Research Portfolio

Umer Gurchani

My research focuses on interdependence in complex systems, with particular attention to network structure, partial observability, and decision-making under uncertainty. I work across political communication and biodiversity conservation, using formal and computational methods to study how incomplete knowledge shapes collective outcomes.

Network analysis Computational modeling Uncertainty-aware conservation

Contact: muhammad.umer.gurchani@ens.psl.eu

Portrait of Umer Gurchani

About me

I work on problems where outcomes depend on relational structure rather than on isolated units. My research combines network analysis, computational modeling, and decision theory under uncertainty to study systems in which interaction structure is consequential but only partially known.

My doctoral research examined political homophily, retweet hierarchies, and the fragmentation of the French Twitter network. My current work moves into ecology and conservation, where I study food webs, biodiversity metrics, and adaptive conservation decisions. Although the substantive domain has changed, the underlying question remains similar: how should one reason and act when system-level dependencies matter but are incompletely observed?

Profile

Networkssocial graphs, food webs, structured interdependence
Methodscrawling, simulation, dynamic modeling, inference
Focusuncertainty, adaptation, prioritization, system structure
Approachformal, empirical, interpretable, interdisciplinary

I am interested in how interdependence structures risk, inference, and action — and in how formal models can help make better decisions when knowledge is incomplete.

Methods

  • Network analysis: community structure, interaction dependence, and graph-derived inference.
  • Dynamic modeling: systems evolving under uncertainty, feedback, and partial observability.
  • Simulation and counterfactual reasoning: comparing adaptive and fixed decision rules.
  • Interdisciplinary computational research: transferring formal tools across political and ecological domains.

Research trajectory

My earlier work focused on political networks, especially homophily, retweeting, and the structure of online publics. My current work focuses on ecological networks and conservation planning under incomplete interaction knowledge. The continuity between them is methodological: in both areas I study how network structure constrains inference and how decisions should adapt when the underlying system is only partially observed.

Current research directions

Ecological networks Food webs Conservation planning Partial observability Phylogenetic diversity Functional diversity Adaptive decision-making Value of knowledge

Current work

My recent work studies biodiversity conservation as a sequential decision problem in interdependent ecosystems. In one project, I examine whether precautionary conservation can perform poorly when ecological dependence is weakly understood, using predator–prey networks, partially observable environments, hindsight-based simulations, and inverse reinforcement learning. In another, I study how conservation priorities should shift between phylogenetic diversity and functional diversity as interaction knowledge accumulates, using a 148-year Northeast Atlantic fish stomach dataset together with simulated food webs.

PhD thesis

Political Homophily and the Role of Retweeters in the French Twitter Network

Université de Montpellier, 2021

In my doctoral work, I examined how political communities form and diverge inside the French Twitter network, and what that divergence implies for the idea of a shared public sphere. The thesis began with a simple but important question: does participation in political Twitter push all communities toward isolation, or is the effect uneven across ideological camps?

Main questions

  • How does political homophily evolve over time in the French Twitter network?
  • Which communities become more isolated, and which remain more embedded in the wider political network?
  • What does the behavior of retweeters reveal about political mediation and hierarchy?

Core hypotheses

  • Homophily would not grow uniformly across all political communities.
  • Some communities, especially at the nationalist or far-right edge, would diverge more strongly from the broader network.
  • Political retweeting would often operate less as horizontal diffusion and more as a top-down mechanism of ideological reinforcement.

Methods used

Graph construction

I separated the French Twitter political space from the broader global network and modeled it as a graph of users and interactions.

Community detection

I identified political clusters and tracked how their structural relationships changed over time.

Focused crawl design

I developed and used a focused back-and-forth crawl strategy to gather network-level data without needing a full platform graph.

Feature extraction & inference

I used communities as graph-derived features for predicting political affiliations on a continuous scale.

Retweet analysis

I analyzed retweet behavior to determine whether it bridged elites and masses or reinforced hierarchical communication.

Interpretive framing

I connected the empirical findings to questions about the public sphere, mediated hierarchy, and refeudalization.

What the thesis found

The central result was that homophily did not intensify equally across all groups. The strongest long-term divergence appeared in communities associated with more extreme nationalist politics. In parallel, the study of retweeting suggested that, at least in some communities, retweets were used less for open circulation and more for top-down ideological reinforcement. Together, these findings pointed toward a differentiated, asymmetric fragmentation of the political public online.

Selected publications

Right-Wing Twitter Users in France Exhibit Growing Homophily Compared With Left and Center Users

Social Media + Society, 2024

View publication

Political Homophily and the Role of Retweeters in the French Twitter Network

Doctoral thesis, 2021

View thesis record

Crawling political communities in Twitter and extracting political affiliations

arXiv, 2021

View paper

Treating Knowledge as a Conservation Asset to Resolve Present–Future Biodiversity Trade-offs

Submitted manuscript, Scientific Reports revision

Uncertainty-aware conservation in predator–prey networks using hindsight-based simulations and inverse reinforcement learning

Working manuscript, Nature Communications version

Three forms of temporal disorientation: A thematic analysis of subjective reports about Covid-19 restriction periods

PLOS ONE, 2025

View publication