3 ms·

I don't get it. Why not?

AlexTWithBeard 8y ago
Well, because it does. Let's say we have a group of 1000 animals: 900 rabbits and 100 foxes. We also know that there are about 10 crimes happening every year and for every crime there's 50% chance of it being committed by a fox. Then according to Bayes' formula: P(fox is a criminal) = P(criminal is a fox) * P(animal is a criminal) / P(animal is a fox) = 0.5 * 0.01/0.1 = 0.05 = 5% chance for every given fox to be a criminal. For every given rabbit the same formula gives 0.5 * 0.01/0.9 which is roughly 0.6%.
__s 8y ago
Let's say I show you a muslim. Do we investigate him as a terrorist? We'll assume 100% of terrorists are muslims. Let's say there are 1000 muslim terrorists, & 100000000 muslims. There's a 0.001% chance of finding a terrorist per random sampling. In order to find 0.001% of terrorists you will have to pick through 99999 innocent muslims. This is immensely far from the original 100% number. If you think 1:100000 is worth it, you'll want to extend your model with: 1. Real numbers 2. What's the probability you misclassify someone innocent? 3. What's the probability you misclassify someone guilty? 4. How many people are going to get busted over some unrelated crime like marijuana, which you would find at the same rate if you investigated non-muslims with the same gusto? 2 & 3 will tend to cause you to fail to achieve the original goal. 4 will cause you to create a systematic racist bias edit: I mismathed my random numbers
- sokoloff 8y ago
  > In order to find 0.001% of terrorists you will have to pick through 999 innocent muslims. 99,999, not 999 innocents and 1:100,000, not 1:1000 (making your point stronger)
- AlexTWithBeard 8y ago
  Not every animal in Australia is a kangaroo. Nevertheless, if you want to find a kangaroo in a wild, heading to Australia, would probably be your best bet.
  - ceejayoz 8y ago
    Sure, but if you're looking for kangaroos because of an incorrect belief that they're mostly responsible for something, and it's actually perpetrated more by mountain lions, you might actually be on the wrong continent. Incels have killed more people than Islamic extremists in the US in the last decade.
- dominotw 8y ago
  > that’d say nothing about the likelihood > There's a 0.001% chance of finding a terrorist per random sampling.
- grkvlt 8y ago
  Re the systematic racist bias in 4. - since you postulate a world in which 100% of terrorists are muslims, why would anyone non-muslim ever be investigated for terrorism? (whether with 'gusto' or not)
  - __s 8y ago
    Because point 4 is about non-terrorist charges. Pretty much what we find with the black population being biased with charges of drug possession. If searches are applied to blacks 10x as often relative to whites, then if neither is more likely to possess marijuana you'll come up with statistics which say 10x as many blacks illegally possess marijuana than whites. When instead one needs to say what percentage of those searched are in possession
    - sangnoir 8y ago
      > if neither is more likely to possess marijuana Research says drug usage rate is similar for black and white people in the US, but prosecution and conviction rates are not[1] 1. http://www.hamiltonproject.org/charts/rates_of_drug_use_and_sales_by_race_rates_of_drug_related_criminal_justice http://www.hamiltonproject.org/charts/rates_of_drug_use_and_...
    - grkvlt 8y ago
      No, point 4 was about the fact that somehow it is racist to prosecute people of type X for crimes they commit if those crimes are discovered because of an investigation into a different type of crime (which only type X people commit) - the alternative being to give them a free pass, I assume? The premise doesn't hold water anyway, since even in this hypothetical world where only type X are terrorists, I assume terrorism is still a rare event? So, the extra crimes discovered in the course of terrorism investigations into type X suspects will be dwarfed by the investigations into normal crimes in the whole population during the course of normal every day police work. // Assuming the police are not incompetent, of course, which may not hold in your hypothetical world where they investigate non type X people for terrorism even though they commit zero percent of terrorist attacks - probably out of some misguided diversity quota or affirmative action policy, I assume? ;)

Well, because it does. Let's say we have a group of 1000 animals: 900 rabbits and 100 foxes. We also know that there are about 10 crimes happening every year and for every crime there's 50% chance of it being committed by a fox. Then according to Bayes' formula: P(fox is a criminal) = P(criminal is a fox) * P(animal is a criminal) / P(animal is a fox) = 0.5 * 0.01/0.1 = 0.05 = 5% chance for every given fox to be a criminal. For every given rabbit the same formula gives 0.5 * 0.01/0.9 which is roughly 0.6%.

__s 8y ago

Let's say I show you a muslim. Do we investigate him as a terrorist? We'll assume 100% of terrorists are muslims. Let's say there are 1000 muslim terrorists, & 100000000 muslims. There's a 0.001% chance of finding a terrorist per random sampling. In order to find 0.001% of terrorists you will have to pick through 99999 innocent muslims. This is immensely far from the original 100% number. If you think 1:100000 is worth it, you'll want to extend your model with: 1. Real numbers 2. What's the probability you misclassify someone innocent? 3. What's the probability you misclassify someone guilty? 4. How many people are going to get busted over some unrelated crime like marijuana, which you would find at the same rate if you investigated non-muslims with the same gusto? 2 & 3 will tend to cause you to fail to achieve the original goal. 4 will cause you to create a systematic racist bias edit: I mismathed my random numbers

sokoloff 8y ago
> In order to find 0.001% of terrorists you will have to pick through 999 innocent muslims. 99,999, not 999 innocents and 1:100,000, not 1:1000 (making your point stronger)
AlexTWithBeard 8y ago
Not every animal in Australia is a kangaroo. Nevertheless, if you want to find a kangaroo in a wild, heading to Australia, would probably be your best bet.
- ceejayoz 8y ago
  Sure, but if you're looking for kangaroos because of an incorrect belief that they're mostly responsible for something, and it's actually perpetrated more by mountain lions, you might actually be on the wrong continent. Incels have killed more people than Islamic extremists in the US in the last decade.
dominotw 8y ago
> that’d say nothing about the likelihood > There's a 0.001% chance of finding a terrorist per random sampling.
grkvlt 8y ago
Re the systematic racist bias in 4. - since you postulate a world in which 100% of terrorists are muslims, why would anyone non-muslim ever be investigated for terrorism? (whether with 'gusto' or not)
- __s 8y ago
  Because point 4 is about non-terrorist charges. Pretty much what we find with the black population being biased with charges of drug possession. If searches are applied to blacks 10x as often relative to whites, then if neither is more likely to possess marijuana you'll come up with statistics which say 10x as many blacks illegally possess marijuana than whites. When instead one needs to say what percentage of those searched are in possession
  - sangnoir 8y ago
    > if neither is more likely to possess marijuana Research says drug usage rate is similar for black and white people in the US, but prosecution and conviction rates are not[1] 1. http://www.hamiltonproject.org/charts/rates_of_drug_use_and_sales_by_race_rates_of_drug_related_criminal_justice http://www.hamiltonproject.org/charts/rates_of_drug_use_and_...
  - grkvlt 8y ago
    No, point 4 was about the fact that somehow it is racist to prosecute people of type X for crimes they commit if those crimes are discovered because of an investigation into a different type of crime (which only type X people commit) - the alternative being to give them a free pass, I assume? The premise doesn't hold water anyway, since even in this hypothetical world where only type X are terrorists, I assume terrorism is still a rare event? So, the extra crimes discovered in the course of terrorism investigations into type X suspects will be dwarfed by the investigations into normal crimes in the whole population during the course of normal every day police work. // Assuming the police are not incompetent, of course, which may not hold in your hypothetical world where they investigate non type X people for terrorism even though they commit zero percent of terrorist attacks - probably out of some misguided diversity quota or affirmative action policy, I assume? ;)