The Ombudsman: Cheating in Management Science

Honesty is vital to scientific work and, clearly, most scientists are honest. However, recent publicity about cases involving cheating, including cases of falsification of data and plagiarism, raises some questions: Is cheating a problem? Does it affect management science? Should anything be done? Comments Postprint version. Published in Interfaces, Volume 13, Issue 4, August 1983, pages 20-29. Publisher URL: http://www.aaai.org/AITopics/html/interfaces.html This journal article is available at ScholarlyCommons: http://repository.upenn.edu/marketing_papers/112 Published in Interfaces, 13 (4), 1983, 20-29. Cheating in Management Science (with Comments by M. K. Starr and M. J. Mahoney) J. Scott Armstrong The Wharton School, University of Pennsylvania Honesty is vital to scientific work and, clearly, most scientists are honest. However, recent publicity about cases involving cheating, including cases of falsification of data and plagiarism, raises some questions: Is cheating a problem? Does it affect management science? Should anything be done? My interest in scientific cheating began years ago, when I was responsible for the marketing research on a government-sponsored research contract. The purpose of the project was to examine an approach to urban transportation, which I will call the Urban Transit Plan (UTP). The government supported this research because of the potential social benefits if commuters were to switch from their automobiles to the UTP. My initial studies showed that more commuters would switch to the UTP from the existing mass transit than would switch from automobiles. Thus, instead of producing social benefits, the UTP would increase social costs. So what happened? First, I was asked by the project leader to “check” my results. What 1 inferred from this request was that 1 should not report these results, but should do studies until I obtained the desired results. I commissioned additional independent studies, but these supported the initial findings. Subsequently, I learned from informal sources that my reports were “lost” when the project leader reported to the government agency. Despite my objections to the project leader, additional reports were also lost. Therefore, I took action to ensure that the government received the reports. My action led to “peer review” whereby the 15 members of the project team, mostly university faculty, were asked by the project leader whether I should be fired. Only two faculty members voted that I should not be fired; however, they look such a strong stand in my favor that 1 was allowed to remain. 1 have felt pressure to cheat on other projects in management science. Furthermore, 1 have heard stories of colleagues who experienced similar pressures. I have observed what seemed to be cases of cheating, and others have told me of cases that appeared to involve dishonesty of one sort or another. None of these cases were publicly revealed. My own efforts to report cheating on the UTP project apparently were of little interest to the government sponsor. Recently, the number of publications on cheating in science has grown at a dramatic rate. A summary of some cases of scientific cheating is provided in Broad and Wade’s Betrayers of the Truth [1982]. This provides vivid descriptions of 34 cases of alleged fraud in science. Almost all of these cases were identified since 1950, and over half have arisen in the last five years. Some of these cases refer to earlier work, dating as far back as Hipparchus in the second century BC. They include such famous scientists as Ptolemy, Galileo, Newton, Bernoulli, Dalton, Mendel, and Burt. An analysis of the Reader’s Guide from 1900 to the present supports the picture in Broad and Wade. We found little discussion of cheating prior to 1950. Some articles appeared between 1950 and 1974, and the number of papers has increased dramatically since 1975. Does this mean that cheating is a serious problem? Initially 1 thought it did. After reviewing the literature and discussing the problem with others, I have concluded that 1 was partially wrong. Science is not threatened by cheating; but scientists are. My initial conclusion was wrong because 1 fell prey to two common errors, even though I was previously aware of these errors. These errors are known in the research literature as “availability” and “vivid examples.” “Availability” states that events that are easy to recall are perceived as more likely. (For recent studies on availability, see Kahneman, Slovic, and Tversky [1982], part IV.) Projects that involved cheating are easier to