I think you are absolutely right. The other point that can be made is that P(person is black | name=Jamaal, person is US citizen) is much higher than P(person i

I think you are absolutely right. The other point that can be made is that P(person is black | name=Jamaal, person is US citizen) is much higher than P(person is black | name=Jamaal), since that is a common name in Arabic and in Muslim countries http://en.wikipedia.org/wiki/Jamal http://en.wikipedia.org/wiki/Jamal. One can argue that the extra "a" in the spelling of Jamaal as opposed to the more common transliteration Jamal is a marker but I don't think regular people notice this. Note that the other weirdness is that the common US usage of "black" to commonly refer to African-Americans is very problematic: An arab may look dark skinned but most would never consider themselves "African" or there are African-Americans that have lighter skin colors than, say, people from Sri Lanka. Similarly for the label "white". For example, when we say "white men's advantage" there is an implicit assumption, not always stated or well-understood by users, that this refers to white men in the US, or perhaps more broadly to countries where there's a large population of "non-whites" that have been historically repressed. Personally, when faced with a form that requires me to fill in race, I mark "other" or "does not want to disclose" rather than "white", since I don't feel I fully understand what the connotations of that label are for the purposes of that form.