Causal Parametric Drift Simulation: A Digital Twin Framework for Classifier Robustness Evaluation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lafrance, Julien, Khoury, Richard, Tremblay, Véronique
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914551567482880
author Lafrance, Julien
Khoury, Richard
Tremblay, Véronique
author_facet Lafrance, Julien
Khoury, Richard
Tremblay, Véronique
contents Machine learning classifiers in dynamic environments face concept drift -- changes in the data-generating process that degrade performance. Conventional evaluation via static test sets or noise perturbations fails to preserve causal dependencies in tabular data, often producing causally invalid assessments. Post-hoc tools like SHAP and LIME offer correlational insights that may not reflect the causal mechanisms driving model failure. We propose a framework that complements existing drift detection by leveraging Structural Causal Models as "Digital Twins" of data-generating processes, enabling precise causal interventions while preserving structural dependencies. Our technique, Causal Parametric Drift Simulation, stress-tests classifiers to identify vulnerabilities before deployment. Experiments on the Open Sourcing Mental Illness (OSMH) dataset demonstrate that this approach exposes latent vulnerabilities invisible to standard statistical monitors.
format Preprint
id arxiv_https___arxiv_org_abs_2605_09663
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Causal Parametric Drift Simulation: A Digital Twin Framework for Classifier Robustness Evaluation
Lafrance, Julien
Khoury, Richard
Tremblay, Véronique
Machine Learning
Artificial Intelligence
62H22, 62D20, 68T05
I.2.6; I.2.0; G.3
Machine learning classifiers in dynamic environments face concept drift -- changes in the data-generating process that degrade performance. Conventional evaluation via static test sets or noise perturbations fails to preserve causal dependencies in tabular data, often producing causally invalid assessments. Post-hoc tools like SHAP and LIME offer correlational insights that may not reflect the causal mechanisms driving model failure. We propose a framework that complements existing drift detection by leveraging Structural Causal Models as "Digital Twins" of data-generating processes, enabling precise causal interventions while preserving structural dependencies. Our technique, Causal Parametric Drift Simulation, stress-tests classifiers to identify vulnerabilities before deployment. Experiments on the Open Sourcing Mental Illness (OSMH) dataset demonstrate that this approach exposes latent vulnerabilities invisible to standard statistical monitors.
title Causal Parametric Drift Simulation: A Digital Twin Framework for Classifier Robustness Evaluation
topic Machine Learning
Artificial Intelligence
62H22, 62D20, 68T05
I.2.6; I.2.0; G.3
url https://arxiv.org/abs/2605.09663