Research
Mathematical optimization for explainable, data-driven decision making.
Journal Articles
One-for-many counterfactual explanations by column generation Forthcoming
A column-generation framework that produces sparse counterfactual explanations covering many instances with a shared, compact set of changes — scaling group explanations to large populations.
Mathematical optimization modelling for group counterfactual explanations
A mathematical optimization framework for generating counterfactual explanations at the group level. Moving beyond individual explanations, we adopt a stakeholder perspective in which counterfactuals are produced jointly under structural and feasibility constraints, revealing close connections with classical facility location problems. Awarded the SEIO–Fundación BBVA Award 2024.
Counterfactual analysis and target setting in benchmarking
We extend counterfactual reasoning to Data Envelopment Analysis (DEA), introducing an optimization-based framework for interpretable target setting in efficiency analysis. Counterfactual targets are computed via mixed-integer convex optimization, providing transparent and actionable explanations for inefficient units.
Collective counterfactual explanations generated for an entire group of individuals at once. We formulate the problem as a mixed-integer model that minimizes total perturbation cost while guaranteeing a favorable outcome, allowing shared influential features and outliers to be identified simultaneously.
A new model for counterfactual analysis for functional data
A counterfactual analysis framework tailored to functional data, where instances are curves rather than vectors. The approach yields interpretable, optimization-based explanations that respect the functional structure of the inputs.
Preprints & Working Papers
Explaining decisions arising from contextual optimization problems under endogenous (decision-dependent) uncertainty, with an application to competitive facility location.
Multistage counterfactual decisions
Extending counterfactual reasoning to sequential, multistage decision settings.
See also my Google Scholar and ORCID profiles.