Your belief implies it is improper to use statistics to describe morally salient human behavior in any way, unless the conclusion is that there is no variation. I think that's silly and intellectually impoverishing. Groups have different average behavior and I don't see the point in blinding myself to that.
We all know that averages don't determine an individual, and we all know that some generalizations should be avoided, but that doesn't imply some broad fatwa, completely irrespective of context, against noting broad variations in group behavior.
The prosecutor attitude is very entertaining though. Feel free to keep doing that.
Point taken on "prosecutor attitude", sorry about that. riding "moral high horse" got to my head.
For me, it's not an easy topic but I would say when negative generalizations escalate to exclusion and dehumanization, that's probably where to draw the line.
I know some people with Audis and I don't think worse of them or anything like that. It's just a funny thing that's hard to ignore after a while on the road.
Regarding my personal beliefs, as I've stated above "I personally believe in statistics and I believe you cannot use* them to make derogatory* generalizations"
1. regarding "use": you cannot use specifically statistics/data to justify any *-ist remark. "I hate bmw drivers because most of them are assholes on the road", an example of that (not what you said!)
2. regarding "derogatory": you can(and should!) use statistics to do (just) generalizations, i.e. in a context of talking about group behaviour. "most of bmw drivers are agressive and dangerous drivers, as shown by this data" is a perfectly fine statement.
My problem is that you can say that sort of statements about certain group of drivers, but if you would have pulled up a similar argument and used (for example) PoC and crime rates, you would have been torn to pieces by everyone. And for me, on the surface, the statement and the structure would be absolutely the same (and it is STILL not escalated to exclusion and dehumanization). So whats the difference? And why different treatment? People above are arguing that buying BMW its a choice and using inherent (immutable) properties (i.e. race/gender) is different for the case of stereotyping, but I cannot see the key difference that makes okay to do one and not okay to do another.
Side point - in general an important thing to remember here is the big, big difference between P(A|B) and P(B|A). It could be the case that all aggressive drivers I ever encounter are BMW drivers, and yet the rate of aggressive BMW drivers could be very low and of little practical predictive value on any individual BMW driver (e.g. 1% of BMW drivers versus 0% for non-BMW drivers).
People are already bad at distinguishing an implication and its converse, and they're quite a bit worse at the statistical version of the same.
Arguably this is a reason we should keep our generalizations to ourselves, even if they're accurate. A lot of people will take them to mean a lot more than they do.
It's not my style, but I do recognize the risk.
Why go around being an enforcer for social rules you don't even know if you believe in?