Template-Type: ReDIF-Paper 1.0

Author-Name: Adolfo De Unánue
Author-Name-First: Adolfo
Author-Name-Last: De Unánue
Author-Email: unanue@tec.mx
Author-Workplace-Name: School of Government and Public Transformation, Tecnológico de Monterrey

Author-Name: Fernanda Sobrino
Author-Name-First: Fernanda
Author-Name-Last: Sobrino
Author-Email: fersobrinno@tec.mx
Author-Workplace-Name: School of Government and Public Transformation, Tecnológico de Monterrey

Title: Machine Learning as Performative Materialist Practice: Thirteen Theses on the Epistemology, Methodology, and Politics of Applied ML

Abstract: This work proposes thirteen theses for rethinking machine learning as a situated, performative, and temporal practice. It argues that models do not represent stable systems, but rather intervene in them, coevolving with the data, institutions, and decisions they help produce. From this perspective, their value should be evaluated based on their concrete effects, their multi-objective trade-offs, and their capacity to improve public action under real material, ethical, and organizational constraints.

Length: 7 pages
Creation-Date: 2026-05
Number: 34

File-URL: https://egobiernoytp.tec.mx/sites/default/files/2026-05/machine_learning_performative_materialist_practice.pdf
File-Format: Application/pdf
File-Function: First version, 2026

Classification-JEL: C45, C53, C63, D81, H83, O33
Keywords: Machine learning, automated learning, performative prediction, data products, complex adaptive systems, public policy, model evaluation, multi-objective trade-offs, fairness, algorithmic governance, temporality

Handle: RePEc:gnt:wpaper:34