In the last few years many cases of scientific fraud have been uncovered. Fraud is defined as: “Wrongful or criminal deception intended to result in financial or personal gain” by the Oxford Dictionary. In the case of scientific papers several forms of fraud are imaginable, such as deliberately citing other sources wrongly in the theory section, plagiarism, obfuscating problems with the methodology of the research, creating data aimed to fit a certain outcome, or deliberately misinterpreting the research results. So, what causes scientific fraud? I think that by putting quantity over quality and ‘publishable’ results, i.e. confirmation, over ‘negative’ results, i.e. falsification, modern science has created a structure in which fraud can have high pay-offs while having low costs. Meanwhile, the costs of uncovering fraud can be high, while the benefits of uncovering fraud are very low. Not all forms of scientific fraud are equally likely to occur, because peer reviewers are experts on the topic, which means that many forms of scientific misconduct will be recognized by them at low cost by simply reading the paper, and seeing the misappropriation of ideas, lacking citations or misinterpretation of results. However, data fraud and opaquely constructing new models until a publishable result is found are more likely to occur, because uncovering such fraud is time consuming.
What can we do to prevent scientific fraud? Key to understand fraud are the expected benefit and the expected cost of fraud for the fraudster, and the expected cost and expected benefit of searching fraud for other scientists. There are several factors which influence the marginal cost/marginal benefit analysis of falsifying data or searching for positive results, and of uncovering fraud.
Firstly, the ‘publish or perish’ attitude has increased the costs of finding negative results. The current rules of science make the cost of being wrong too high in two ways. Being wrong means your research is not published and cited, which means the research has no value for anyone, even though negative results based on sound methodology are actually valuable. Moreover, the pressure to publish has increased the cost of being wrong. In a system in which fewer publications are demanded, finding a negative result is just a clue to where to search next, or to change the methodology. Meanwhile, if publishing high quantities is key for academic careers, research spent on what turned out to be negative results has immediate negative consequences. Fraud is a way to prevent incurring these costs, so high costs of falsifying results increase the benefits of fraud. We know that individuals facing expected costs are more likely to undertake risky behaviour to avoid those costs (see Thinking Fast and Slow by Kahneman). Fraud is exactly such risky behaviour. Of course, giving scientists too much credit for negative results creates less incentive to find positive results, which will lead scientists to be less careful in selecting hypotheses to test. So the benefits of being right should always be higher than the benefits of being wrong, but the cost of being wrong is exceedingly high at the moment.
Secondly, the expected marginal costs of fraud are low. Although the costs of being uncovered are high – it is a huge humiliation and it spells the end of the fraudster’s career – the expected cost of fraud depends on the chance of being uncovered, which is relatively low. In turn, the expected cost of fraud depends on the benefits which others receive for uncovering fraud. A key mechanism in science to prevent fraud is anonymous peer review, which lets anonymous reviewers read papers before they get published. Unfortunately the cost of uncovering fraud for peer reviewers is high: it takes much time to study data collection, the data itself, and the possible models that researchers have tested. Moreover, the publish or perish mentality has led to an increase in the production of scientific papers. Yet, the very scientists who are supposed to ensure that other researchers are not fraudulent, are themselves under the exact same pressure to produce large quantities of research, so they have less time to spend on peer review. Meanwhile the rewards for uncovering fraud are low or absent: there is no reward for peer reviewing or uncovering fraud. Economists would say that in the current situation finding fraud has positive externalities, that is, others benefit from one person uncovering fraud, while the persons who could uncover fraud are insufficiently rewarded for doing so. The market for uncovering fraud fails.
So what can be changed to prevent fraud? Firstly, we need to internalize this positive externality. We need to make sure that the person who uncovers fraud is rewarded by those who benefit from it. It is in the very nature of science to be both competitive, researchers want to be the first to find a certain result or to show that someone else was wrong, and cooperative, research is almost always based on earlier research. Like the invisible hand of Adam Smith in economics, the benefits of cooperation ironically depend on competition between scientists which keeps scientists sharp. We need to make science more competitive, by rewarding researchers more for finding others’ mistakes, so science can be more cooperative, because with less fraud scientists can put greater trust in others. Secondly, we need to stop disproportionately valuing quantity over quality and confirmation over falsification, because this increase the costs of finding (actually valuable) negative results, which makes scientists more likely to be fraudulent to get positive results.
Bottom Line: By putting quantity over quality, and confirmation over falsification, scientists feel a higher pressure to publish. By changing this, scientists will feel less pressure to be fraudulent. Moreover, by rewarding scientists for uncovering fraud, we can incentivize them to uncover fraud.
Acknowledgements: I would like to thank Jora Broerse and Dr Brandon Zicha of LUC The Hague for their comments.
Image: Former German defense minister Freiherr zu Guttenberg, who resigned over plagiarism. Copyright by World Economic Forum, swiss-image.ch/Photo by Sebastian Derungs