Thus, whenever you look in a computer science textbook for an algorithm which only gives approximate results, you will find that the algorithm itself is very vaguely specified, since the result is just an approximation anyway.
(I would have said: “When a concept is inherently fuzzy, it is a waste of time to give it a definition with a sharp membership boundary.”)
(I would have said: “When a concept is inherently fuzzy, it is a waste of time to give it a definition with a sharp membership boundary.”)
Thus we merely require citizens to “be responsible adults” before they can vote rather than give a sharp boundary such as 18 years old, college applications tell you “don’t write a long, rambling essay” rather than enforce a 500-word limit, and food packaging specifies “sometime in September” for the expiration date.
Sharp membership boundaries are useful to make it easy to test for the concept. Even if the concept is fuzzy and the test is imperfect, this doesn’t need to be a waste of time.
Sharp membership boundaries, however, often result in people forgetting the fuzziness of the concept—there are some people who vote without being responsible adults, because they can; an essay can be boring and rambling at 450 words or impressive and concise at 600; and food can be good a bit past its expiration date (it doesn’t usually go in the other direction in my experience, presumably because the risk of eating spoiled food vastly outweighs the risk of mistakenly tossing out good food, so expiration dates are the very early estimates).
Though sometimes it’s even more useful to acknowledge that the sharp-boundaried concept we’re testing for is different from, though perhaps expected to be correlated with in some way, the fuzzy concept we were initially interested in.
That helps us avoid the trap of believing that 17-year-olds aren’t responsible adults but 18-year-olds are, or that 550-word essays are long and rambling but 450-word essays aren’t, or that food is safe to eat on September 25 but not on September 29. None of that is true, but that’s OK; we aren’t actually testing for whether voters are responsible adults, essays are long and rambling, or food is expired.
Well, sure. But that doesn’t mean it’s very strong evidence: I’d expect to see an average human (or nation) do something stupid almost as often as they do something intelligent.
We are obviously starting from very different premises. To me, the fact that lots of people do something is very strong evidence that the behaviour is, at least, not maladaptive, and the burden of proof is very much on the person suggesting that it is. And the more widespread the behaviour, the stronger the burden.
Alternatively, you could just look at the evidence. When legal systems have replaced bright-line rules with 15-factor balancing tests, has that led to better outcomes for society as a whole? Consider in particular the criteria for the Rule of Law. In the mid-20th century, co-incident with high modernism and utilitarianism, these multi-part, multi-factor balancing tests were all the rage. Why are they now held in such disdain?
Unfortunately, the fact that lots of people do something may merely be an indication of a very successful meme: consider major religions.
I will certainly grant that having a sharp restriction is better than a 15-factor balancing test, but I’m not arguing for 15-factor balancing tests.
I’d go further, but I’ve just noticed that I don’t really have much evidence for this belief, and I should probably go see how accomplished Chinese universities (which judge purely off the gaokao) are versus American universities first.
When a concept is inherently approximate, it is a waste of time to try to give it a precise definition.
-- John McCarthy
Thus, whenever you look in a computer science textbook for an algorithm which only gives approximate results, you will find that the algorithm itself is very vaguely specified, since the result is just an approximation anyway.
(I would have said: “When a concept is inherently fuzzy, it is a waste of time to give it a definition with a sharp membership boundary.”)
Thus we merely require citizens to “be responsible adults” before they can vote rather than give a sharp boundary such as 18 years old, college applications tell you “don’t write a long, rambling essay” rather than enforce a 500-word limit, and food packaging specifies “sometime in September” for the expiration date.
Sharp membership boundaries are useful to make it easy to test for the concept. Even if the concept is fuzzy and the test is imperfect, this doesn’t need to be a waste of time.
Sharp membership boundaries, however, often result in people forgetting the fuzziness of the concept—there are some people who vote without being responsible adults, because they can; an essay can be boring and rambling at 450 words or impressive and concise at 600; and food can be good a bit past its expiration date (it doesn’t usually go in the other direction in my experience, presumably because the risk of eating spoiled food vastly outweighs the risk of mistakenly tossing out good food, so expiration dates are the very early estimates).
Though sometimes it’s even more useful to acknowledge that the sharp-boundaried concept we’re testing for is different from, though perhaps expected to be correlated with in some way, the fuzzy concept we were initially interested in.
That helps us avoid the trap of believing that 17-year-olds aren’t responsible adults but 18-year-olds are, or that 550-word essays are long and rambling but 450-word essays aren’t, or that food is safe to eat on September 25 but not on September 29. None of that is true, but that’s OK; we aren’t actually testing for whether voters are responsible adults, essays are long and rambling, or food is expired.
Just because humans do it doesn’t mean it’s a good idea.
To clarify, I also think all of these are good ideas; not necessarily the best possible, but definitely useful.
It doesn’t prove it’s a good idea, but it’s evidence in its favour.
Well, sure. But that doesn’t mean it’s very strong evidence: I’d expect to see an average human (or nation) do something stupid almost as often as they do something intelligent.
We are obviously starting from very different premises. To me, the fact that lots of people do something is very strong evidence that the behaviour is, at least, not maladaptive, and the burden of proof is very much on the person suggesting that it is. And the more widespread the behaviour, the stronger the burden.
Alternatively, you could just look at the evidence. When legal systems have replaced bright-line rules with 15-factor balancing tests, has that led to better outcomes for society as a whole? Consider in particular the criteria for the Rule of Law. In the mid-20th century, co-incident with high modernism and utilitarianism, these multi-part, multi-factor balancing tests were all the rage. Why are they now held in such disdain?
Unfortunately, the fact that lots of people do something may merely be an indication of a very successful meme: consider major religions.
I will certainly grant that having a sharp restriction is better than a 15-factor balancing test, but I’m not arguing for 15-factor balancing tests.
I’d go further, but I’ve just noticed that I don’t really have much evidence for this belief, and I should probably go see how accomplished Chinese universities (which judge purely off the gaokao) are versus American universities first.