Microsoft open sources InterpretML, a python package it says will help find bias and regulatory issues in machine learning frameworks
Microsoft is the latest developer to open source a tool attempting to bring explainability to complex, hard-to-understand machine learning processes.
InterpretML is built with an algorithm Microsoft calls the Explainable Boosting Machine (EBM), which Microsoft calls a fast implementation of algorithms it's previously used to introduce explainability into machine learning medical tools.
The goal, according to Microsoft Research, is to make the most advanced and potentially dangerous tools intelligible:
In machine learning, there is often a tradeoff between accuracy and intelligibility: the most accurate machine learning models usually are not very intelligible (for example, deep neural nets, boosted trees, random forests, and support vector machines), and the most intelligible models usually are less accurate (for example, linear or logistic regression).