The best books for data scientists trying to be ethical people

This book is now 50 years old, but its message is as relevant and important now as when it was written. In a series of witty essays that border on rants, Andreski attacks much of social science as fluff obscured by technical jargon and methodology. In particular, he laments the growth of quantitative methods as an attempt to add objectivity to social science and make it appear “harder.” True objectivity is about more than mechanical number-crunching, he says; it’s about a commitment to fairness and resisting the temptations of wishful thinking – a challenge anyone who works with data concerning people and their lives should take seriously.

People need less Dawkins in their lives and more Lewontin, whose thought-provoking, accessible writing about evolutionary biology stands in fierce opposition to the trend toward genetic determinism that seems to be the rage nowadays. We are not simply our genes, Lewontin says, because the effects DNA has on our lives are mediated by social and environmental factors, many of which we can influence. While it’s nominally about biology, I also read this as a critique of causal inference, generally. What we consider a “cause” reveals our ideological commitments to certain aspects of the world being maintained, and we should be careful what causal lessons we draw from data.

The thing you should know about science is that it’s a human enterprise. As a result, it’s dependent on human factors like social consensus and prejudice. In this series of case studies of famously expensive and difficult-to-replicate experiments probing the limits of scientific understanding from biology to theoretical physics, Collins and Pinch show how scientific knowledge gathering is rarely straightforward because there are always alternative explanations available for the data. Was the phenomenon real or was the experiment set up badly? We can never know for sure, but we decide collectively what we believe. Scientists are experts participating in human culture, they argue, not mysterious clergy issuing declarations of absolute truth.

The fact that race is a social construct and not a biological reality seems to be a lesson that we are destined to learn and re-learn many times. Saini uses a personal, journalistic style to tell the story of the pernicious myth of biological race in the sciences, drawing a continuous line from scientific racists like Francis Galton in the 1800s to present-day medicine and right-wing politics. The story is alternately funny and horrifying, with incredibly timely significance. It should be read by all data-adjacent individuals as a cautionary tale about avoiding the mistakes of the past and present. 

If you’ve never thought of “intersectional feminism” or “the gender binary” as essentially data-scientific terms, please allow this book to correct that. Data science is a locus of power, and that power can be wielded in the service of oppression or liberation. This book raises essential questions about the predominantly white, male, technocratic interests served by the traditional narratives of data analysis and what feminism and data science have to offer each other. Bottom line: the data doesn’t speak for itself, never has, and never will.