"Models are opinions embedded in mathematics"

Cathy O’Neil, Weapons of Math Destruction

I am easily distracted. I think I recall once being incredibly efficient. I remember having 12 clients simultaneously on the books with unique demands. No problem. I would work on a few in the morning, more in the afternoon, and if time was slipping away I would balance my laptop on my knees while hanging out with my young family watching a movie or seeing what Thomas the Tank Engine was up to.

Fast forward to now and I have Clubhouse in the background where I think the real creativity fail is when speakers think they aren’t simply hawking their skills or companies. I also write most days while also hopping on and off panels discussing diversity, equity, and equality. The illusion is that I am not compromising the quality of what I am writing or the quality of what I am sharing on Zoom. I am sure I am.

This multi-tasking existence is no longer viable. First of all, I don’t have the patience for it. Listening to leaders in healthcare and medicine discuss racial differences in outcomes but neglect to include meaningful metrics and variables is infuriating. I am a person of color and I received stellar healthcare recently for an open-heart procedure. When we use race as a proxy for something, without bringing other variables to the conversation we are leaving data (and insights) on the table. For example, I have 3 degrees (2 post graduate), a decent amount of subject matter knowledge (cardiology writer), and I am a clear communicator. Does being Black put me at additional risk?

Think of the book Moneyball by Michael Lewis. The winning algorithm defers from using home runs as the metric of interest. The low-hanging fruit would be to value the player that can hit a home run. If we are only looking at this statistic we might overlook the higher rate of strike-outs among the heavy hitters. It is the same with race. When we focus on a political/social construct and use it as a poor proxy for biology or a wide host of other measures the anticipated outcomes will be misleading and distorted.

Isabel Wilkerson has written a masterpiece, so timely and necessary—Caste: The Origins of Our Discontents. You don’t need to read too far into the text to recognize the caste system in the United States. Perhaps you are a data scientist? The ability to recognize the complexity of slavery and the distortion of humanity existing during the era of “All Men Are Created Equal” where voting rights are still being challenged centuries later—will help you recognize the futility of dividing people by color.

My point is to name the measure you seek to evaluate. When you use a proxy for the measures that might be able to illuminate a discussion the real relationships remain hidden.

Do you know what is hiding in your data model?