DANIEL BUENOResearchers who falsify data in scientific articles usually adopt a specific style of writing in order to try to hide their tracks. This is the main conclusion of a study by Jeff Hancock and David Markowitz, professors in the Communications Department of Stanford University, in the United States. In an article published in the Jornal of Language and Social Psychology in November 2015, they identified distinctions in writing style between fraudulent and non-fraudulent articles. The authors looked at 253 papers published in various periodicals in the field of biomedicine that had been retracted from 1973 to 2013. Using computational linguistics techniques, the authors compared those documents with articles that had not been retracted and were published in the same journals during the same time period, covering similar subjects. The results show that the retracted articles exhibited a high level of what they call “linguistic obfuscation.” “Scientists who falsify data are aware that they are guilty of misconduct and don’t want to be caught. One strategy they can use is to try to obscure the fraud by using certain words or phrases in the text,” explains Markowitz on the Stanford University website. That tactic had already been observed in financial reports. “We wanted to see whether the same thing happens in scientific articles.”
They noticed, for example, that fraudulent papers contain more instances of technical jargon, an average of about 60 more specialized terms than in other articles. One possible explanation is that those words, not common in everyday communication, help simulate the scientific backing for the article. There also are fewer terms that express emotions or value judgments, such as “success” or “improve,” in the retracted papers. According to the study authors, using fewer words that sound positive, like simply stating that the results obtained are “satisfactory,” helps lessen the likelihood that the reader will focus attention on the falsified data in the article. “Our work is a contribution to a research effort that is trying to understand how language can reveal social and psychological dynamics, like fraud,” Markowitz explains. However, he emphasizes the need for more study about the subject so that this approach can be used to detect fraud.
Republish