Also in this Issue
Duncan Luce Wins National Medal of Science
By Roberta L. Klatzky
The National Medal of Science has been given to more than 400 distinguished recipients for their contributions to science and engineering. Although initially restricted to the physical, biological, mathematical, or engineering sciences, the domain was expanded by Congress in 1980 to include the social and behavioral sciences. Criteria for the award include an individual's total impact, significant achievements, distinguished service to advancing science and engineering, and peer recognition. This year's recipient, Duncan Luce, exemplifies every one of those criteria. The award joins many other honors awarded to Luce, including election to the American Academy of Arts and Sciences, the National Academy of Sciences and the American Philosophical Society, the 2004 Norman Anderson Award of the Society of Experimental Psychologists, the 2003 Frank P. Ramsey Medal of the Decision Analysis Society and the 2001 Gold Medal for Life Achievement in the Science of Psychology of the American Psychological Foundation.
Luce received a Mathematics PhD in 1950 from the Massachusetts Institute of Technology. His university appointments have included Columbia, Harvard, the University of Pennsylvania, and UC Irvine. In 1988 he founded and assumed the directorship of the UCI Institute for Mathematical Behavioral Sciences, which he headed for 10 years. Currently he is UCI Distinguished Research Professor Emeritus of Cognitive Sciences and Economics.
Luce's work is both theoretical and empirical and above all, mathematically infused. It crosses boundaries among psychology, mathematics, and economics, always with an eye to understanding human behavior from a rigorous theoretical perspective. When asked to cite some of his research contributions that were "favorite children," Luce's list revealed the breadth of his contributions: reanalysis of Fechner's scaling method, the choice axiom, additive conjoint measurement (with J. W. Tukey), contributions to games and decision making, and his book Response Times. Of the choice axiom, which commonly bears his name, he noted that it is important not because it is always right - "it isn't" - but because it both provided a somewhat novel framework for thinking about probabilities and inspired others to study and apply several variants.
Luce also cited as career highlights the Handbook of Mathematical Psychology (with R. R. Bush and E. Galanter) and Foundations of Measurement (with D. H. Krantz, P. Suppes, and A. Tversky), both published as three-volume sets. He has had a number of significant collaborations, including two decades of interaction with Louis Narens on fundamental aspects of measurement and recent association with A. A. J. Marley and several functional equations experts, especially János Aczél and C. T. Ng, on issues of measurement in utility theory and psychophysics.
Although the National Medal of Science is given in the category of "Behavioral or Social Sciences," a perusal of winners from psychology reveals a heavy emphasis on mathematical theorizers, including William Estes, Patrick Suppes, Roger Shepard, and Herbert Simon. Perhaps this is because the representatives of science and engineering who constitute the award committee see common ground with the methods of mathematical psychology, Luce conjectured. The mathematical approach, he suggested, points to one of the challenges facing psychological science today, namely, the lack of unified ideas and themes.
"Physicists agree on several - three or four - theoretical problems to attack; in psychology every theorist has a view." The lack of consensus, said Luce, may reflect the field's tendency to pursue research that rejects null hypotheses. He indicated that Psychology is perhaps one of a few sciences to focus mostly on the goal of rejecting something, rather than finding out what is approximately true. Mathematical psychology, on the other hand, tends to rely on model fitting and evaluation, an approach that intrinsically seeks to accept a null hypothesis that is framed by the model itself, that is, by its axioms and mechanisms. If the model is rejected, it's "back to the drawing board." Of course, issues of sample size and power are important in model testing and must be taken into account, but the goal of the model builder is to construct a "proper" model, not to reject an improper one. The broad impact of Luce's work attests to the value of the modeling approach.