,
13 tweets,
4 min read
Read on Twitter
Great new paper by @dashunwang and colleagues in next week's issue of Nature [1/12] 
One of the most robust finding in the "science of science" field is the correlations between team size and citation rates. But impact and novelty (or at least "interesting novelty") are not one and the same. [2/12]
And it turns out the type of work that small and large teams tend to be associated with differ systematically: small teams are more likely to "disrupt"; large teams are more likely to "consolidate." [3/12] nature.com/articles/s4158…
But what the heck is "disruption" in science and technology, and how might you measure it? Here @dashunwang et al. leverage a recent metric developed by @jdos_wot for patents [4/12] pubsonline.informs.org/doi/10.1287/mn…
Key intuition: if my forward cites reference my backward cites, then my work is cumulative; if they do not, it's disruptive. The disruption index (if you don't like the term, think of it as "interesting novelty") is distributed on the interval ]-1;1[. [5/12]
But how do we know that this index has a promising signal-to-noise ratio? @dashunwang et al. do a lot to validate the index. For instance, review articles consolidate; but *reviewed* articles disrupt (on average); Nobel-winning papers disrupt even more. Plus many more. [6/12]
Ok, it's a cool new metric. But it's not theirs. So why is the paper in Nature? Because thy found a great application for it. It turns out that team size correlates negatively with disruptiveness, using this new index. [7/12]
The finding is super robust: articles, patents, GitHub code repositories; natural sciences, life sciences, social sciences; across quantiles of the citation distribution. [8/12]
And here is the kicker: the result holds *within* authors as well as *between* authors. The same science dude/dudette is more likely to do disruptive work (yuck this word again) when working in a small team than when working in a large team. [9/12]
Anyway you should read the whole thing. Most exciting thing I have read this year. Lots of avenues for follow on (consolidative?) work. I liked it so much that Nature asked me to write a 1,000 words about it. [10/12] nature.com/articles/d4158…
Tentative policy implication: if you are a science funder and the rules of your solicitations end up de facto favoring large teams (e.g., because you buy into the Pablum about the inherent virtues of interdisciplinarity) you should stop. [11/12]
There is nothing wrong per se about interdisciplinary collaboration, but funders should let it emerge organically. Nudges in that direction might be counterproductive. [12/12]
Bonus tweet: Since the d-index is distributed over ]-1;1[, I predict that the inverse hyperbolic tangent transformation stands to gain popularity in the social sciences. And that, surely, is disruptive.
