With big data, comes formulas to parse through the data trying to make sense of it. Those algorithms can help make sense of the data and help filter through the noise to find trends. But those algorithms can also be easily misused and have harmful, if unintended consequences. Cathy O’Neil explores these problems in Weapons of Math Destruction.
Ms. O’Neil is a former academic, hedge fund quant, and data scientist for startups. At the hedge fund she was looking at ways to make money by predicting financial trends. As a data scientist she was looking to predict people’s purchases and browsing habits to monetize those habits. Those are good algorithms where the people using them understand them and update the algorithms as they see problems and can check to see if they are working properly by verifying outcomes.
Many data driven outcomes fail. Those are the focus of Weapons of Math Destruction.
I’m a big fan of the Slate Money podcast with Cathy O’Neil, Felix Salmon, and Jordan Weissmann. When the publisher offered me a review copy of Ms. O’Neil’s book, I jumped at the chance.
It’s a short book and even though the title makes it sound like it’s full of math, it’s not. It’s more about social policy. She points to the danger of decision makers blindingly following algorithmic output that they do not understand.
Take the US News and World Report listing of best colleges. This is preeminent guide for high school students trying to figure out which college to attend. The rankings do not look at actual student outcomes. The publisher does not look at happiness, learning or improvements. It looks at proxies for the outcomes like SAT scores, alumni contributions, and graduation percentages. Colleges started making decisions to improve their rankings in the algorithm by improving those measured proxies. As anyone writing checks for their college bound children, affordability is not one of the measured proxies.
As you might expect there are big chunks of the book dedicated to failings in the financial sector by relying on poorly designed algorithms. The biggest problems often being false assumptions plugged into the data running through the algorithm.
With the election season upon us, the chapter on political algorithms is fascinating. The data scientists of commerce were brought into political campaigns. They are able to micro-target potential voters sending different messages to different groups highlighting the positions that would make the candidate more favorable to that segment of the population. Civic engagement and the political process is increasingly being algorithm driven.
Since our world is increasingly being driven by big data, it’s important to understand what is happening behind the decision-making. Weapons of Math Destruction is an excellent tool to help you understand data driven decision-making.
