There are some contexts in which the difference between 99.999% and 99.9% is about the same as the difference between 10% and 90%. However, I do not think this is one of them. I repeat: the letter you are talking about did not say anything about probabilities; it said “some scientific theories have a shedload of evidence and are well enough established that we can reasonably call them facts; here are some familiar examples; well, global warming is also in that category”.
I think it’s probably 100x more certain that the earth is more than a few thousands of years old than that our universe began with a big bang ~14Gya. Does that mean the people who wrote that letter were wrong to group those together? Nope; all that matters is that both are in the “firmly enough established” category. So, they suggest (and I agree), is global warming.
(Not every detail of global warming. Not any specific claim about what the global mean surface temperature will be in 50 years’ time. But the broad outline.)
The outside view suggests that most of the time experts are a bit overconfident.
The outside view suggests to me that much of the time experts are horribly overconfident, and some of the time they are distinctly underconfident (at least in what they say). The picture doesn’t look to me much like one of consistent slight overconfidence at all.
If those IPCC scientists are that good at not being overconfident, why don’t we tell the psychologists to listen to them to deal with their replication crisis?
Hey, psychologists! Go read the IPCC reports, and follow their example!
There you go. I did. It won’t actually do any good, because the problem isn’t that no one has ever told psychologists to be cautious and avoid overconfidence. And that’s the answer to your question “why don’t we …”, and you will notice that it has nothing to do with the people who wrote the IPCC being overconfident.
the problem isn’t that no one has ever told psychologists to be cautious and avoid overconfidence.
The problem is probably that psychologists afterwards always nod their heads and say: “uhm, uhm, that’s interesting… please tell me more about your feelings of anxiety.”
There are some contexts in which the difference between 99.999% and 99.9% is about the same as the difference between 10% and 90%.
It’s not 99.9% in the IPCC report.
Events that happen with 0.005 probability are worth planning for when they have high impacts. We care about asteroid defense when that probability is much lower.
Humanity has a good chance of getting destroyed in this century if decision makers treat 0.001 the same way as 0.0000000001.
The outside view suggests to me that much of the time experts are horribly overconfident, and some of the time they are distinctly underconfident (at least in what they say)
In what examples are are experts underconfident when they give 0.9 or 0.95 probabilities of an event happening.
I wasn’t trying to suggest it was; my apologies for (evidently) being insufficiently clear.
Events that happen with 0.005 probability are worth planning for when they have high impacts.
Yup, strongly agreed. But here the low-probability events we’re talking about here are things like “it turns out global warming wasn’t a big deal after all”. It would be sad to have spent a lot of money trying to reduce carbon dioxide emissions in that case, but it wouldn’t be much like (e.g.) being hit by an asteroid.
In what examples are experts underconfident when they give 0.9 or 0.95 probabilities of an event happening
I don’t have examples to hand, I’m afraid (other than the IPCC example we’re discussing here, though actually all we know there is that their probability estimate is somewhere between 0.95 and 1, and is probably below 0.99 since they didn’t choose to say “virtually certain”. (Only “probably” because when they list what the terms mean they say 95-100% and not 95-99% for “extremely likely”, and the best explanation I can see for that is that they are reserving the right to say “extremely likely” rather than “virtually certain” sometimes even though they think the actual probability is over 99%. This is one reason why I suspect them of understating their certainty on purpose: they seem to have gone out of their way to provide themselves with a way to do that.)
There are some contexts in which the difference between 99.999% and 99.9% is about the same as the difference between 10% and 90%. However, I do not think this is one of them. I repeat: the letter you are talking about did not say anything about probabilities; it said “some scientific theories have a shedload of evidence and are well enough established that we can reasonably call them facts; here are some familiar examples; well, global warming is also in that category”.
I think it’s probably 100x more certain that the earth is more than a few thousands of years old than that our universe began with a big bang ~14Gya. Does that mean the people who wrote that letter were wrong to group those together? Nope; all that matters is that both are in the “firmly enough established” category. So, they suggest (and I agree), is global warming.
(Not every detail of global warming. Not any specific claim about what the global mean surface temperature will be in 50 years’ time. But the broad outline.)
The outside view suggests to me that much of the time experts are horribly overconfident, and some of the time they are distinctly underconfident (at least in what they say). The picture doesn’t look to me much like one of consistent slight overconfidence at all.
Hey, psychologists! Go read the IPCC reports, and follow their example!
There you go. I did. It won’t actually do any good, because the problem isn’t that no one has ever told psychologists to be cautious and avoid overconfidence. And that’s the answer to your question “why don’t we …”, and you will notice that it has nothing to do with the people who wrote the IPCC being overconfident.
The problem is probably that psychologists afterwards always nod their heads and say: “uhm, uhm, that’s interesting… please tell me more about your feelings of anxiety.”
:D
It’s not 99.9% in the IPCC report.
Events that happen with 0.005 probability are worth planning for when they have high impacts. We care about asteroid defense when that probability is much lower.
Humanity has a good chance of getting destroyed in this century if decision makers treat 0.001 the same way as 0.0000000001.
In what examples are are experts underconfident when they give 0.9 or 0.95 probabilities of an event happening.
I wasn’t trying to suggest it was; my apologies for (evidently) being insufficiently clear.
Yup, strongly agreed. But here the low-probability events we’re talking about here are things like “it turns out global warming wasn’t a big deal after all”. It would be sad to have spent a lot of money trying to reduce carbon dioxide emissions in that case, but it wouldn’t be much like (e.g.) being hit by an asteroid.
I don’t have examples to hand, I’m afraid (other than the IPCC example we’re discussing here, though actually all we know there is that their probability estimate is somewhere between 0.95 and 1, and is probably below 0.99 since they didn’t choose to say “virtually certain”. (Only “probably” because when they list what the terms mean they say 95-100% and not 95-99% for “extremely likely”, and the best explanation I can see for that is that they are reserving the right to say “extremely likely” rather than “virtually certain” sometimes even though they think the actual probability is over 99%. This is one reason why I suspect them of understating their certainty on purpose: they seem to have gone out of their way to provide themselves with a way to do that.)