As we have written previously: “Modern research-intensive universities present a paradox. On the one hand, they are dynamic, vibrant institutions where researchers use cutting-edge methods to advance knowledge. On the other, their traditions, structures, and ways of working remain rooted in the nineteenth century model of the independent scientist” [3]. In our view this underlying culture, where research groups operate effectively as artisanal small businesses, when combined with human cognitive biases and the current incentive structures, interact to contribute to problems of poor reproducibility and replicability.
The research ecosystem is therefore complex and highly interconnected. Incentives, for example, may be embodied in institutional hiring and promotion practices, but are also influenced by the demands of research funders, the (explicit or implicit) requirements of journals and publishers, and the enthusiasm of researchers themselves to make discoveries that will advance their field. In the majority of cases, these influences are unconscious or implicit, and while introduced for the best of reasons, they often have unintended consequences that have a negative impact on research quality.
To bring about change, efforts will need to be coordinated across the research ecosystem [4]. For example, open research practices (e.g., sharing data and code) will require supporting digital infrastructures, training in the skills necessary to make use of these infrastructures, mandates from funders and publishers to require open research practices where appropriate, monitoring of performance to create motivation for improvement (which in turn requires coordination across institutions, funders and publishers), and recognition of these practices in institutional hiring and promotion processes (again to create an incentive).
In other words, no one actor in the research ecosystem can bring about meaningful change on their own. Indeed, individual actors may place themselves at a disadvantage if they work in ways that benefit their research, if this is not aligned with what is rewarded within current incentive structures. For example, whilst there are advantages to individual researchers in engaging in open research practices [5], institutional recognition of this will be necessary for career advancement. And if this recognition is not sector-wide, then a researcher working in alignment with their own institutional framework may be at a relative disadvantage when moving to another institution if such practices are not valued there.
There is therefore a need—for example, through meta-research activity—to understand how the research system affects the quality and robustness of research outputs, and consider how a programme of measures across different sectors can work in concert to improve research quality. We need to distinguish between research integrity (the processes, working practices and incentives that contribute to the quality of research outputs) and researcher integrity (i.e., the specific behaviour of individual researchers). Whilst both are important, we feel that focusing on the former (i.e., the system) will generate more benefits.