3 ms·

You should read about the CIA's study about words of estimative probability : https://waf.cs.illinois.edu/visualizations/Perception-of-Probability-Words/ https:

You should read about the CIA's study about words of estimative probability : https://waf.cs.illinois.edu/visualizations/Perception-of-Probability-Words/ https://waf.cs.illinois.edu/visualizations/Perception-of-Pro... Basically, when you use a term like "seemingly likely", you do have a number range in mind. Always. Be it 50%-70% or 40%-60%. It's impossible not to. And it became a problem for CIA agents communicating with each other, because different people have different sense of what numbers fit a given term. It's better to be clear about what you mean by giving explicit numbers. It's understood in this sort of context anyway that your number is not calculated from anything, that it's only the rough middle of a range that sounds right to you.

bee_rider 5y ago
> Basically, when you use a term like "seemingly likely", you do have a number range in mind. Always. Be it 50%-70% or 40%-60%. It's impossible not to. I don't see how this is supported in the article. And the experiments they discussed -- asking people what probability they assign to certain expressions -- can't possibly support this statement. If you ask people to assign a number to an expression, they'll give you a number, but that doesn't mean they were originally thinking of one, or that the one they come up with is meaningful. > And it became a problem for CIA agents communicating with each other, because different people have different sense of what numbers fit a given term. It's better to be clear about what you mean by giving explicit numbers. It's understood in this sort of context anyway that your number is not calculated from anything, that it's only the rough middle of a range that sounds right to you. My experience is with these sorts of number-like estimate things is that I should try to figure out what plain English phrase they are trying to over-specify. I mostly come to this conclusion by talking about current events with people (elections outcomes that they thing are sure things, for example), and comparing the numbers reported (by 538 for example) to odds that they are familiar with (D&D critical hits are a good one, as an event that is rare but memorably encountered). This tends to lower confidence, which indicates to me that people don't really have a feel for these numbers. I don't really think 70% and 60% are different from an operational point of view anyway. They are both events where we think the outcome has some skew but must plan on seeing the less likely outcome pretty often.

bee_rider 5y ago

> Basically, when you use a term like "seemingly likely", you do have a number range in mind. Always. Be it 50%-70% or 40%-60%. It's impossible not to. I don't see how this is supported in the article. And the experiments they discussed -- asking people what probability they assign to certain expressions -- can't possibly support this statement. If you ask people to assign a number to an expression, they'll give you a number, but that doesn't mean they were originally thinking of one, or that the one they come up with is meaningful. > And it became a problem for CIA agents communicating with each other, because different people have different sense of what numbers fit a given term. It's better to be clear about what you mean by giving explicit numbers. It's understood in this sort of context anyway that your number is not calculated from anything, that it's only the rough middle of a range that sounds right to you. My experience is with these sorts of number-like estimate things is that I should try to figure out what plain English phrase they are trying to over-specify. I mostly come to this conclusion by talking about current events with people (elections outcomes that they thing are sure things, for example), and comparing the numbers reported (by 538 for example) to odds that they are familiar with (D&D critical hits are a good one, as an event that is rare but memorably encountered). This tends to lower confidence, which indicates to me that people don't really have a feel for these numbers. I don't really think 70% and 60% are different from an operational point of view anyway. They are both events where we think the outcome has some skew but must plan on seeing the less likely outcome pretty often.