In a new paper, Carlos, José, Jacopo, Kent and I tackle this simple problem: if we take a Lotka-Volterra system with random parameters, and let the dynamics elapse, how big will the final, persistent community be?
Biological networks show strong departures from simple models of random graphs. For example, they display broader degree distributions, high modularity, and strong preponderance of certain motifs.
One might be tempted to interpret these features as a signal of a selective force pruning the space of possible networks, resulting in empirical networks possessing certain features.
In one of my all-time favorite papers, Ricard Solé & Sergi Valverde proposed an alternative explanation: these features might be by-products of how the network was assembled. They dubbed this the “network-spandrel” hypothesis, referencing the famous paper by Gould & Lewontin.
In a new paper just published in Ecology Letters, Dan Maynard, Carlos Serván and I show a simple model of ecological assembly where by slightly tweaking the rules of assembly we can obtain dramatically different network structures—a paradigmatic case of network spandrels:
In a new paper published today in PNAS, Jacopo and I try to extract as much information as possible from an ostensibly meager source of data: a list of the names of all the researchers working in Italy, at the Centre National de la Recherche Scientifique in France, and at public universities in the US.
We show that by using simple randomizations, one can highlight several interesting facts about these different academic systems.
The work is a bona fide exercise in style: by introducing subtle variations of the randomization algorithm, we show that in Italy researchers work in the region where they were born and raised, while in the US geography does not influence the distribution of researchers; in France, we can detect academic couples working in the same unit; we demonstrate that academic nepotism in Italy (the focus of a previous paper of mine) is declining; finally, we show that in the US immigration is field-specific.
It is not unusual to read the tirade of a senior scientist complaining that science was better back then, when papers were fewer, and ideas better (a perfect example of this genre is here). Usually, the conclusion is that we should publish less, lest producing lower-quality science.
These considerations are based on a quite precise hypothesis—that a scientist can either produce many papers, or produce fewer good ones. Detecting such a trade-off in actual data is quite difficult, though, as scientists vary dramatically in productivity, as well as field of study.
Matt and I tried a different route, and compared scientists with themselves: does a scientist produce better papers in the years when she’s most productive? For testing our method, we took the members of the National Academy of Sciences, and reconstructed their publication history. (The rationale being that their best papers must be of high quality).
We found that these scientists tend to produce their most recognized work in years when they’re most productive. However, they also tend to produce their least impactful articles during the same productive years. This is consistent with the “random impact” hypothesis: by publishing many papers, scientists sample their distribution of good ideas more thoroughly, leading to higher maxima and lower minima.
You can read the paper here: