"It is important to maintain the discipline of not trying to explain too much,” said Max Levchin, chief executive of Affirm. Adding human assumptions, he noted, could introduce bias into the data analysis.

This might be true, but I've heard this mentality before, and I feel like it draws this incorrect picture of completely fair robots scanning the universe of variables. Ultimately it's a [probably-white-upper-middle-class] human who intuits which variables are worth further study.

[probably-white-upper-middle-class]: pick your favorite adjectives to describe startup machine learning engineers.