97 Things About Ethics Everyone in Data Science Should Know

Chapter 84. Ethical CRISP-DM: A Framework for Ethical Data Science Development

Collin Cunningham

Data Scientist, Amazon Web Services

Good data science creates the illusion of something human; something beyond a cold, colorless process that disallows empathy. The goal of a model, however, is singular: make decisions that have previously minimized loss functions (or something equally mechanical). Therefore, we must systematically enforce empathy and ethics where there is none.

The cross-industry standard process for data mining, more commonly referred to as CRISP-DM, is a widely used methodology in analytics development. The steps of CRISP-DM are:

• Business understanding

• Data understanding

• Data preparation

• Modeling

• Evaluation

• Deployment

Although CRISP-DM was developed for data mining, successful data science projects knowingly or unknowingly follow these procedures in some way. To make more ethical decisions in handling data, we can augment this process by considering a question at each step. By doing so, we create a concrete ethical framework for doing data science.

Business Understanding

What are potential externalities of this solution? Every successful data science project must start with an understanding of the problem as well as of the environment within which it exists. This is the foundational step in positioning a project for success in terms of both effective modeling ...