How likely is ‘likely’?

The inherent ambiguity of probability language in news and intelligence reporting often leads to significant misinterpretations of risk. Adam Kucharski, a professor of epidemiology and author of *Proof*, explains that while numerical probabilities like 80% offer clarity, verbal descriptors such as "realistic possibility" are highly subjective. Historical precedents, such as a 1951 CIA report on a potential Soviet invasion, demonstrate how intelligence analysts and policymakers can hold wildly different numerical interpretations of the same phrase. Data from a study of over 5,000 participants reveals that while terms like "about even" or "almost certain" have high consensus, "realistic possibility" is the most volatile, with interpretations ranging from 10% to near certainty. Standardizing these terms through "probability yardsticks" and demanding numerical accountability from experts can prevent communicators from using vague language to avoid responsibility for inaccurate predictions.