INSIGHTS | BOOKS

D A T A S C I E N C E Truth or Truthiness Distinguishing Fact from Facts versus fallacy Fiction by Learning to Think Like a Data Scientist Howard Wainer A statistics-driven treatise teaches readers how to spot lies, Cambridge University Press, half-truths, and outright deception 2016. 228 pp.

By Christopher J. Phillips ters under the headings of “thinking like a complex questions. Even when the data data scientist,” “communicating like a data are incomplete, we don’t have to settle for harles Dickens opened his 1854 novel scientist,” and “applying the tools of data sci- “truthiness,” he argues. Hard Times with the memorable fig- ence to education.” His chosen examples are Wainer spent 5 years on a school board, ure of Thomas Gradgrind admonish- thoughtful and touch on relevant and engag- and often the book reads as if he’s writing to ing a schoolmaster to “teach these ing problems, especially in medicine and edu- fellow board members: people who have ba- C boys and girls nothing but Facts. Facts cation. (Wainer is currently with the National sic statistical knowledge, who are often faced alone are wanted in life” (1). Prosely- Board of Medical Examiners and spent more with incomplete data, and who are ultimately tizing “men of facts and calculations” were than two decades with the Educational Test- responsible for making and justifying di cult ripe for Dickensian satire in the mid-19th ing Service.) decisions. His examples form a convincing century, thanks, in part, to the mathemati- Aside from brief case studies, Wainer’s argument that no matter how intractable a cal sciences, which had been established as book also functions as a popularization of the problem seems, careful use of data can help powerful, even definitive, ways of knowing work of his colleagues (primarily Paul Holland sort things out. about the world. and Don Rubin), an homage to well-designed Wainer is also intent on debunking the be- Since antiquity, mathematics had been an data displays (from college acceptance letters lief that more data is always better: “the mind- exemplar of certain knowledge. less gathering of truckloads of Reasoning like a mathematician data,” he writes, “is mooted by the on August 4, 2016 meant reasoning reliably from as- gathering of even a small amount sumptions to conclusions. Geom- of thoughtfully collected evidence.” etry, for example, was taught for The book is a fun read unless centuries as a model for reason- you happen to be a policy-maker ing in general. who leans to the right politi- Howard Wainer’s new book, cally. The only examples Wainer Truth or Truthiness: Distinguish- provides of people with such “an ing Fact from Fiction by Learning excessive dimness of mind” that to Think Like a Data Scientist, they won’t ever be capable of suggests that geometry’s role in “connecting the dots of evidence”

the search for truth has been re- are congressional Republicans. http://science.sciencemag.org/ placed by data science. That is, This is a shame, because his ul- replaced by a “complex mixture timate message is that whatever of ideas and methods drawn one’s political background might from many related fi elds,” espe- be, policy decisions require care- cially statistics, that can be used ful analysis of available evidence. to evaluate and make inferences Of course, sometimes issues

from numerical data. There’s a are inherently political. Good, re- Downloaded from bit of an irony here, given that liable evidence is important to statistics and probability were “Truthiness,” a term coined by Stephen Colbert in 2005, is defined as many of the debates that Wainer long considered sciences of un- “truth that comes from the gut, not books.” wades into—the value of teacher certainty, as distinguished from tenure, the measurement of surgi- geometry’s eternal truths. Nevertheless, at a to inside-out plots of baseball statistics), and cal outcomes, the achievement gap in public time when PolitiFact’s “Truth-O-Meter” pro- a repository for his favorite quotations (with schools, the possible seismic e ects of frack- vides an essential guide to politicians’ stump sources ranging from Blanche DuBois to the ing—but these are also issues that unavoid- speeches, Wainer’s book is welcome indeed. Roman philosopher Seneca). ably engage humanistic debates about rights, It is perhaps especially relevant in a divisive Books like this face the inevitable prob- priorities, and values. election year, when evidence, truth, and ex- lem of how to spread the gospel of data Wainer’s point, though, is that for the pertise seem under assault and yet desper- analysis to neophytes without reading like questions where evidence does exist, we ately needed. an introductory statistics textbook. There is should know what to do with it. He plausibly Truth or Truthiness consists of a series of certainly little here that will be new to pro- suggests that data science provides a good short, lively chapters that discuss how data fessional statisticians; at the other end of the toolbox for that task. Policy-makers and the scientists would approach various contem- spectrum, readers who have never taken a electorate would do well to follow his advice. porary quandaries. Wainer groups the chap- statistics course will fi nd few details about REFERENCES how to actually use most of the tools Wainer 1. Charles Dickens, Hard Times: For These Times (London, mentions. Rather, it seems that he’s writing The reviewer is at the Department of History, Carnegie Mellon Bradbury and Evans, 1854). University, 240 Baker Hall, 5000 Forbes Avenue, Pittsburgh, PA more to convince policy-makers that data

15213, USA. E-mail: [email protected] science provides reliable ways to approach 10.1126/science.aag2776 PHOTO STOCK ARCHIVE/ALAMY AF PHOTO:

550 5 AUGUST 2016 • VOL 353 ISSUE 6299 sciencemag.org SCIENCE

Published by AAAS Facts versus fallacy Christopher J. Phillips (August 4, 2016) Science 353 (6299), 550. [doi: 10.1126/science.aag2776]

Editor's Summary

This copy is for your personal, non-commercial use only.

Article Tools Visit the online version of this article to access the personalization and article tools: on August 4, 2016 http://science.sciencemag.org/content/353/6299/550 Permissions Obtain information about reproducing this article: http://www.sciencemag.org/about/permissions.dtl http://science.sciencemag.org/ Downloaded from

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2016 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS.