🧵14
Hope you enjoyed the thread!
Huge thanks to my coauthors:
@mariapalazzi.bsky.social, Albert Solé-Ribalta, Markus Nordberg, @agusticanals.bsky.social , and especially Javier Borge-Holthoefer

🧵13
Results (2/2):
The nested country–paper network reflects diversity but is fragile to geopolitical shocks.
If the top 7 generalist countries disappear, scientific production declines rapidly.

🧵12
Results (1/2):
The author–paper network, being modular, resists random turnover.
But it fails faster under targeted attacks—e.g., removing key researchers with high degree and who connect different groups.

🧵11
We track how many papers stay connected (Y-axis) as a proxy for cohesive scientific production.
We compare real networks (Original) with synthetic ones (EE, CP) that lack modular/nested structure to assess the role of these patterns in ATLAS’s robustness.

🧵10
We simulate attacks via percolation💣
◾️ Remove authors randomly (A), by importance (B), or by contribution to modularity (C)
◾️ Remove countries randomly (D), by importance (E), or by their role in nestedness (F)

🧵14
Hope you enjoyed the thread!
Huge thanks to my coauthors:
@mariapalazzi.bsky.social, Albert Solé-Ribalta, Markus Nordberg, @agusticanals.bsky.social , and especially Javier Borge-Holthoefer

🧵13
Results (2/2):
The nested country–paper network reflects diversity but is fragile to geopolitical shocks.
If the top 7 generalist countries disappear, scientific production declines rapidly.

🧵12
Results (1/2):
The author–paper network, being modular, resists random turnover.
But it fails faster under targeted attacks—e.g., removing key researchers with high degree and who connect different groups.

🧵11
We track how many papers stay connected (Y-axis) as a proxy for cohesive scientific production.
We compare real networks (Original) with synthetic ones (EE, CP) that lack modular/nested structure to assess the role of these patterns in ATLAS’s robustness.

🧵9
ATLAS is not just Big Science—it’s also international diplomacy.
Its scientific output can be affected by elections (US, Russia) or events like Brexit.
So… how resilient is this mixed system against different types of attacks?

🧵8
How does this affect network structure?
Surprisingly, both levels remain stable over time.
Emergent patterns:
🔹 Modularity (Q_ap) in the author–paper network
🔸 Nestedness (N_cp) in the country–paper network

🧵7
Looking at persistence over time:
👩‍🔬 Authors change constantly
🏢 Institutions and countries are far more stable
So, ATLAS shows high turnover of individuals but strong institutional stability.

🧵6
At the country level, we observe a nested network. These are common in ecology and support diversity.
Some countries (generalists) have broad presence, while others (specialists) focus only on core topics.

🧵5
We found that specific collaboration dynamics give rise to distinct structures.
At the author level, we see a modular structure: well-defined groups with dense internal links and few connections between them.

🧵4
For over a decade, ATLAS has produced ~1300 scientific papers.
Thanks to internal databases, we reconstructed bipartite coauthorship networks at two levels:
🔸 authors ↔ papers
🔸 countries ↔ papers
And here’s where it gets interesting…

🧵3
We analyze ATLAS, part of the LHC, the world’s largest particle accelerator. In 2012, the Higgs boson was discovered there.
ATLAS stands out from traditional science due to its scale: dozens of countries and thousands of researchers.
#BigScience

🧵2
We are surrounded by organizational systems. Here are some examples.
In recent years, there’s been growing interest in how network structure shapes key properties—such as robustness or diversity—in these systems.