What is going on with the increase in retractions?

It is no secret that the number of scientific publication retractions is on the rise and that there is a disproportionate number of retractions associated with higher impact journals.  Retractions, though, may result from any number of factors ranging from downright scientific malfeasance that has horrible repercussions as in the case of the STAP stem cell controversy to honest error and misinterpretation of the data as with the arsenic DNA example (albeit the article has not been retracted).   There is even some dispute as to the factors that weigh most heavily in an article being retracted.  For example, earlier work done by Fang et al. reported that up to 67% of retracted articles are due to scientific misconduct.  Another study found that ‘error’ is the leading cause and that misconduct only accounted for 20% of all retractions. 

Yes, overall misconduct numbers are low and as with the above examples, science does, for the most part, correct itself.  I.e.. The system is working.  With 2% of researchers saying they’ve produced falsified data and another 14% saying they’ve witnessed misconduct by a colleague, it indicates there is a lot of fraud that is still not detected.   Scientific misconduct has downstream effects that the scientific community needs to limit, outside of financial costs, which, in and of themselves, are large.  One example is the propagation and persistence of bad data.  If you cite a retracted paper in your own studies, what does that say about your own work?  For one case study, over half a researchers retracted articles were being cited 5 years after the retractions.   Connected to the propagation of bad data and of more concern is the erosion of public confidence. The Autism and the MMR vaccine link is still being fought and having major public health implications. With the advent of the internet and such groups as Retraction Watch, the dirty laundry of the not so neat process of science, so to speak, increasingly will be aired for the public to see, comment, and vote upon. 

Does misconduct arise due to bad people making bad choices or situations (personally or institutionally driven) that cause good people to make bad choices?  For those that do commit misconduct (which can be a broad swath of misdeeds and open to interpretation), is it a conscious decision -- “Today, I will fabricate data.” -- ?  Or, is it more of a gradual slide?   How much of it is self-deception versus deliberate deception?  To what links will they go to perpetuate the fraud or not be detected?

Janet Stemwedel summarized a limited empirical study published in Science and Engineering Ethics that analyzed 92 case files from the Office of Research Integrity for misconduct causality.  They found a total of 44 self-perceived justifications that they grouped into 7 explanatory clusters:   (1) personal and professional stressors, (2) organizational climate, (3) job insecurities, (4) rationalizations A, (5) personal inhibitions, (6) rationalizations B and, (7) personality factors. 

In On Fact and Fraud: Cautionary Tales from the Front Lines of Science, the author found 3 commonalities to the studied cases:   1) The feeling of career pressure (e.g. tenure, promotion, job-security, funding, etc); 2)  Self-deception that the answer the researcher knew to be correct would be borne out once all the ‘work’ was done; and 3) the feeling that the experiments they were working on did not have to be reproducible.  

The conduct of scientific research is a human endeavor and, therefore, will always be flawed and subject to abuse. Yet, how can scientific misconduct be drastically reduced?  Because there is no single justification or reason for misconduct, there is no panacea.  No system that is devised will eliminate fraud.  A common refrain trumpets the need for more reproducibility through transparency and more robust Post Publication Peer Review process.  Simply encouraging reproducibility does not guarantee an arrival at the ‘truth’, as there can be methodological biases that will ensure reproducibility of an answer, but not necessarily the correct answer.  Although there are creditable PPPR sites such as PubMed Commons, the unfiltered, to be understated, PPPR sites can induce any scientist to shudder.  Yet, a more ‘robust’ PPPR environment will likely persist and grow.

Ethical training for students and post-docs is another suggestion.  If someone is a cheat by the time they get to grad school, though, ethical training is not going to do much more than provide insight in how not to get caught.  One area that can and likely should be improved is the statistical and experimental design training for grad students.  If the experiment is done correctly, the likelihood of mistaken interpretation and self-deception will be reduced greatly; and, as Richard Feynman said: 

The first principle is that you must not fool yourself--and you are the easiest person to fool. So you have to be very careful about that. After you've not fooled yourself, it's easy not to fool other scientists. You just have to be honest in a conventional way after that.