Mauboussin on the Language of Likelihood

In a recent article for the Harvard Business Review, authors Andrew and Michael Mauboussin discuss the language of probability and share analysis that shows how its subjective nature can lead to “widely different interpretations.”

“Today people in the worlds of business, investing, and politics continue to use vague words to describe possible outcomes,” the article says, citing an explanation by psychology professor Phil Tetlock: “vague verbiage,” he says, “gives you political safety.”

The authors created a survey to track the relationship between words and probabilities akin to one created by Yale professor Sherman Kent in the 1950s for the intelligence community (surrounding the potential for a Soviet attack on Yugoslavia). Messrs. Mauboussin created an updated survey with a view toward increasing the sample size (to include individuals outside of the intelligence and scientific communities) and to see whether we could detect any differences by age or gender or between those who learned English as a primary or a secondary language.”

The authors shared the following three main takeaways:

  1. “Use probabilities instead of words to avoid misinterpretation.” In general, the authors found that the word “possible” and variations thereof “have wide ranges and invite confusion.”
  2. “Use structured approaches to set probabilities:” The article notes the importance of updating “subjective probability estimates” whenever new, relevant information is obtained.
  3. “Seek feedback to improve your forecasting.” When forecasts are made, it’s important to keep track of how accurate they prove to be. “Opinion writers and public intellectuals often talk about the future, but typically they don’t express their convictions precisely enough to allow for accurate performance tracking,” the article argues. The best forecasters, it says, check their track records with a metric (such as a Brier score).

The article concludes: “The next time you find yourself stating that a deal or other business outcome is ‘unlikely’ or, alternatively, is ‘virtually certain,’ stop yourself and ask: What percentage chance, in what time period, would I put on this outcome? Frame your prediction that way, and it’ll be clear to both yourself and others where you truly stand.”