[...]If you are a policeman, clearly you are going to opt for population A if you want to make a bust.[...]
Whether I have to make 1000 controls for a 39 percent chance to catch one dealer, or 1000 controls for a 95 percent chance to catch one dealer – both are incredibly ineffective. This is exactly the problem.
If I am only capable of random controls with low chances, I have to control very, very many people to make a hit (and each control of an innocent person is something, that should be avoided if possible, because it is a form of harassment). Now by going from one low probability to a somewhat less lower probability by ignoring the group of the lower probability and putting all the burden of unjustified control to the other group you create a huge sense of frustration, stress, injustice and anger. For good reason! You make a 60:40 relation to a 100:0 relation with this approach. The problem is not with the dealers, but with the false positives. 600 innocent people of group A have to be harassed for one true positive, but 0 innocent people of group B get harassed and 0 people of group B get busted, because they are not even controlled anymore (as hits are less likely). And now, by making hits only in group A, the ratio of convicted drug dealers gets pushed even more into the direction of group A, allegedly confirming the efficiency of racial profiling. It is utterly wrong. Morally and mathematically. It is a pseudologic abuse of science to discriminate a group of people. And the desire for discrimination arises from hate. That is, why racial profiling is forbidden in modern democracies and it is not a matter of free speech, in my eyes.
If you wish for a more efficient handling of your police with drug dealers, you really do not want them to perform random controls (whether racially biased or not)!
Correct, but in reality, the police doesn't go do some random control in the streets of some randomly selected rural area. They will target locations where they are likely to find drug dealers, target behaviors that are likely to be drug dealers, etc. So the numbers aren't those from your theoretical example.
Though in reality, I mostly hear about racial profiling in France in the context of looking for illegal immigrants where the odds are even more skewed against a population than your example.
The moral argument is orthogonal from the efficiency argument, and I totally agree with the frustration generated by misguided checks (and am reminded of those every time I take a plane).
Whether I have to make 1000 controls for a 39 percent chance to catch one dealer, or 1000 controls for a 95 percent chance to catch one dealer – both are incredibly ineffective. This is exactly the problem.
If I am only capable of random controls with low chances, I have to control very, very many people to make a hit (and each control of an innocent person is something, that should be avoided if possible, because it is a form of harassment). Now by going from one low probability to a somewhat less lower probability by ignoring the group of the lower probability and putting all the burden of unjustified control to the other group you create a huge sense of frustration, stress, injustice and anger. For good reason! You make a 60:40 relation to a 100:0 relation with this approach. The problem is not with the dealers, but with the false positives. 600 innocent people of group A have to be harassed for one true positive, but 0 innocent people of group B get harassed and 0 people of group B get busted, because they are not even controlled anymore (as hits are less likely). And now, by making hits only in group A, the ratio of convicted drug dealers gets pushed even more into the direction of group A, allegedly confirming the efficiency of racial profiling. It is utterly wrong. Morally and mathematically. It is a pseudologic abuse of science to discriminate a group of people. And the desire for discrimination arises from hate. That is, why racial profiling is forbidden in modern democracies and it is not a matter of free speech, in my eyes.
If you wish for a more efficient handling of your police with drug dealers, you really do not want them to perform random controls (whether racially biased or not)!