Mateo Dulce Rubio
¡Hola! I am a Ph.D. candidate in Statistics and Public Policy at Carnegie Mellon University, advised by Edward H. Kennedy. My research develops flexible and robust statistical methods for humanitarian and policy applications, using tools from nonparametric statistics, causal inference, mathematical optimization, and responsible machine learning.
My work includes creating an AI-informed tool for estimating landmine contamination risk, used in Colombia and Afghanistan to identify priority areas for mine clearance in collaboration with the UN Mine Action Service and UNOPS. I also work on developing doubly robust estimators for causal inference and population size estimation under capture heterogeneity and recapture dependence, with applications in conservation, public health, and human rights. Finally, I investigate limitations of current algorithmic decision-making from observational data and devise new methods to address issues like selection biases, missing data, privacy, and algorithmic discrimination.
I am a K&L Gates Presidential Fellow in Ethics and Computational Technologies. I have worked at RAND Corporation and Apple on natural language processing and foundation models. Before that, I co-founded the Centro de Analítica para Políticas Públicas and worked at Quantil mostly on crime modeling research in Bogotá. My background is in Mathematics (BSc) and Economics (BSc, MSc cum laude) from Universidad de los Andes in Colombia.
* Dulce is my first (paternal) last name and Rubio is my second (maternal) last name.
★ I am on the academic job market for tenure track postitions and postdocs!
Recent News
October 2024 | My co-author Santiago Cortes-Gómez will be presenting our work "Statistical Inference Under Constrained Selection Bias" at EAAMO'24! |
October 2024 | Invited to give a talk at the INFORMS Doing Good with Good OR competition on "Identification of Hazard Clusters for Priority Landmine Clearance as a Quadratic Knapsack Problem". |
August 2024 | I'm organizing a topic-contributed session at JSM on "Recent Advances in Capture-Recapture Methods for Population Size Estimation" (8/5, 10:30 am). |
July 2024 | New preprint out on "Population Size Estimation with Many Lists and Heterogeneity: A Conditional Log-Linear Model Among the Unobserved". Feedback is welcome! |
June 2024 | I was named a finalist for the INFORMS 2024 Doing Good with Good OR competition for my work on "Identification of Hazard Clusters for Priority Landmine Clearance as a Quadratic Knapsack Problem". |
May 2024 | Our work on "Statistical Inference Under Constrained Selection Bias" was accepted at ICML'24! |
April 2024 | I will spend next summer as a machine learning intern at Apple (Seattle), working on evaluation of large language models. |
January 2024 | Won the ASA Student Paper Competition Award in the Social Statistics Section for my work on nonparametric capture-recapture methods and will be presenting at JSM 2024 . |
January 2024 | Our paper "RELand: Risk Estimation of Landmines via Interpretable Invariant Risk Minimization" was accepted for publication at the ACM Journal on Computing and Sustainable Societies and will be presented at COMPASS in July! |
October 2023 | Awarded the K&L Gates Presidential Fellowship in Ethics and Computational Technologies. |
Selected Research Projects
ML system to support humanitarian demining, currently being tested in Colombia.
ACM JCSS
2024
AI for Social Impact Book. Edited by Milind Tambe, Fei Fang & Bryan Wilder. https://ai4sibook.org/
AI4SI Book
2022