Towards Intersectional Artificial Intelligence

Abstract

Introduction. The rise of Artificial Intelligence systems for information processing and the empirical relevance of the intersectional approach encourages the search for a confluence point between both analysis methods. Objectives. The analytical potential offered by machine learning systems, given their high capacity for processing massive data, added to the better disposition of intersectional approaches to address social problems by focusing on the different axes of identity/oppression that vertebrate the position of people, invites us to propose a symbiosis between them until reaching an Intersectional Artificial Intelligence. Methodology. To this end, a conceptual approach to both methods is made, and the three moments in which the intersectionality of artificial intelligence systems could be tested are studied: in the configuration of the training databases, in the discovery of correlations between variables during the development of the models and, finally, in the audit phase as a category of system reliability. Results. After reviewing the doctrinal and empirical assumptions already developed, it is observed how it is possible to place at the service of society an Artificial Intelligence that, far from causing biases, contributes to the visibility of forgotten realities and discriminated groups from an Intersectional perspective. Conclusion. In a democratic society, an Intersectional Artificial Intelligence is not only possible but desirable as a tool to promote diversity and inclusion.