I struggle with the framing of “algorithmic bias.” As if it were possible to develop algorithms without bias. The whole point is to create bias based on data. Do you know if anyone is taking on the question: what kinds of bias do we want, what kinds do we not want and why?