Abstract
Research on verbal probabilities and standard scales issued by national and international authorities suggest that only events with probabilities above 60% should be labelled “likely”. We find, however, that when people apply this term to continuous variables, like expected costs, it describes the most likely (modal) outcome or interval, regardless of actual probabilities, which may be quite small. This was demonstrated in six studies in which lay participants (N = 2,228) were shown probability distributions from various domains and asked to generate or to select “likely” outcome intervals. Despite having numeric and graphically displayed information available, participants judged central, low-probability segments as “likely” (as opposed to equal or larger segments in the tails) and subsequently overestimated the chances of these outcomes. We conclude that high-probability interpretations of “likely” are only valid for binary outcomes but not for distributions of graded variables or multiple outcomes.
What is a “likely” amount? Representative (modal) values are considered likely even when their probabilities are low