Stephen E. Fienberg (1942–2016) Statistician Who Campaigned for Better Science in Court

OBITUARY COMMENT Stephen E. Fienberg (1942–2016) Statistician who campaigned for better science in court.

tephen Fienberg was the ultimate public on 14 December at home with his family in 90% of the first 500 cases examined had con- statistician. One of his most signifi- Pittsburgh, after a long battle with cancer. tained erroneous statements in laboratory cant social contributions was making Between 1982 and 1987, Fienberg chaired reports and analysis. Sa convincing case that bad science sends an NAS panel which resulted in his 1989 His work is also helping to fill the scientific innocent people to prison. He used statis- book The Evolving Role of Statistical Assess- holes he identified. In 2015, he drew together tics to assess the foundations of legal a four-university consortium to evidence — fingerprint analysis, hair develop a statistical foundation for matching, polygraphs — and found how examiners should use common them wanting. His work helped to sorts of courtroom evidence, such limit the use of shoddy techniques as fingerprints or ballistics data, to and forced reviews of past cases. match a pattern from a crime scene More than most statisticians, to a suspect or their belongings. Fienberg focused on policy. He Fienberg was an indefatigable improved how the census counted mentor, replying to students’ e-mails homeless people, promoted data at all hours of the night. He often sharing among scientists and devoted himself to work that was nec- UNIV. MELLON OF CARNEGIE COURTESY assessed online privacy measures. essary rather than prestigious. Survey He testified before Congress 15 research, for example, is no longer a times, and consulted for numerous hot topic for theoretical statisticians federal agencies, working on prob- because the theoretical underpin- lems related to health, employment, nings of choosing a survey sample justice and agriculture. He took part are thoroughly worked out. However, in dozens of US National Research Council ments as Evidence in the Courts (Springer). doing a good survey is difficult, and (NRC) committees and panels, including At the time, the book received little notice, Fienberg continued to work on problems chairing the Report Review Committee, but he continued to develop this line of work. such as questionnaire design until his death, and contributed to hundreds of others. In In 2003, he worked on the NAS report ‘The helping to make surveys more reliable. His 2015, he was one of 13 scientists convened Polygraph and Lie Detection’, which helped modelling techniques have been cited by the US National Academy of Sciences to limit the use of the instrument. across a range of academic disciplines, and (NAS) for a summit on the need to improve In 2015, he was a senior adviser for a also in reports on human-rights violations research integrity in the wake of misconduct report by the President’s Council of Advisors presented in war-crimes trials and historical scandals and concerns about reproducibility on Science and Technology that highlighted clarification commissions. of results. problems across a range of forensic-science In October, Carnegie Mellon hosted a He also made important theoretical techniques. He also served as the only symposium to celebrate Fienberg’s career. advances in statistics. His 1975 book, statistician on the Department of Justice’s One attendee recalled that, when she began Discrete Multivariate Analysis: Theory National Commission on Forensic Science work on an NAS panel, she was warned to and Practice (MIT Press), is a canoni- until his death. He was the driving force be prepared to deal with this impossible guy cal approach to the analysis of categorical behind draft recommendations on the on the review committee who always had data — information such as race or size that use of probabilistic statements in court; difficult questions — Steve Fienberg. can be lumped into categories. His work on the commission is expected to vote on the At that same gathering, Fienberg election polling, replicability of research recommendations this April. explained how his illness had shaped a findings, privacy in digital networks and In many forensic disciplines, examiners new interest. He supervised his own care, the application of Bayesian methods were link evidence from a crime scene to evidence combing the literature to find medicines all ahead of their time. from a suspect by stating, for instance, that approved for other uses that had shown off- Fienberg was born in 1942. At Canada’s there’s only a one-in-a-million chance that label success in treating cancer. He urged his University of Toronto, in his hometown, he hair from a crime scene and hair from a sus- audience to take up a question that he’d been studied mathematics and science. For his pect are not from the same person. In reality, thinking about: developing rigorous analyses PhD, at Harvard University in Cambridge, with the exception of DNA testing, there is to evaluate the effectiveness of treatments in Massachusetts, he worked on an NRC study no research to back such claims: the likeli- an age of personalized medicine — one in on the possible association of anaesthesia hood of coincidental matches of forensic evi- which more people are doing as Fienberg did and postoperative liver failure. Next came dence is unknown. Fienberg’s work helped and becoming a ‘sample of one’. ■ a joint appointment in statistics and theo- to bring to light these problems — and the retical biology at the University of Chicago, resulting wrongful convictions. Robin Mejia did a PhD in biostatistics Illinois. From there, he went to the Uni- For example, his assessment of what in May 2016, inspired in part by Steve versity of Minnesota in Minneapolis and, constituted a statistically defensible analysis Fienberg. She now manages the Statistics in 1980, to Carnegie Mellon University in of microscopic hair comparisons was used and Human Rights Program at Carnegie Pittsburgh, Pennsylvania, where he spent the by the FBI to review past investigations. In Mellon’s Center for Human Rights Science. bulk of his academic career. Fienberg died 2015, the bureau announced that more than e-mail: [email protected]