| Summary: | The public expects science to reduce or eliminate uncertainty (Kinzig & Starrett, 2003), yet scientific forecasts are probabilistic (at best) and it is simply not possible to make predictions with certainty. Whilst an ‘unlikely’ outcome is not expected to occur, an ‘unlikely’ outcome will still occur one in five times (based on a translation of 20%, e.g. Theil, 2002), according to a frequentist perspective. When an ‘unlikely’ outcome does occur, the prediction may be deemed ‘erroneous’, reflecting a misunderstanding of the nature of uncertainty. Such misunderstandings could have ramifications for the subsequent (perceived) credibility of the communicator who made such a prediction. We examine whether the effect of ‘erroneous’ predictions on perceived credibility differs according to the communication format used. Specifically, we consider verbal, numerical (point and range [wide / narrow]) and mixed format probability expressions. We consistently find that subsequent perceptions are least affected by the ‘erroneous’ prediction when it is expressed numerically, regardless of whether it is a point or range estimate. Our findings suggest numbers should be used in consequential risk communications regarding ‘unlikely’ events, wherever possible.
|
|---|