3 ms·
Why would that argument not apply to "It's 1960, and I don't want to seat black customers at my fancy restaurant because it's statistically likely to be less pr
by geofft 10y ago
Why would that argument not apply to "It's 1960, and I don't want to seat black customers at my fancy restaurant because it's statistically likely to be less profitable than seating white customers"?
- blfr 10y agoBecause black customers pay the same prices and do not require additional accommodations. It would apply if the restaurateur was required to maintain separate dining rooms for whites and blacks or additional water fountains.
- geofft 10y agoEveryone pays the same prices for what they order, but they can choose to order more or less expensive foods, more or less expensive wines (or none at all), etc., and they can tip differently. If you have more customers trying to make reservations than tables to seat them, it is economically rational to seat the customers most likely to buy your more expensive items. We decided (correctly, IMO) that it is more important for society to fight entrenched racism than to enable business owners to make that particular locally-rational economic decision.
- reinhardt 10y agoIt would apply just as well, hence the "Why not the government?". You probably don't want to be the first/only fancy restaurant to fight racism.
- m4chth1s 10y agoI've never worked in the restaurant industry, but, I've heard from people who have that it's STILL extremely common for black customers to be considered undesirable. I.e. a server will get upset at a host(ess) if they get "too many" black customers assigned to their tables. Why? Servers claim that black customers - on average - tip less. I don't know how prevalent this belief on the part of servers is. And I don't know how much of a numerically accurate basis in reality it has. But, as a thought experiment, imagine for a moment that there exists an identifiable group [black people, star trek fans, old people, dentists, ...] that are likely to lead a server to go home with less money at the end of the night. There can be a really fine line between statistics/machine learning and "bias"/"racism". Where it gets even more complicated is that there might be variables which are not expressed as "is the customer black" which ultimately predict/are highly correlated with the question "is the customer black" - but are also predictive of the question "is the customer profitable".
- thaumasiotes 10y ago> There can be a really fine line between statistics/machine learning and "bias"/"racism". There's no line. Learning and bias are the same thing.
- Dylan16807 10y agoThe question is whether the bias is actually justified by experience.
- geofft 10y agoShould it be legal to discriminate against a race if you have data indicating that they're economically disadvantaged?
- Dylan16807 10y agoSociety's answer to that is a resounding "sometimes".
- geofft 10y agoI don't think there's a line between machine learning and bias. In the example I gave (black Americans in the 1960s), I'm pretty confident that any working machine-learning algorithm would conclude that this population is economically disadvantaged and therefore a less-profitable target market. Such a conclusion would also reinforce structural racism. And that's the way we treated this in the 1960s. The government didn't ask Woolworth whether they had legitimate economic reason to refuse to seat those black college students, or whether their objection was based on them being college students, or because they only ordered coffee and not a meal, or anything else. As long as the effect of your policy was that you were discriminating on race, it became illegal. It doesn't matter whether you had valid reasoning based on race and economic status being correlated, or you were a card-carrying Klansman. A good-faith policy that discriminated on race is still illegal. It seemed to work pretty well, and I don't think many of us regret the loss of ability of restaurateurs to make these particular economic decisions. The fact that we can now offload these decisions to algorithms (which aren't likely to be "racist" in the conventional sense) instead of humans shouldn't change things.