97 Things About Ethics Everyone in Data Science Should Know

Chapter 66. Automatically Checking for Ethics Violations

Jesse Anderson

Managing Director, Big Data Institute

Data science likes to be quite meta sometimes. There is a push to use machine learning models to check the actions of data scientists or other machine learning models for ethics violations. The watchers are being watched by the machine learning model.

I’m often asked whether it’s really possible for a machine learning model to automatically check for ethics violations. This question usually comes from companies that are worried about the sheer number of queries and overall discovery that a data scientist needs to run. With data democratization, even more people will have access to the data, which means even more possible ethics violations. It will be virtually impossible for the management team or general counsel to review every single query.

In my opinion, it isn’t possible to rely on a machine learning model to find ethics violations. The people who write the machine learning model are the same ones the machine would be watching for potential violations. If they aren’t the ones who wrote it, they will have enough of a background to know how not to have their query be flagged as an ethics violation. The data scientists especially will be able to make educated guesses as to which algorithms are being used and will know what the weakness of each algorithm is.

Despite all ...