Productivity and reproducibility: conflicting pressures on scientists?
Dorothy V M Bishop
In 2015, a symposium was organised by the Academy of Medical Sciences in London on the topic of "Reproducibility and reliability of biomedical research." Speakers from a wide range of backgrounds presented evidence that there is indeed a "reproducibility crisis": over time, scientists have generated more and more publications, but confidence in findings has declined. The report on the meeting concluded that there was no single cause for this situation: rather, a toxic mix had grown up over time, whereby top-down incentive structures encouraged researchers to over-hype work and cut corners, and meanwhile education in methods of research and analysis had suffered. Just as there is no one cause, there is no one solution, but it is clear that institutions can play a part by giving greater value to work that is done carefully and reproducibly, with open methods, analyses and data. Working that way is satisfying but it takes time, and often seems impossible when there are pressures to produce a high volume of work that is "groundbreaking" and original. I will make some recommendations for a move to "Slow Science" that can help us recover from a period where an emphasis on productivity has hindered rather than helped scientific progress. These include pre-registration of studies, use of simulation when planning data analyses, making scripts as well as data and publications openly available, and paying attention to prior literature to create a cumulative body of research.