I found Dan Ariely’s TED Talk on dishonesty (https://www.youtube.com/watch?v=onLPDegxXx8&feature=youtu.be) very compelling, particularly the part where group mentality can have a major effect on the prevalence of cheating.
He begins by asserting that most people are likely to cheat, though not in proportion to a cost-risk-benefit analysis. Instead, people tend to cheat a marginal amount in most circumstances in order to be able to both reap the benefits offered by cheating, as well as maintain the image that they are ‘a good person’.
What does affect this marginal amount, it seems, is the perceived level by which other ‘good people’ cheat. He gave one of his studies as an example, where students taking an exam were likely to cheat more extensively if an obvious cheater was perceived to be part of the in-group (a fellow university student), as opposed to when that person was part of an out-group (a student from a rival university).
I think this phenomenon has great relevance to the scientific community, particularly in regards to lab environments. This phenomenon speaks to a relativistic kind of morality, as well as an unwillingness to go against the grain. If a particular lab has a specific method that they use for analyzing data or preparing figures, it may be difficult for a newcomer to challenge that method, even if the newcomer suspects that it may not be completely accurate. It is all too easy to accept a method as just the way the lab does things; stirring up trouble about this point could lead to a variety of problems: ridicule at the newcomer’s lack of understanding of the technique, backlash from other lab members, even resentment at the implication that the data the lab generates may be flawed in some way. As such, it may be difficult for labs to perform quality checks on themselves. Even if all the members of a lab truly believe that their methods are utterly sound, they may be unwilling to go to great lengths – using time and resources – to try alternate methods to test their belief.
Any uncertainty at the data-producing level is only compounded as any papers move on to higher levels of review, as the process of fact-checking becomes more arduous, as more reverse-engineering of protocols and raw data is required. Therefore, I think the lack of replicability in scientific data (as discussed at length in the Economist’s article http://www.economist.com/news/briefing/21588057-scientists-think-science-self-correcting-alarming-degree-it-not-trouble) is largely a ground-up issue. Until researchers are willing to hold themselves, their labs, and their techniques to the highest possible level of integrity, the quality of work that the scientific community as a whole produces cannot be improved.