I attended a talk by Prof James Evans about this back in 2019. I think about it most weeks.
What does the paper say, and what are the policy and funding implications?
Prof James Evans and his team Lingfei Wu and Dashun Wang had a front-page Nature article with the following finding in 2019:
Large teams develop science incrementally, small teams disrupt science
As Prof Evans says, people often argue that you need large teams that reach across the scientific and technological system to “provide complex solutions to 21st century problems”. His paper critiques this perspective.
Yes large teams do work, for particular goals.
BUT small, non-overlapping teams develop the most disruptive science.
The team found that small and large teams are best for different types of work
Small teams are best for:
- Exploration
- Subversion
- Originality
- Dissent
- Rebellion
Large teams are best for:
- Exploitation
- Succession
- Relevance
- Conformity
- Discipline
It’s Subversion vs Succession, Exploration vs Exploitation
Why? Small teams have less to lose and more to gain. Large teams have the opposite.
Small teams don’t seem to search and cite the same literature as large teams
Interestingly, we see that small teams have a different approach to engaging with the existing literature. For example, small teams are more likely to reach further into the past or cite less popular papers.
In contrast, large teams are more likely to cite “yesterday’s hits” – and speak to “yesterday’s audiences”, as Evans puts it.
This benefits large teams, as they can amplify their findings to existing, bigger audiences.
Small teams can find it hard to get people to listen.
Why do we see these different citation practices?
One reason may be that a disruptive small team may be doing more first principles thinking, or spawning a whole new field that’s not familiar with the existing history.
Policy implications when funding teams
Policymakers should be considering their overall goals in distributing particular lots of money. What do you want to achieve?
Do you want to optimise an existing approach, or discover a new one?
- To optimise: Fund big, networked teams
- To disrupt: Fund lots of small, decentralised teams
Another way to look at this, is to say that large teams hill climb, whereas individuals and small teams search and scatter. By exploring the landscape more widely, this wider search can unearth new approaches. Then it’s time for the larger teams to come in and develop those new approaches.
Hill climbing is good, as long as you’re on the right hill!
You have to have a taste for risk, to fund the kind of science that’s going to supplant current ideas
I chatted with James afterwards, and he shared this insight:
In his view, the personality types at DARPA are different from those at NIH.
- NIH has people who understand where to put money in order to scale promising developments.
- DARPA has more risk takers. More of the VC mindset.
Of course, we can see a parallel in organisational theory too. Large corporations are great for scale and distribution. Small start-ups are often where the most disruptive new paradigms emerge.
You need both large & small teams to advance science. But they’re suited to different outcomes.
Concerningly, Prof Evans notes that large teams are flourishing, while small teams are shrinking systematically.
What does this mean for current fields of science?
The big one I’m thinking of is AI. At the moment, huge amounts of time, talent and money are flowing into scaling existing AI paradigms – even though they have well-documented limitations, such as high training costs, brittleness and hallucinations.
Hype and groupthink reduce appetite to fund potentially disruptive new approaches, even when they may be most needed.
Conclusion: To fund new scientific breakthroughs, we should fund more small, decentralised, non-networked teams.
Even when (especially when!) they challenge the prevailing scientific consensus.
Prof Evans’ 30 min talk at the Researcher2Reader conference covers a lot more than what’s here, so I recommend giving it a watch.
Here’s where he starts talking about this study.
Further reading:
Prof. James Evans & team Fengli Xu and Lingfei Wu, have a 2022 PNAS article developing this thinking. It explores how flat teams drive scientific innovation.
Photo by Marco Montero Pisani on Unsplash