Gowri Gopalakrishna, an epidemiologist at the Free University of Amsterdam and a co-author of the Dutch study, thinks that the percentage of researchers who confessed to falsifying or fabricating data could be an underestimate. Gopalakrishna says that the survey’s questions were more direct than those in earlier surveys on the issue. For this reason, she is hesitant to compare the results to other studies, except for one from 2001 (ref. 3). In that study, which used methods similar to those in this survey, found that about 4.5% of respondents admitted to falsifying data at least once.
This is an unusual survey to report. It will still widely underestimate how much falsification, and the context varies for each “study”. Sometimes, you just need to fit the data and go with the disclaimer that “no animals were harmed during making of this film”. Then you wonder- why were they used in the first place or were they as critical to the plotline? Likewise, for the studies too. Surveys, inherently, are dependent on the respondents, and I have a strong feeling it was published for the sake of publication. No one will own academic fraud in public.
Here’s something interesting:
More than half (51%) of respondents to the Dutch survey also reported frequently engaging in at least one of 11 ‘questionable research practices’ (QRPs), which include using inadequate research designs or deliberately judging manuscripts or grant proposals unfairly. QRPs are considered lesser evils than outright research misconduct, which includes plagiarism and data falsification or fabrication.
(emphasis mine)
Why is it considered “lesser evil”? Scientific research and progress depends on trust and a “fair assessment” and therefore decisions seems like a collective responsibility under the ambit of a “committee”. It is either “swim or sink” together and can hamstring a genuine question or something innovative. It is too generic to label them as academic fraud but a transparent criterion will alleviate many problems which plague the academia.
RadOncNotes 