Collectivism and Big Team Science
The first annual Big Team Science conference starts today. In its honor, I want to highlight some possible benefits of big scientific collaborations that I haven’t seen discussed.
An obvious benefit of big collaborations is that they can collect larger datasets than were previously possible. More data means more statistical power. Big teams can also facilitate, in the social sciences, the breadth of sampling needed to generalize a result beyond a few cultures.
Big collaborations also facilitate specialization of researcher roles, such as in data analysis, statistics, and study protocols. This is important because specialization has been a key to increased productivity across many human endeavors, and may also increase scientific productivity. This, too, has been discussed.
What I haven’t seen discussed is how big collaborations might improve science through mutual aid, as a result of their collectivism. UPDATE 29 OCTOBER: It’s already happening.
“Collectivism” and “mutual aid” are loaded terms for some; mutual aid was popularized by one of the founders of anarchism, and collectivism may recall certain catastrophes of the 20th century, but here I want to use both terms as rather neutral descriptors.
Individualism in science
For the last few hundred years, science has been dominated by rather individualistic societies like the US and the UK. Even in those countries, science isn’t a wholly individualistic enterprise, because many researchers work in teams. But in most domains, the research teams tend to be small, and they are more likely to be competing with each other rather than to be helping one another (the “neoliberal” system).
In many places, in most fields, a researcher with a permanent position will apply for a grant with a few others, which if won provides support for a small group of PhD students and postdocs. But only a minority of researchers can expect to win grants each round — recent success rates where I live in Australia, for example, are less than 20%. That’s for just three years of funding, and the selection process is quite unreliable, so even top researchers can’t be assured of continual funding.
The feast-and-famine system hinders long-term plans and pushes researchers to concentrate on incremental research, as only that will consistently deliver the results and associated publications needed to have a chance to receive funding in the next grant cycle.
Some researchers manage to devote some percentage of their grant income to new, risky lines of work, or work that goes in a different direction than their publication record. This is something that researchers wish they could do much more of. Seventy-eight percent of researchers who responded to one survey said that they would change their research program “a lot” if they had continual, permanent funding.
To have a good chance at having continuous funding, researchers spend much of their time applying for grants. There is little time for actual science because of grant-writing and other demands such as as ethics applications, expense management, grant management, teaching, and project management.
The combined burden on researchers today could make the most self-reliant of individualists look to others for help! Homo economicus, who by definition looks out only for himself, will join a collective in the right circumstances. Mutual aid can emerge in a society filled with entirely self-interested, rational actors, as scholars in the fields of game theory and evolution have shown.
Funging funds
As individuals, researchers can’t normally do much about the problems of intermittent funding or not enough time for research. Pooling of resources can help, and this has long been done at the small scale of within a university. A university’s researchers will sometimes shift funds around a team or institute to keep everyone employed and their lines of research going. A progressive dean might do the same.
I call this shifting around of funds “funging”, my verbing of “fungible”, which unlike funging is actually a word. Funging is possible partly because typically funding bodies don’t require that researchers stick to the items and costs they wrote into their budgets, and partly because university finances aren’t audited very carefully.
Funging can also happen in large-scale collaborations, even though moving money between universities can be more difficult. The advent of remote work facilitates this. Imagine a research assistant who has been working on a large-scale collaboration project, helping out with various organizational tasks, which might include drafting IRB/ethics applications or managing the data coming in from different labs.
Let’s say that the funding for the project is coming largely, although not entirely, from just one or two of the labs, and it’s now run out before the project has completed.
Other labs in the collaboration might step in to continue paying the research assistant by diverting a portion of the funds they have from somewhat-unrelated research grants. If the research assistant is in a different city or country, in the days of mandatory office attendance and paper-only HR forms this wouldn’t have been possible, but now that remote work is widespread at universities, it is more feasible.
Funging non-monetary resources, such as labor, is even easier. When one person is no longer able to carry out a task in a large collaboration, someone else may be able to fill in.
So far, I’ve been referring to labs shifting resources to each other to further a project they’re already all contributing to. But as the bonds of mutual trust increase, individuals may become willing to provide resources for projects they aren’t directly involved in. The collaborative’s resources begin to resemble a joint fund that members use to tide over individual shortfalls, much as labor unions have assistance funds to help members that are in a bad spot.
In 2021, the main basic science agency of the Australian government disastrously failed to announce, until Christmas Eve, the outcomes for grants that were applied for in February. The existing funding for many of the applicant small research teams was due to run out on the first of January. Not knowing whether they would still have a job after Christmas, before the outcomes were announced many postdocs had to leave academia for more secure jobs. In a large-scale collaboration with more diverse sources of funding, thanks to mutual aid, one can imagine these postdocs having a bit more job security in academia, and the principal investigators having more confidence that they could continue their research into the new year.
From conservatism to innovation
Many large-scale collaborations were set up over the last several years to collect big datasets using established methods applied to very conventional, mainstream lines of work. There was a widespread view in the associated fields that for these methods/questions, a lot of data, in the context of a rigorous protocol, would yield important results. What brought these large collaborations together was, then, essentially incremental research.
Perhaps a goal of largely-incremental work is the only thing that can bring together into a collaboration a large number of researchers unknown to each other. In such a situation of methods that are already developed or rather straightforward, an individual researcher can be fairly confident that the collaboration’s members will agree on most things relevant to the project, even if an individual researcher does not know whether they have major differences with the others in their theoretical perspectives or their research styles.
As the project is carried out, however, inevitably there are dozens of little decisions to be made on details of a study’s methods, the logistics of how to get things done, and how to distribute the work within the collective. The researchers have to learn to work together and they develop decision-making procedures, at least implicit ones.
By the end of such a project, a high level of trust will have developed among some of the participating researchers, if their interactions went well and resulted in a happy outcome. Of course, some large teams will instead dissolve due to disagreements, or fail to develop effective decision-making and efficient procedures, causing many to drop out. But those that survive may in many cases have developed into an effective organization.
With this foundation of organization and human relationships established by an initial project, the large-scale collaboration may be able to move beyond the conservatism of its initial project, into the domain of risky research.
Imagine a sort of tree of knowledge, and imagine that this tree is specific to a research area. Any low-hanging fruit may have been stripped from the tree long before the big team came along. A big collaboration is like a ladder that makes it easy to pick fruit that’s hanging higher up. Due to the larger amount of data the collaboration can obtain, research questions beyond the reach of an individual lab can be answered.
This benefit of a large collaboration may continue after the initial project, as there is likely to be more fruit very much within the collaboration’s reach. That’s great, but what I’m interested in here is the potential to take on projects that would be risky even for the big team. New directions or methods of uncertain viability, or ones that carry a risk of uninteresting results, the kind of thing that individual researchers try to squeeze in on the side. High risk, but a chance of a big reward, perhaps a breakthrough.
With its large pool of resources, a big team has a greater ability to complete risky work. This may be especially true in the social sciences and psychology, where expensive equipment isn’t needed to get a project done. The ability to collect data from broad samples of human participants can be enough, together with the associated labor of professors and PhD students.
Professors whose salaries are paid by their universities on an ongoing basis typically have considerable freedom in what they work on. So for a research collaboration that includes many professors, although shifting money among universities can be difficult, shifting work between universities is not.
For example, large-scale collaborations in psychology have benefited from the labor of faculty and students at American liberal arts colleges who rarely receive any research funding, as some of us discovered when we launched a few large-scale collaborations eight years ago. If a team is organized well, even members with few resources can both contribute and benefit. Thus, while any individual professor may not be able to devote enough time to get a risky project done, the collective can.
For members to agree to pursue something risky, however, they need to have a lot of confidence in the collective generally, and in the legitimacy of how its decisions are made.
Building organizations with the required good governance is hard work and doing it right is not easy, so I don’t want to minimize the difficulties here. Some big team scientists have recently highlighted the potential for unaccountable leadership and mega-mistakes. I don’t know that any effective templates for avoiding this have been developed. But that’s part of what the Big Team Science conference is about.
Collective action
Many of the difficulties that scientists face, such as intermittent funding, are not normally thought of as collective action problems. A prominent exception is the problem of open access publishing and academia’s over-reliance on journal prestige. In that domain, reformers have long bemoaned the failure of individual researchers to stop submitting their work to corporations that suck millions from our universities.
Today, researchers could choose to support new and often better (in their expectations of rigor and transparency) academic journals and publishing platforms, but as is often pointed out, the career incentive for submitting one’s latest article to an expensive journal with a legacy of prestige is high, and there is no reward for supporting more recent initiatives, at least not until a majority do so.
Fortunately, large-scale scientific collaborations don’t have that problem of needing to reach a distant tipping point before benefits accrue to the individual. Large-scale collaborations can form at a rapid rate because being part of one is likely to yield a reward of an important scientific result, and a publication in, yes, a legacy prestige journal.
As we have seen, then, for short-term gain, some researchers will end up a part of a collective that can address other problems in academia.
Regarding the intermittent funding problem, the only remedy usually considered is reform of government funding policies. In other words, the problem is one of political lobbying and winning elections. But given how sclerotic some funding agencies behave regardless of who is at the top, we shouldn’t look to them to solve our problems, even the problems that they created.
For more than a decade, Michael Nielsen has been explaining how scientists can help themselves with new organizational forms—back in 2011, he published a book entitled Reinventing Discovery: The New Era of Networked Science.
In his writing, Nielsen has used the example of how the invention of insurance enabled a new era of risk-taking in trade. Commercial insurance was developed in the 1300s in Italy so that merchants could survive the occasional disaster of their goods being lost at sea. Today, for a small recurring payment, people can insure all sorts of things, including a loss of income by taking out an annuity or unemployment insurance.
Although insurance contracts could conceivably provide some bridging funds when researchers fail to win a grant, conventional science funders don’t allow researchers to spend their funds on that. Scientists also aren’t allowed to take out loans for their labs.
We researchers are therefore in a position a bit like that of businesses in some Islamic countries. Because loans that include interest payments are prohibited by the Koran, Islamic jurists have had to come up with other, creative ways of financing things.
Most recently, the cryptocurrency movement has developed new forms of organizations, which are now being applied to the funding of work that traditionally goes unfunded, like the development of treatments that, while saving lives, have a market of sick people that is too small to interest traditional pharmaceutical companies. “Smart contracts” bind members of these organizations, Kickstarter-like, to fund certain things when specific criteria are met.
So big team science is just one kind of new research collective, but it is possibly the one with the most momentum.
Big team science versus traditional scholarly societies
If collectivism like this is so feasible, and allows researchers to help each other and do important work that wouldn’t otherwise be done, then why doesn’t that already happen via the existing and longstanding scientific societies? I think there are a few reasons for this.
First, scientific societies aren’t really about getting research done. They often conceive of themselves as advancing their field by organizing avenues for disseminating and discussing work, through academic journals and an annual conference. They also fundraise, lobby for more funds for their field, and give awards. The awards traditionally go to individual researchers rather than teams.
And although scientific societies are often specific to fairly small fields, this still might be too broad a remit. They don’t focus on particular research questions.
Another part of the problem is that most scientific societies are old and run largely by people who succeeded in the current system and thus aren’t focused on solving the problems that beset the majority.
Some new scientific societies may be different, such as the Neuromatch society. Neuromatch started in order to put on an online conference during the pandemic in 2020.
Founded by a core of researchers with coding skills, Neuromatch provides computational neuroscience training and networking opportunities online. Using the (largely non-financial) resources they possess as a large collective, they have turned their attention more recently to improving scientific publishing, and are working on a new publishing platform.
It is highly unlikely that any one organization or platform will solve the open access problem, that problem is simply too big. But that organizations like Neuromatch would even consider it does illustrate their broad potential. Let’s keep pushing for new, social solutions to our longstanding problems.
