Share

Listen: Small research teams are better than big ones at disrupting science

Interviewer: Nick Howe

Scientific research today seems more complicated than ever. There’s simply too much out there for one person to be an expert on everything. This may be one of the reasons why research teams are getting bigger and bigger. And big teams can bring benefits. For example, there’s a strong positive correlation between the numbers of authors on a paper and the number of citations it receives. But are there any advantages of working in a smaller research group? Are we losing something by focusing on bigger teams? A paper coming out in Nature this week has been looking at how the size of teams, in all areas of science and technology, affects their outputs. I spoke to the corresponding author of this paper, James Evans from the University of Chicago, who’s been crunching the numbers.

Interviewee: James Evans

So, we looked at 65 million teams – scientific teams, inventive teams, teams that develop software – and we identified the degree to which the products from each of those teams either developed things or disrupted the frontier of ideas and technologies. And we found that with each additional member on those teams, there was a dramatic drop in the likelihood that their products would disrupt the frontier of science.

Interviewer: Nick Howe

But what does it mean to be disruptive? James and his colleagues were trying to distinguish between papers that either upset the scientific status quo, making a potentially big sudden change, and papers that consolidate and develop existing knowledge. So, what is the value of disruption?

Interviewee: James Evans

For the vast majority of science and technology, our purpose is not just to get a few more eyeballs, in the same way that a big production studio is more likely to pick Transformers 9 and fund that over an independent script, in the context of developing new films. If our purpose is to develop and feel new ideas which have the potential to really change science and technology, our findings show that it’s much, much more likely that those are going to be sourced from small teams.

Interviewer: Nick Howe

But how do you measure how disruptive a paper is? In this research, disruption is all about the kinds of citations a paper gets. The authors argue that disruptive papers are often the beginning of something new, so authors tend not to cite research that has come before it, whereas consolidating papers form part of an ongoing body of work, so when they get cited, it will be alongside older and newer research. So, James uses the way in which a paper has been cited as a metric for disruptiveness. Pierre Azoulay from the Massachusetts Institute of Technology, who’s written a News and Views article on this research, thinks that there’s a convincing argument to suggest that citations and disruptiveness are linked.

Interviewee: Pierre Azoulay

And so, they’ve done sort of a number of checks that really establish the plausibility of this sort of disruptiveness interpretation. So, for example, they’ve looked at Nobel-winning contributions and conversely, they’ve looked at review articles, right, so articles are in some sense written without necessarily introducing new ideas. So, for example, they’ve been able to check that the Nobel-worthy contributions are very disruptive according to this index, whereas the review contributions are very consolidative according to this index.

Interviewer: Nick Howe

So, having checked the metric by assuming that Nobel prizes are ‘disruptive’ and review articles are not, the team are able to back up their claim that small teams are more disruptive. But does this new finding hold up cross the board? In some fields, large teams are often the norm – think of the huge collaborations in physics projects like the Large Hadron Collider – whereas in mathematics papers, typically there’s only a few authors. How universal is this newfound correlation?

Interviewee: Pierre Azoulay

They’ve established that that’s a pretty robust finding, so it holds across disciplines – it holds for publications, for patents and for code repositories. And then they do something that to me really sort of nails it, which is that in the case of papers, they show that the result holds even for a given author, meaning, let’s imagine that you observe that individual in a small team and in a large team, you will see that the articles that were written with a large team tends to be more consolidative than the articles that were written when that same individual was part of a small team.

Interviewer: Nick Howe

So, this paper suggests that larger teams are less disruptive, and this holds across different fields of study and when looking at the same authors. But this does beg the question – why? Why do small and large teams behave differently and what could be driving this? James has some ideas.

Interviewee: James Evans

So, for example, small teams are much more likely to search much more deeply into the distant past, they’re much more likely to pick up and cultivate ideas that were unpopular that they built on, and small teams are more likely to be flat, so a higher proportion of team members are likely to perform multiple roles on the team, they’re less likely to be well funded. And it turns out that all of these features end up contributing to the fact that their work – if and when it becomes important – becomes disruptively important.

Interviewer: Nick Howe

James has some thoughts on other potential mechanisms. Perhaps large teams with big budgets are less able to change their approach – if you’ve spent billions building infrastructure to look for, say, gravitational waves, it’s not easy to then change direction and look at particle physics. A small team’s flexibility might allow them to follow new, riskier avenues. At the moment, it’s unclear what the mechanisms are, so what happens next? Here’s Pierre.

Interviewee: Pierre Azoulay

We need to proceed with caution here. It’s not a policy paper. It doesn’t establish sort of a causal mechanism. What we need now is to have a few important funders who take those results as sort of a hypothesis to be tested and actually run experiments to think about what happens when you sort of vary the extent to which you push scientists to form large teams versus small teams.

Interviewer: Nick Howe

One thing that isn’t clear is how much are the findings of this research to do with the changing nature of science? Fifty years ago, when smaller teams were the norm, was research more disruptive? There is also a potential problem with the metric of disruptiveness. It can only be applied to papers after they have been out for years and have accrued citations – it can’t be used to assess recent papers. While small may be splendid, big isn’t necessarily bad. Pierre and James think we need both little and large teams for the advancement of science.

Interviewee: Pierre Azoulay

Science needs both kinds of contributions. It’s not like incremental innovation and elaborating initial ideas, it’s not like this is not an important activity. It’s a super important activity and we need a balance.