|Authors||E. Løhre and K. Teigen|
|Title||Stronger forecasts are more certain|
|Afilliation||Software Engineering, Software Engineering|
|Publication Type||Proceedings, refereed|
|Year of Publication||2016|
|Conference Name||ESCON Transfer of Knowledge Conference|
|Place Published||Lisbon, Portugal|
Predictions are made with different degrees of certainty. Take predictions of climate change: while climate scientists are highly certain that the global mean temperature will increase during the 21st century, there is more uncertainty about how much the temperature will increase. So imagine two climate scientists giving different predictions. Scientist A says that the temperature will increase by “at least 1.5°C”, while B says “at least 2.0°C”. Which scientist would you say is more certain? We propose that lay people will often believe that scientist B is more certain, due to an association between prediction strength (i.e., the magnitude of the predicted outcome) and prediction certainty.
Our hypothesis stems from the idea that people often view probabilities as causal tendencies (propensities), so that higher probabilities are associated with stronger outcomes. Previous research (e.g., Keren & Teigen, 2001) has for instance demonstrated that a high probability (90%) prediction of a sports game is seen as more accurate if the team in question wins by a large rather than a small margin. We now demonstrate that the inverse relation also holds, so that people infer degree of certainty from the strength of the predicted outcome.
In three experiments, participants were told that climate change predictions had been revised from T1 (the time of the first prediction) to T2 (the time of the second prediction). In Experiment 1, we informed participants that a prediction had become more certain or more uncertain from T1 to T2. More certain [uncertain] predictions were associated with stronger [weaker] outcomes. In Experiment 2, we found that when the predicted outcome had become stronger after being revised (e.g., a higher sea level rise was predicted at T2 than at T1), people thought the prediction had become more certain than when the predicted outcome had become weaker. We also found in Experiment 2, and replicated with a similar design in Experiment 3, that this “strength effect” influences perceptions of the speaker. Predictions of stronger outcomes make the scientist appear more trustworthy, and make participants believe that scientists’ ability to predict is improving and that the forecast should be relied upon to a greater degree. Hence, these findings not only increase our understanding of how lay people think about probability and uncertainty, they also have clear implications for those involved in the communication of uncertain outcomes.