Machine Learning Predicts Police Violence in Latin America In a groundbreaking application of machine learning to the study of state violence, recent research delves deeply into the factors driving extrajudicial killings in Colombia and Mexico. Using advanced

"These insights hold promise for future interdisciplinary research as well as practical policy design (...) it represents a notable advance in empirical understandings of state violence, employing state-of-the-art machine learning tools"

scienmag.com/machine-lear...

Si tan solo hubiera sabido Si tan solo hubiera sabido

#LaCronicaDeHoy "Esta investigación es relevante para entender las dinámicas del conflicto en México más allá de las acciones por parte de las organizaciones criminales."

www.cronica.com.mx/opinion/2025...

My most recent published article has received some media attention! 👮🚓

Check it out here 🧵

See my full article here: nature.com/articles/s4159…

#PoliceViolence #MachineLearning

Predicting police and military violence: evidence from Colombia and Mexico using machine learning models In this section, I show the results of the predictive performance of four machine learning models trained to identify instances of police and military misconduct in Colombia and Mexico. The models include a regularized logistic regression model (Lasso), an extreme...

Predicting police and military violence: evidence from Colombia and Mexico using machine learning models
->Nature | #MachineLearning | More info from EcoSearch

So excited to see my first solo-authored paper in English officially accepted!

Stay tuned for more research on crime, institutions, and quantitative methods

While these analyses are predictive and not intended to establish causal relationships, I also employ feature importance analysis to reveal the main drivers of police and military violence in both countries

The study combines Lasso, Random Forests, XGBoost, Neural Networks, and a Super Learner ensemble to anticipate misconduct by the military and police at the municipal level. #MilitaryMisconduct

Using detailed subnational data from Colombia and Mexico, I built several machine learning models to predict violence by armed forces, achieving over 92% accuracy 📈 #PoliceViolence

I'm excited to share that my article "Predicting Police and Military Violence: Evidence from Colombia and Mexico using machine learning models" has been officially accepted for publication at @nature.com HSScomms! 📚

See a draft version here: papers.ssrn.com/sol3/papers....

In Latin America, police and military abuses remain widespread. Thousands of civilians have been victims of extrajudicial executions and human rights violations in Colombia and Mexico alone

What if we could predict where violence is likely to happen?

Versión preliminar aquí: papers.ssrn.com/sol3/papers....

Más cosas por venir en temas de crimen, política y otros métodos cuanti 🎯 #violenciapolicial

Aunque los análisis son predictivos y no buscan establecer relaciones causales, también miro la importancia de variables para identificar los principales factores que explican la violencia en ambos países

Combino diferentes métodos de IA como Lasso, Random Forests, XGBoost, Redes Neuronales y un Super Learner ensemble para anticipar la violencia militar y policial a nivel municipal

Utilizando datos subnacionales de Colombia y México, construí varios modelos de aprendizaje automático para predecir violencia por parte de las fuerzas armadas, logrando una precisión superior al 92% 📈

Mi artículo "Predicting Police and Military Violence: Evidence from Colombia and Mexico using machine learning models" ha sido aceptado para publicación en
@nature.com (Humanities & Social Sciences Communications) 📚 y ahí estudio justamente la predicción y prevención de la violencia policial

En América Latina, la violencia policial y militar continúa siendo un reto de política. Miles de civiles continuan siendo víctimas de violaciones a los DD.HH.

Prevenir la violencia de las Fuerzas es clave para mantener la confianza en las instituciones y el Estado de Derecho

📸France24

SIGLA is excited to invite you to the 2025 Hackathon! It's an exciting event to develop innovative ways to analyze data on legal and political institutions in Latin America. Interested members of the Georgetown community can register for this event using the QR code on the flyer!

“Lo obvio y lo correcto.” Mi columna en @ElPaisAmericaCo: a pesar de la violencia en el Catatumbo, volver a la erradicación forzosa de la hoja de coca sería un error. Mejor que el Estado ejerza su función reguladora del mercado.

elpais.com/america-colo...

SIGLA – Your gateway to governance in Latin America If you have ideas for the SIGLA website or suggestions about the data SIGLA provides, we would love to hear from you!

Excited to join Bluesky! We’re SIGLA (States & Institutions of Governance in Latin America), a multilingual database on LatAm political and legal institutions. Based at Georgetown University, we aim to advance research & government transparency. Explore www.sigladata.org and follow us for updates!