Is algorithmic transparency and accountability necessary and/or feasible?