Some of the marginalized groups most likely to be harmed by AI are also most wary of it. A new study’s findings raise questions about equity and consent in technology design.