertainty communication – the effects for numerical uncertainty communication were small. Interestingly, the communication of uncertainty through a numerical range created no decline in trust in the source of the numbers. This indicates that people distinguish between the numbers themselves and the source in their judgments: whereas the numbers were seen as less trustworthy, in the case of numerical uncertainty communication the trustworthiness of the source was unaffected. There were no significant moderating factors in terms of demographics or numeracy. Conclusions Our initial pilot study results provide a first indication that communicating uncertainty does affect people’s interpretation of numbers and of the organisation or source behind it. We plan further experiments imminently testing the influence of the magnitude of uncertainty, more variations of numerical and verbal forms of communication and then also graphical formats. After those experiments, we plan further testing in real-world settings (eg. online news websites, with patients) and with different audiences to test the generalisability of the conclusions.