Getting Uncomfortable with Data
Many data science speakers preach to the choir. They tell us that data science is the sexiest job of the 21st century, or that data is eating the world. Others try to offer practical advice — I generally put myself in this category. We provide tips and tricks across the stack — from building infrastructure to building teams.
But it’s a rare data scientist who challenges our core values and exhorts us to get uncomfortable with the fundamental tools of our trade.
A few days ago, Cloudera hosted Wrangle, a conference for and by data scientists. The talks were consistently excellent, filled with war stories from some of the industry’s top companies in the field.
She made several points that I hope every data scientist internalizes:
- Algorithms can hurt people. We may think we live in a world of software and data, but the work we do can cause harm in the real world. It’s our job to mitigate that harm.
- Algorithms scale human judgment. That means they also scale and reinforce human bias. As data scientists, we may be part of the problem. Instead, we have to be part of the solution.
- Remember the people on the other end of your algorithm. For example, if you’re working on credit scoring or facial recognition algorithms, you can’t ignore how those algorithms may be applied.
- We need to construct fairness. When bias is deeply baked into your data, it’s your job to explicitly construct fairness into your approach. This is hard, but not impossible.
- If you don’t build it, maybe no one will. You can drive up the market price of creating harmful technology.
We data scientists create algorithms in our own images. It’s a god-like power, and we can’t let that get to our heads. With great power comes great responsibility. A responsibility to create the world we want to live in.