Topic of race becomes a lot fuzzier when you have the option to select languages. For instance, I speak Spanish and English which gives me the option to select the language for interacting with the device. For English, I expect an English speaking person in the right context of country of origin (British vs. US vs. Australian, for example); the same would be to Spanish between Latin American or from somebody from Spain.

However, I do not visualize a person’s race when I make the selection. I just care about understanding the information that is provided while I am using the device. In my opinion, proper diction is more important than accents that are related to a given ethnic group.

I guess you could have the option to personalize it a bit further than gender and be able to select between accents that relate to different races in addition to gender, but it would be something that could further enhance your experience as a user rather than an essential usability feature.