David H. Blackwell (1919–2010)
David Blackwell came to statistics late, but his career in mathematics got an early start. The oldest of four children, Blackwell grew up in Centralia, Illinois, and considered himself fortunate to attend an integrated school at a time when racial segregation was the norm.
Blackwell was always adept at mathematics, but he didn’t much care for algebra or trigonometry. “I could do it,” he said; while he found the math useful and interesting, geometry was what really caught his attention. His high school teacher made the subject beautiful for him. “Geometry connected with activity and motion,” Blackwell said, and he loved the idea of the “helping line,” a line that can be drawn to clarify a problem that had previously seemed unsolvable.
Another teacher encouraged Blackwell’s interest in math by advising the high school’s Mathematics Club. Mr. Huk would look for problems in School Science and Mathematics, the official journal of the School Science and Mathematics Association, and bring them in for the club to solve. Blackwell’s name appeared in the journal three times with others, and once he solved a problem on his own and published the solution in SSM with Huk’s help.
At age 16, Blackwell began his mathematics degree at the University of Illinois. His parents were very certain that their children should attend college, but Blackwell said they didn’t have much to do with what he studied. Blackwell took jobs to help pay his way through school, and earned his bachelor’s degree in three years by taking classes during the summer months.
Blackwell continued to study mathematics at the University of Illinois, earning a master’s degree in less time than usually needed. By age 22, he had a PhD in mathematics, for which he wrote a thesis on Markov chains, under the supervision of Joseph Doob.
After completing his studies, Blackwell served as Rosenwald Postdoctoral Fellow at the Institute for Advanced Study in Princeton for one year. He always knew he would teach at some level; he originally planned on elementary or high school math, but from Princeton, Blackwell sent a letter to each of the 105 black colleges in the country, looking for a faculty position.
At first glance, this is a striking example of the employment discrimination of that time, but Blackwell saw it more simply at the time; he has said that he just accepted that a black teacher would only be able to teach at a black college. Blackwell did apply to the University of California at Berkeley and Jerzy Neyman interviewed him, but prejudices at that time were still too strong to allow an African American on the faculty.
Out of the 105 letters, Blackwell received three job offers. He chose Southern University in Baton Rouge, Louisiana, over West Virginia State College and Clark College in Atlanta, Georgia. “It was the first offer I got,” Blackwell said. After Southern University, Blackwell did teach at Clark College for a year.
In 1944, Blackwell went to teach at Howard University in Washington, D.C., which he called “the ambition of every black scholar.” It took only three years for Blackwell to become a full professor and head of Department of Mathematics.
Blackwell credits a 1945 ASA-sponsored lecture by Abe Girshick with turning his interests toward statistics. Girshick gave a talk on sequential analysis that revolutionized the way Blackwell thought about sampling in two ways. First, where some scientists would limit their sampling before they even began, Blackwell remembers Girshick recommending sampling “until you’ve seen enough.” Blackwell saw statistics applied to a real problem, and it intrigued him.
Second, Girshick announced a theorem that Blackwell thought was false. He worked up a counter example and sent it to Girshick. It was wrong, but sending it was right. “Instead of dismissing it as the work of a crank,” Blackwell said, “he called and invited me to lunch.” Thus began a cooperation that lasted about a dozen years. In 1954, the two published Theory of Games and Statistical Decisions. By this time, after several summers of working on “games” for the RAND Corporation, Blackwell had become a leading expert in the area.
At the hands of Jimmy Savage, Blackwell had also become a diehard Bayesian. “Jimmy convinced me that the Bayes approach is absolutely the right way to do statistical inference,” he said. Savage had been at the Institute for Advanced Study with him, and they worked together again at RAND. Blackwell said he thinks he was looking for Bayes all along, and Savage was the one who brought him to it.
He remained at Howard until 1954, when he took a professorship at U.C. Berkeley. Over the years, Blackwell had remained friends with Neyman and his colleagues. In 1956, Blackwell became the chair of the Berkeley’s Department of Statistics, but the administrative role was less fulfilling. For about a year after stepping down, he said, “my first thought was, ‘I’m no longer chairman,’ and it made my day.”
Blackwell’s extensive teaching career has yielded many exciting moments, but one of the ones that stands out is a time when he was working on a problem involving choosing experiments, and he and several of his peers couldn’t seem to find the solution. He mentioned it to a graduate student, and the student was tempted to try the problem. Blackwell said he was initially inclined to discourage his pupil, who wasn’t among the top students in the program, but he was insistent. The student found the solution, and Blackwell chuckles about it to this day, so pleased that one of his students solved the problem that it outweighs not solving it himself.
This student’s abilities were a kind Blackwell could appreciate. Every year, the department had a meeting to decide which graduate students to support. “We had more applications than money,” Blackwell said. He supported this insistent student based on one thing: “he was the best “kriegspiel” player in the department.” Kriegspiel is a war game; historically, the term refers to war games conducted by the Prussian and German militaries.
Over the course of Blackwell’s career, computers have prompted much of the changes to the professions. The focus has shifted, he said. “The field of theoretical statistics that I work in has all but disappeared.” Now, there is more emphasis on problem-solving, which is data-heavy.
Blackwell suspects this change is a good thing, though he claims never to have published a paper with data in it. Statistical decision theory, the field in which he has done so much work, may now be pretty well understood, he said, and maybe it’s less necessary now that statisticians can do so much with computers.
Blackwell is a longtime member and former vice president of the ASA and an elected member of the National Academy of Science (the first and only African American mathematician) and the American Academy of Arts and Sciences. He holds 12 honorary doctorate of science degrees and has served the profession as president of the Institute of Mathematical Statistics, the International Association for Statistics in Physical Sciences, and the Bernoulli Society. Prior to retiring, Blackwell published 80 papers.
When students ask his advice about the profession, he tells them what he would tell them about any other profession: keep trying different things and try to find something you like. “And go, Bayes!” he added.
Learn more about David Blackwell from the University of St. Andrews, Scotland.