Published by Modern Times Review
Unequal distribution of (inherited) wealth has lately taken much of the blame for winners taking it all, and then some. In her disturbing account of the advent of data modelling, data scientist Cathy O’Neil shows how numbers themselves have become commodities, further widening the gap between the haves and the have-nots.
Inequality is on the rise all around the globe. One of the major scholars researching inequality is economist Thomas Piketty, who gained worldwide fame with his book Capital in the 21st Century. Thanks to him, we now know for sure inequality is on the rise. According to Piketty, this is caused by a market economy based on private property benefiting the people who own capital. According to Piketty, ‘[w]ealth accumulated in the past grows more rapidly than output and wages. […] The entrepreneur inevitably tends to become a rentier, more and more dominant over those who own nothing but their labor.” Or, as he states, “[t]he past devours the future”. You’re probably familiar with this view on inequality, thanks to the attention Thomas Piketty has received over the last couple of years. However, you may be less familiar with a very different take on the causes for increasing inequality explained in Weapons of Math Destruction.
Learning by doing, or not
Written by former mathematics professor and data scientist Cathy O’Neil, this book will open your eyes to the flaws of big data analysis. After all, humans designed the algorithms that come up with the forecasting models of big data. Right off the bat, O’Neil mentions the difference between Google – that has millions and millions of users to test and improve its algorithm – and, for example, a data model for a single school in Washington DC with a few thousand students. According to the author, this has severe consequences. “Promising efficiency and fairness, they distort higher education, drive up debt, spur mass incarceration, pummel the poor at nearly every juncture, and undermine democracy.”
“They distort higher education, drive up debt, spur mass incarceration.”
Mathematical models predict the likelihood of somebody underperforming on the job, defaulting on a student loan or committing a crime. Sometimes the model creates a self-fulfilling prophecy when it is faulty to begin with and has difficulties to learn from its mistakes, if running through a feedback loop. In some instances, the faulty models cause real destruction. The “harms are legion. They unfold when a single mother can’t arrange childcare fast enough to adapt to her work schedule, or when a struggling young person is red-lighted for an hourly job by a workplace personality test. We see them when a poor minority teenager gets stopped, roughed up, and put on warning by the local police, or when a gas station attendant who lives in a poor zip code gets hit with a higher insurance bill,” according to O’Neil.
Inequality here seems to cause more inequality in a never-ending downward spiral where the wealthy get wealthier and the poor poorer, just like in Piketty’s book. The consequences of increasing inequality are major, Piketty explains; “The consequences for the long-term dynamics of the wealth distribution are potentially terrifying, especially when one adds that the return on capital varies directly with the size of the initial stake and that the divergence in the wealth distribution is occurring on a global scale.”
Piketty continues, “[g]rowth can of course be encouraged by investing in education, knowledge, and nonpolluting technologies. But none of these will raise the growth rate to 4 or 5 percent a year.” According to Piketty, a growth rate of 4 or 5 percent is needed to increase the economic position across the board, and not just for the wealthy. It is not easy to hit those numbers, though. Piketty explains that only countries catching up with more advanced economies, like China, have seen such numbers. More likely is an annual growth rate of 1 to 1.5 percent in the long run “no matter what economic policies are adopted.” So economic growth is not going to help us in the long run. What will? Piketty proposes a progressive annual tax on capital. This will make it possible to avoid an endless inegalitarian spiral while preserving competition and incentives for new instances of primitive accumulation.
Concerning the data models, there are also measures we can take to help prevent a downward inequality spiral. First of all, not all big data models are evil, and second, big data is here to stay. So with this in mind, what can we do to prevent the models from running wild? Models need to be transparent and illustrate the data used, as well as the results of the modelling. They must also be audited and changed for the better when flaws emerge. Or, as O’Neil states, “These are powerful engines, after all. We must keep our eyes on them.”
O’Neil goes on to point out that some models can have an incredibly positive impact on society. One example is of a child and family services organisation preventing child abuse in Florida, USA; the model was initiated in 2013, in Florida’s Hillsborough County, after nine children over the previous two years had died from abuse in the area. The researchers identified several indicators for predicting abuse, such as a history of drug abuse or domestic violence, a living-in boyfriend, and a parent that had been in foster care. The families this model highlighted received extra resources and, during the two years after introducing it, no child died in the risk area. With these measures we can tend to data models before they spiral out of control and inequality is beyond retreat.