Small teams vs. large teams in science

Here we analyse more than 65 million papers, patents and software products that span the period 1954–2014, and demonstrate that across this period smaller teams have tended to disrupt science and technology with new ideas and opportunities, whereas larger teams have tended to develop existing ones.

That is from a new Nature paper by Lingfei Wu, Dashun Wang, and James A. Evans.  Here is the NYT write-up by Benedict Carey.


Yes if you have been in the lab doing software.
The small lab group develops a great break through white paper, but leaves a horrendous pile pf spaghetti code in the lab.

That really is a different sphere. Software development is like an engineering team designing a structure with the various players doing their part to build it. It is all about process and systems. It has to be otherwise it becomes a horrendously expensive failure.

To come up with a new ideas, insight or discovery is a different endeavor.

It is related though, by way of the Mythical Man Month. And by the way of the problem of parallelism.

You may parallelize, but only some (well understood) problems.

When I worked in data processing managers who literally had no idea what we were doing would often opt to assign more people to the projects to get them finished on time. It was a running joke that adding people to a DP project would extend it's duration exponentially.

"no nor shall oblivion every lay them to sleep; the god is mighty in them, and he grows not old." --Sophocles

One popular computer software development process claims to be effective for teams of 3-9 people in fields far away from software development - scrum. See e.g. Any social scientists looking for a study topic could do worse than make an assessment of this.

The mythical man-month does not plumb the darkest depth of software development management, for which Dilbert is a remarkably faithful source. I have twice seen manager react to a lack of progress by instituting daily hour-long progress meetings.

I suspect that correlation is stronger than causation here.

As a lead you would only venture to organize and run a large team for a well understood and defined problem.

It's why prospecting could be one guy and a mule, whereas mining employs hundreds.

Given it takes decades to build a large team, how can a large team be disruptive with their papers written a decade or more into the pproject?

I'm sure plenty of people, scientists and their sponsors, are waiting for the disruptor who comes up with the cheap method of colliding particles at higher energy than in the CERN collider.

The GOP decided that creating jobs in Texas to do what is done at CERN sooner and at higher energy cost too much and killed jobs, and were counting on the Texas physicist disruptor to do it for a hundred thousand dollar private grant in Texas by 1995.

One large team is getting smaller, a team that started being built in 1993, from existing, smaller, NASA teams, has stopped collecting Mars data from Opportunity, with papers still to come over the next decade.

A private Mars program that is disruptive was started in 2002 after Elon bought existing efforts with his own cash, eventually sinking overr $50 million paying a "small" team to get a small dummy payload into orbit, but no papers, but with that success got added funding privately and from NASA, and now employs 5000 people, but producing few papers.

Elon needs lots of research done, producing papers with scientific results to keep refining the design of his Mars exploraation program, before he can deliver the Mars settlement that will enable lots of research on Mars, with papers likely following in significant numbers. Based on the past 50 years of research done by large NASA teams, we should expect disruptive science results to follow a Mars colony, except that with it spread over 75 years, it will be dismissed as not disruptive by those looking for free lunch science.

Although I bet the claim will be made that a few scientists looking at existing Mars data will make some disruptive paper on Mars, ignoring the tens of thousands on the team over 50 years that produced the data they used.

Good analogy.

if you are interested in this question, for further research, I recommend that you observe the argument between, on one side, Lisa Randall and Lubos Motl, who rationally expect their peers to know how to employ the viable technologies of today, versus those who say the most expensive colliders should be delayed (Sabine Hossenfelder and Peter Woit), as Other Technologies, not described in detail, will, as those people say, some day supersede what we now can do....

It might just come down to this, depending on how well informed you are: do you trust people who are the smartest people you have met with respect to their subject (let us say the subject is the expense of tens of billions of dollars on colliders) or do you assume that someday somewhere the next generation of people who obediently strive to get an A grade in every stage of their education as a physicist will somehow come up with a better way to spend money on collider-type physics experiments at the tens of billions of dollars level?

I say this: do not trust anybody who cannot tell you the difference between, on the one hand: truly understanding a subject as friends of 50 years understand each other, or. on the other hand, being somebody who:

understands the arts of persuasion.

If I was not clear enough, understanding the arts of persuasion is not really an accomplishment of very much merit. For obvious reasons.

(Final note: if you are a famous physicist, who happened to luck out and be fairly young when the most famous quantum physics textbooks were written by people your age, do not testify at a Congressional hearing in support of collider funding if you are not able to pretend you do not think you are by far a superior human being, in both religious and secular ways, because you aced your differential geometry classes, and so on and so forth. First off, you are not, for that reason, superior. Differential geometry, as it is currently understood, is fairly trivial, and while you may not know that many people know that ---- well, many people do know that, trust me. Second, even if you were as intellectually superior as you thought you were, Professor W., acting like you were as intellectually superior as you thought you were - whether your pride was sincere or not - was counterproductive.)

If you are the one or two people who have read the foregoing comment and have understood it:

you are welcome, by the way.

There are things in this world that are more underestimated than unconditional kindness ----

trust me ----

there are many such things.

But you cannot buy at any price good advice ....
don't be proud ....
don't see yourself as understanding more than you know ...

because, once you have learned to be kind to a creature who never had a friend in this world, you will understand ....

Differential geometry, as it is currently understood, is fairly trivial.
But the only intellectual miracle you will ever experience is not trivial at all ...
There is nothing more important than being a friend to a creature who never had a friend in this world.

Let us now say good-by to the years when we knew that what we read and heard (what we once read and once heard, or often read and often heard) was something a human being or an AI had said....

but before we say good-bye in that context ....

let me remind you.

Many people who have posted here - between one and two hundred --- have called me names, good names sometimes, angry insulting names sometimes. I apologize if I should have paid more attention, but I was not all that encouraged by the praise and I was, at most, mildly amused by the criticism.

Keep reading if you want to understand something important ....

But here, in 2019, the last online year before any of us are reasonably certain, without human presence, that we are reading online the words of a real human being ....

here in 2019, I say to you, one last time ....

It is no small thing to be a friend to a creature who never had a real friend in this world.

That, my friends, is the sort of thing a human creature says.

God loves us all. I wish you understood that as well as I do. And maybe you do. God is angered and disgusted by bullies and by people who take advantage of other people. You bullies, you people who take advantage of others in a sinful way, you can be better!
Anyway, whatever you do, please, for the love of God, refrain from wanting to be the sort of person who rejoices in being a bully or in being a person who takes advantage of other people.

good news.

Scott Alexander on this from last November

"Explanation 1 – that progress depends only on a few geniuses – isn’t enough. After all, the Greece-today difference is partly based on population growth, and population growth should have produced proportionately more geniuses. Explanation 2 – that PhD programs have gotten worse – isn’t enough. There would have to be a worldwide monotonic decline in every field (including sports and art) from Athens to the present day. Only Explanation 3 (“low-hanging fruits”) holds water."

The low hanging fruit evidence is only true if you think real innovation has been universal. If however, there were only a few countries with the right institutional and cultural framework to produce real, macro innovations (cue Joel Mokyr) then problems with regulations or institutions or culture in just a few countries could severely disrupt world innovation and productivity. After all both the first and second Industrial Revolutions were driven by a small handful of countries. Why not the decline of innovation?

Excellent comment.

@CMOT - I'm reading a good book by Lucio Russo, "The Forgotten Revolution: How Science Was Born in 300 BC and Why It Had to Be Reborn" (1996: Springer --note the academic publisher--translated from the Italian), and the thesis is that modern inventions have existed well before the modern era but have died out. I've not finished the book but a plausible thesis is they died out due to trade secret rather than patents. Recall Roman concrete (quicklime + ash + heat) had to be reinvented after the Roman empire fell and after the Dark Ages by medieval inventors, since the invention was lost due to trade secret. So it's not "low hanging fruit" but rather lack of followup. If society spent as much money fighting cancer as they have building better war machines, we'd have a cure for cancer by now (such as the CAR-T promising line of research).

Bonus trivia: CNTRL + F + "patents" --> *finally* gets a hit! :)

When literary is rare and paper and travel are extremely expensive a lot of ideas are going to look like they are trade secrets, when in fact they are just very expensive to communicate to the person who needs to learn them.

I've never seen any statement of the "low-hanging fruit" argument that wasn't simply circular reasoning.

It's really quite something.

Scott seems like a nice guy but he has no idea what he is talking about.

Longer excerpt from Alexander on The Three Reasons for the slow down (the paper Tyler links to above would fall into the #2 reason):

"1. Only the best researchers in a field actually make progress, and the best researchers are already in a field, and probably couldn’t be kept out of the field with barbed wire and attack dogs. If you expand a field, you will get a bunch of merely competent careerists who treat it as a 9-to-5 job. A field of 5 truly inspired geniuses and 5 competent careerists will make X progress. A field of 5 truly inspired geniuses and 500,000 competent careerists will make the same X progress. Adding further competent careerists is useless for doing anything except making graphs look more exponential, and we should stop doing it. See also Price’s Law Of Scientific Contributions.

2. Certain features of the modern academic system, like underpaid PhDs, interminably long postdocs, endless grant-writing drudgery, and clueless funders have lowered productivity. The 1930s academic system was indeed 25x more effective at getting researchers to actually do good research.

3. All the low-hanging fruit has already been picked. For example, element 117 was discovered by an international collaboration who got an unstable isotope of berkelium from the single accelerator in Tennessee capable of synthesizing it, shipped it to a nuclear reactor in Russia where it was attached to a titanium film, brought it to a particle accelerator in a different Russian city where it was bombarded with a custom-made exotic isotope of calcium, sent the resulting data to a global team of theorists, and eventually found a signature indicating that element 117 had existed for a few milliseconds. Meanwhile, the first modern element discovery, that of phosphorous in the 1670s, came from a guy looking at his own piss. We should not be surprised that discovering element 117 needed more people than discovering phosphorous."

I suspect the results in the linked paper are due to a version of Alexander's reason 1.

Most researchers are in working in a publish or perish environment and need a sort of background level of published results just to justify their existence and budgets. The less important the findings the easier it is to share credit and be sure that everyone can at least stay in the game. Reciprocity means that circles of non producers can stay in the game.

I'd test this theory by looking for logrolling, sumo wrestler style 'cheating' reciprocity, and network effects.

Perhaps high energy particle physics is the exception to this. I would also include the Manhattan Project which was a very large team and perhaps was one of the most disrupting technological advances of our lifetime.. CERN, Fermilab, SLAC, and a bunch of others are modern examples of large teams.

High energy physics (HEP) is, in fact, a perfect example of this. The typical HEP project involves thousands of people, and hundreds of paper authors, to characterize a result that is exactly predicted by the Standard Model -- so much so that detection bounds and uncertainties are exquisitely mapped out ahead of time. Opinions vary on whether this is extremely dull, or whether it's technically amazing that all this works so well -- I often think both things at once at high energy physics seminars. On top of all this is a hope of discovering something new & important, which hasn't happened in collider experiments in many decades, and for which there's good reason to be pessimistic looking ahead.

The standard model philosophically or (ego from sound) derives from the overlap of persistence and sacrifice. Or, from a classical liberal perspective, Hades...the unseen--the Bach G major--sculpture as the pacifist direct response to slavery ((as if a humpback whale could not admire figurines)),, Hades...the question...what is a standing ovation like after a tennis match? Indeed, the poet tasks God. Or to use the version of literature, the efficient displacement--galloping. As Aristotle writes, "Happiness is a habit; it isn't random."

Average vs. marginal. The real question is to what extent the field is tilted for or against small innovators or large ones. IF large teams used to be innovative but now are not, that is a big problem. If this is the equilibrium ideal response given today's scientific problems there is no issue. Correlation here almost literally tells us nothing that would help in making policy or deciding on what problems exist.

Good contender for obvious conclusion of the year. Prior art: the law of large numbers, regression to the mean, any human committee ever assembled.

If an idea is groundbreaking, wouldn't the principle authors be less inclined to let others in? Strong ideas cause fewer authors.

Comments for this post are closed