Radically Collaborative Science

Some contemporary research is carried out by very large research groups consisting of multiple labs, from different scientific fields, working in different parts of the world. In such contexts, social and material pressures shape the collection, analysis, and interpretation of data; and local norms shape each researcher’s decisions about when to treat something as data, and when to treat it as noise (Douglas 2004; 2009).

The scientific questions addressed by large-scale collaborations require contributions from multiple researchers, who draw on tools and techniques from different disciplines, and who are working in different locations. And since these research communities are constituted by scientists who operate within partially distinct credit economies, and who make judgments on the basis of partially distinct networks of disciplinary skills and norms, their research outputs are typically shaped by “a chaotic web of micro-interests and local values that penetrate the study bottom-up” (Winsberg et al.

In this context, subtle forms of epistemic distortion can emerge, even where everyone is committed to engaging in good epistemic practices. There is no obvious way to keep track of the impact of the adjustments that are made across the collaboration; and there is no way to know whether their effects will aggregate, cancel each other out, or amplify one another. At each stage of the research process, incoming data that has been shaped in accordance with local norms can be re-evaluated in accordance with locally salient considerations; and the impact of this fact is often obscured due to the individual or team who is responsible for writing papers and producing other research outputs. Since it is unlikely that they have participated in the collection or analysis of data, writers function as a distinct component of the collaboration, and they tend to write in accordance with their own normative assumptions about what is worth presenting and what can be ignored. Of course, the output of this process is typically inter­pretable. But it is often impossible to determine how the collaboration arrived at the reported results.⁴

To clarify the ethical implications of this situation, we return to Dang’s (submitted) tripar­tite distinction regarding different kinds of responsibility. In radically collaborative research, papers and other research outputs are typically attributable to widely distributed networks of researchers, technicians, statisticians, and writers.

The scope of these kinds of research projects can thus compromise the otherwise stable link between the best features of the scientific credit economy and the process of scientific research. This can occur when highly distributed and massively interdisciplinary research leads to the breakdown of the norms that typically constrain individuals in the pursuit of scientific credit, because the communal oversight of adherence to those norms has been weakened. First, co-authors on such projects will often object to the retraction of a paper, attempting to hold on to any credit they have received for a massively multi-authored paper. Second, there is little motivation to take part in attempted replications of large-scale collaborative research projects with problematic implications. The credit one receives as the 164th author in a field of 400 authors is both minimal and diffuse; and the credit one would receive for pur­suing a replication, which is not likely to establish priority, is even more minimal. Third, no one can reasonably be punished if problems arise within a large-scale collaboration, as both answerability and accountability become so diffuse in a network of hundreds of distributed researchers that it only makes sense to hold the collaboration as such accountable; unfortu­nately, attempts at censoring radical collaborations are likely to fail, since such collaborations are often kludged together for a specific purpose, and they often dissipate after the projects are completed. Fourth, and finally, this kind of research has a serious impact on the checking mechanisms that are typically at play in the peer review process. When “research projects engage a greater proportion of the scientists working in a specialty, there are fewer and fewer competent scientists available to referee the resulting research” (Wray 2017: 129ff). In the con­text of massively interdisciplinary research, which draws on many different research areas, it is unlikely that anyone is competent to evaluate the research project as a whole; consequently, the more limited pool of referees is likely to impose significant constraints on the ability of the scientific community to track cases of where questionable research practices are employed, or where fraud has been carried out.

Such collaborations thus compromise the checking mechanisms that serve to prevent fraudulent claims from being entered into the scientific record. And where problems emerge— whether unintentionally, or as a result of fabrication and manipulation—it becomes difficult, and perhaps impossible, to figure out who is answerable-responsible for the scientific claims that are made, and who should be held accountable when problems arise, even if it is possible to attribute a paper to the widely distributed research group. Intriguingly, similar problems arise where heavily managed and systematically biased research is produced in accordance with a single set of non-scientific values. Here too, questions arise about who is responsible for the production of a scientific claim, who is answerable for it, and who should be held accountable when problems arise.

25.4