[protomyth]

The first is that most of the algorithms they’re using are probably based heavily on machine learning, making them inscrutable not just to the general public but also to any experts they have auditing them.

I still think this is going to be a big problem with regards to legislation and safety boards. I'm pretty sure they are not going to accept the black box decision making nature. They see the results of prejudice or unsafe decisions, real or imagined[1], and will probably see lack of ability to explain these actions clearly as hubris on the part of companies. The algorithm did it will not hold as an excuse.

1) iPod shuffle play randomness