Probabilistic Expression¶

Why It Matters¶

Verbal probability terms are ambiguous - "likely" means 30-80% depending on who's speaking. Binary predictions (it will/won't rain) destroy information about confidence and make calibration impossible. Single-point predictions with no error bars imply false precision. Probabilistic expression forces you to quantify your uncertainty and makes your predictions testable against reality.

The Key Move¶

Convert any qualitative prediction to a number between 0-100%. If that feels uncomfortable (it should), that discomfort is information about your actual uncertainty. Force yourself to stake a specific probability, then track whether events at "70% confidence" actually happen 70% of the time. Calibrate your probability words to numbers and use them consistently. When presenting to others, resist the pressure to collapse probabilities into yes/no.

Classic Application¶

Precipitation forecasting. Instead of "rain or no rain," meteorologists give Probability of Precipitation (PoP): the chance that at least 0.01 inches falls at a point in the forecast area. "40% chance of rain" means rain occurs 4 out of 10 times with this pattern. This allows users to make rational decisions (outdoor wedding, yes or no?) based on their own risk tolerance.

Surprising Application¶

Project management. Instead of "will we hit the deadline?" or "estimated completion date," express as probability distribution: 20% chance we finish early, 50% on time, 25% one month late, 5% cancelled. This enables decision-makers to plan contingencies proportional to risk. A "90% on time" project needs different preparation than "60% on time." The probability forces honest assessment of risks usually hidden in single-date estimates.

Failure Modes¶

False precision - stating "73.4% confidence" when your true uncertainty is much wider creates illusion of rigor. Misunderstanding conditional probability - "60% rain" doesn't mean "it will rain 60% of the day" or "60% of the area gets rain." Probability without clear reference class - 70% likely compared to what base rate? Treating probabilities as preferences - "I give this 90%" when you mean "I hope this happens" rather than a calibrated forecast.

Go Deeper¶

Tetlock & Gardner, Superforecasting, Chapters 3-4; Murphy, "What Is a Good Forecast? An Essay on the Nature of Goodness in Weather Forecasting," Weather and Forecasting (1993).