Robert Schlaifer, who was probably the most influential person in my career, and I’ve met a lot of people. He was a—got his PhD in Ancient Greece. If you look in the Encyclopedia Britannica, there are a lot of articles there written about the ancient times signed R.O.S.—“Robert O. Schlaifer.” During the Second World War he got a job working with the underground laboratory here at Harvard doing editorial work, or writing up reports for the research. He was a handyman. He ended up writing a tome, big, fat tome about aviation, the economics of aviation engines, the procurement and the economics of that. So he had that type of mind where he could learn things quickly.

For some reason, they hired him at the Business School to be a utility infielder. That he would teach what they needed to be taught. He did a course in accounting, in marketing, and then they assigned him to teach a course in statistics. He had no statistics previously. He had one course in calculus in mathematics, but that’s about it. I know the instructor who taught him that.

So like a classical Greek scholar, he looked at the history of statistics, what the great minds said about statistics, and he said, “That’s all well and good. That’s if you really are interested in inference as a social scientist. But if you’re in a business school, you like what you’re doing to tie to business decisions. There shouldn’t be a divide.”

He went down the same path I went down. Only he sort of invented what is now called Bayesian—it’s really doing analysis with allowing judgmental probabilistic inputs by experts. And he became a specialist at that. Well, when he found out that that was my closet dream, we embraced each other, and we worked like mad together.

We then went ahead and said, “Well, we’re going to revolutionize the field of how statistics is taught, especially at the Business School,” but there was a lot to be done. There were techniques, standard problems that were solved in the classical way, but not in the newfound Bayesian way, which allowed judgmental probabilities. And people criticized us, though, and said, “You know, you two will probably have more converts on the philosophical side, but it’s not going to amount to very much, because the problems are too hard. And you can’t really get important people, experts, to give you responsible judgments.”

So we started a program in 1958 of showing what the classical statisticians did we can do also, only better. And we had a laboratory. I did some consulting for DuPont, and I tried to get the people in DuPont to work into this field. And yes, we were able to make the mathematical analysis tractable. And yes, we could get engineers to give judgmental inputs. They were willing to do that, excited about doing it, except they were no damn good. They were inconsistent. And we got into the whole question about judgmental inputs, and how to make respondents consistent that later took off. It’s now a hot topic for people working in judgmental decision making.