Eh, I think from a Bayesian standpoint its valid to think that if somebody has a last name that correlates with a group that doesn't drink, and that person also

Eh, I think from a Bayesian standpoint its valid to think that if somebody has a last name that correlates with a group that doesn't drink, and that person also argues against drinking that they're more likely to be a member of that group. But that's the key part "more likely", and not something definitive. It's the thinking that that inference is definitive is where it crosses the line from trying to be understanding about another culture to becoming a racist argument. I agree though that its definitely flirting with that dangerous assumption.