Give them PROOF

A data scientist or machine teacher might reasonably be expected to answer the question

"How do we know what led the machine to reach each decision?"

This feels like a fair request to ask of a neural net that could decide if your tires stay on the road, the length of criminal sentences and the composition of prescription medications.

It's so clear when you explain it like that.

Much like humans, however, the more we learn about the way they work the more confident experts are in asserting the vastness of the undiscovered.

We are flawed decision makers, teaching our invention based on data from our flawed past. As we pour resources into the new age of computation, let us teach first from our mistakes, the ways we err'ed, mistreated each other, and made systems that are incapable of distributing resources to meet need. That is why the paper on teaching models to avoid forgetfulness behaviors was so promising.

You can read "Overcoming Multi-model Forgetting " here: https://arxiv.org/pdf/1902.08232.pdf