It’s a bit debatable if this is really a “data” problem that could be solved in principle by gathering enough data or more of a “philosophical” problem where we aren’t exactly sure what bias means.