Monday, August 3, 2009

Quantifying Uncertainty

Scientists and engineers are familiar with quantifying uncertainty in situations where we have data and measurements available. If we know the measurement standard deviation and the number of measurements we made, we can calculate a confidence interval in which the true answer lies.

Things are more complicated when we are asked to make a judgment call. For example, what if you are asked when you will get a report finished. You might answer, "probably tomorrow". The problem with this is that "probably" can mean something quite different to different people. "Probably" to you might mean "maybe" to another. Some will use the word "unlikely" instead of "might" or "might not".

Although it will seem uncomfortable, try to put some sort of odds on your judgment calls. If you answer, "there's a 90% chance I'll be finished tomorrow" it's a lot less ambiguous than "I'll probably be finished tomorrow". Even though the odds might only be 70% or 80%, they're certainly not 20% or even 50%. Attempting to apply some sort of numerical as opposed to verbal uncertainty will make your communication clearer.

Somewhere I've come across a study of what various verbal uncertainty terms meant to different people. I'll try to find it again and post it here.

No comments: