Users follow prompts, so "people choose a race when presented with a racialized emoji picker" doesn't mean "people want their emojis racialized". It means, people understand what they are being asked to do, and want to get back to their business.

You would need to have a theory of benefit, and then design a survey which tests users to understand if that benefit was achieved. You, frankly, are confused about causation -- we put a dialog in front of people and they made a selection, therefore people wanted this dialog?