The New Notion of Fairness in Predictive Analectics

There are allegations that amazon is promoting its private label products over other third party sellers on the same market place. If i were asked to defend amazon, i guess maybe i could say, we have a ranking algorithm at amazon, and it includes, we can't tell you what it is, it's our secret,. But a large portion of it is based on conversion raid or sales. And it just so happens that amazon base excels really well. Is that sufficient? Or can we learn anything more than that? Actually, there is a lot more to learn than that.