Event

Social experts within and without: social epistemologies, the Netflix competition, and the making of machine learning:
From 2006 to 2009, the Netflix video service sponsored a competition to improve predictions about which films their customers would rank highly and lowly. Following strands in computational statistics and machine learning from the 1990s onward, the Netflix competition reduced the diverse values of knowledge—simplicity, predictive ability, mechanical insight, to one: a single metric to be maximized. The competition drew upon and, indeed, consecrates, a renewed form of instrumentalism that builds upon three related presuppositions, an agreed-upon single metric, a perspectivism, and a sense of collective epistemology. To the surprise of many statisticians and others in the 1990s, and to many of the contributors in the Netflix competition, the best ways to maximize this metric turned out to require bringing together many models into ensemble, indeed, to bring together the product of diverse human labors. Values of collective epistemology were built into the trained algorithmic systems themselves. The Netflix competition allows us to see the social worlds that create algorithmic models that are themselves social. Through it, we can see the coming together of social organizations and the coming together of different algorithmic predictors, as well as the stories about such organizations: the vernacular social theorizing within algorithms, the notions of collective expertise that yield ensembles of reified expertise. The well-known problems with machine learning systems end up in many ways stemming from the inadequate sociality of machine learning and its makers.

Matthew L. Jones focuses on the history of recent information technologies and intelligence as well as the history of science and technology in early modern Europe. He received his A.B. and Ph.D. from Harvard (1994, 2000) and an M.Phil. from Cambridge, after which he taught at Columbia for twenty-three years. Along with Chris Wiggins, he is the author of How Data Happened, a history of the science, politics, and power of data, statistics, and machine learning from the 1800s to the present (W. W. Norton, 2023). He has published two books previously, The Good Life in the Scientific Revolution: Descartes, Pascal, Leibniz and the Cultivation of Virtue and Reckoning with Matter: Calculating, Innovation, and Thinking about Thinking from Pascal to Babbage (both with Chicago). He has received fellowships from the Mellon Foundation, the Guggenheim Foundation, the Sloan Foundation, and the National Science Foundation, and is currently a CIFAR fellow in the Future Flourishing(Link is external) project.

Professor Jones is completing a book on state surveillance of communications from the 1970s until the present, including the explosion of state-sanctioned hacking, focused on signals intelligence agencies of the US and its allies.