Tweaking the models to lead towards the world we'd like to live in seems like a noble goal, but I wonder how you should incentivize this. If discriminatory models provide higher profits, the idealistic models won't be used.