AI bias solved? New study proposes radical fairness framework