Interpretability

An Introduction to Machine Learning Interpretability

An Applied Perspective on Fairness, Accountability, Transparency, and Explainable AI

Written by O'Reilly, a computer publisher and a provider of Computer events, and Dataiku, a company I never heard of before.

It was written with some exotic word processing techniques that generated an outline along the left margin. 

Interpretability: what a strange word! I had to look it up in a dictionary to see if it was real. 

But in the context of Artificial Intelligence (AI) it is enormously important, as this paragraph explains:

Machine learning promises quick, accurate, and unbiased decision making in life-changing scenarios. Computers can theoretically use machine learning to make objective, data-driven decisions in critical situations like criminal convictions, medical diagnoses, and college admissions, but interpretability, among other technological advances, is needed to guarantee the promises of correctness and objectivity. 

Without interpretability, accountability, and transparency in machine learning decisions, there is no certainty that a machine learning system is not simply re-learning and re-applying long-held, regrettable, and erroneous human biases. Nor are there any assurances that human operators have not designed a machine learning system to make intentionally prejudicial decisions.

This addresses the question "Can we trust Machine Learning?" Strange things are going on here, that we do not understand. And the more we learn how this is done, the harder it is to trust it.