Who is thinking about AI through the social sciences in Spain

When people talk about Spanish leadership in artificial intelligence, the conversation tends to drift towards models, supercomputers, semiconductors or companies. The map changes if the question changes. Where is AI being used to study society? And where are the social sciences and the humanities intervening in how artificial intelligence is designed, governed and interpreted?

These are two different directions. In the first, AI works as an instrument for reading social processes at a previously impossible scale. In the second, the disciplines that study society claim a place in defining the technology itself.

There is no single leading university. There are, rather, several poles with different specialisms. Barcelona brings together supercomputing, computational social science and social simulation at European scale. Granada has turned computational method into a curriculum. Madrid combines training in computational social science, digital infrastructure for the humanities and a heritage laboratory at the National Library. Anthropology appears in the lists of disciplines and in the curricula, though it mostly comes in through the door of method.

We do not aim to produce an exhaustive census, but to follow some of the places where the relationship between AI and the social sciences becomes especially visible. Going through them allows a sharper question than that of leadership. Which of the two directions carries more weight in what is being built today?

Barcelona, when the social sciences reach the supercomputer

The Barcelona Supercomputing Center (BSC-CNS) is still associated with physics, climate, biomedicine or engineering. In November 2025, when the centre turned twenty, Spain’s Recovery Plan described it as an institution of 1,400 professionals of 60 nationalities whose activity, organised around the computer, earth and life sciences and engineering, has expanded to Computational Social Sciences and Humanities.1

That expansion has a date and a name. In May 2023, the BSC announced that Mercè Crosas was joining to lead a new computational social sciences programme, which it presented as an exceptional initiative within a European supercomputing centre. Crosas had led research data management at Harvard University and the office of data science and technology at its Institute for Quantitative Social Science, and from 2021 to 2023 she served as Secretary for Open Government of the Generalitat de Catalunya. The announcement placed the programme in dialogue with economics, political science, sociology, anthropology, geography, psychology, cognitive science and the digital humanities.2

Today the programme already includes the humanities in its name. From it, Crosas leads initiatives that apply supercomputing, artificial intelligence and the analysis of large datasets to the study of human societies. She was also the first confirmed speaker at the International Congress on Archives, ICA Barcelona 2025, a detail that brings the programme closer to archives and heritage as well.3

Here scale matters. The BSC can bring the social sciences together with supercomputing, large datasets and AI models in a way that is hard to reproduce in a conventional faculty. The European project WHAT-IF shows where that combination is heading.

Funded by Horizon Europe with €3.2 million between January 2025 and December 2027, under a call devoted to computational social science approaches in research on democracy, WHAT-IF is coordinated by Vrije Universiteit Amsterdam and brings together fifteen institutions. The three Spanish ones are in Barcelona. They are the University of Barcelona, the BSC itself and the Universitat Oberta de Catalunya.4

Its aim is to build an empirically grounded simulation that functions as a digital twin of the political information environment. To do so it combines data donation, agent-based modelling and large language models, in order to test how different policy and regulatory interventions would affect democratic citizenship and to provide evidence for improving media regulation.45

The project record explains why a twin is needed. Platform data are proprietary, existing studies have limited external validity and societies cannot be used as experimental playgrounds.4 The solution is to ask people to donate their own data and to rebuild with them, and with artificial agents, an environment in which to run tests.

From an anthropological standpoint, this design raises questions that are not technical. Who donates their data and who does not. What form of citizenship is represented when the environment is rebuilt from those who agree to hand over their digital trace. What changes when whoever reads, shares or argues inside the simulation is a language model. The digital twin, moreover, is no longer a laboratory curiosity. In September we found it in the Madrid regional government’s announcements, applied to health and regional management.

WHAT-IF also illustrates the second direction of the map. It does not merely use AI to study society. It seeks to produce knowledge for governing the information environment in which AI itself already circulates. That is why the BSC is probably one of the Spanish centres we would do well to follow closely.

Granada, where method is already a curriculum

The surprise on the map is Granada.

The Unit for Computational Humanities and Social Sciences (U-CHASS) at the University of Granada presents itself as a unit dedicated to transforming the social sciences and humanities through advanced computational and artificial intelligence methodologies. It brings together more than twenty faculty members from linguistics, literature, history, library and information studies, computer science and media studies, among other disciplines. It organises its work in three areas. The first addresses disinformation and social polarisation; the second, digital humanities and cultural metrics; the third, the science of science.6

Its scientific director is Nicolás Robinson-García, Senior Lecturer in the Department of Information and Communication Sciences at the UGR and head of the EC3 group, which aims to understand and improve the processes by which science is evaluated.7 Under the name C-CHASS, the unit is also listed on the university’s research portal as a unit of excellence for 2024–2027.8

What sets it apart is that this agenda has already entered formal teaching. Since the 2024–25 academic year, the UGR and the University of Salamanca have offered the Master’s Degree in Data Science Applied to the Social Sciences (DaSoc), coordinated by Rosa Soriano of the UGR and Modesto Escobar of the University of Salamanca.9 It carries 60 credits and has a hybrid format, with the foundation subjects taught in person during the first semester and the rest delivered virtually and synchronously.10

The curriculum published in Spain’s Official State Gazette includes among its six compulsory subjects Epistemología y ciencias sociales computacionales (Epistemology and computational social science), alongside foundations of programming, statistical learning and machine learning applied to social analysis, text and social network analysis, and the collection, storage and pre-processing of digital social data. The ten electives include Etnografía digital (Digital ethnography), Investigación social con datos masivos y densos (Social research with big and thick data), Ciencia de datos e intervención en bienestar social (Data science and social welfare intervention) and an applied workshop on modelling and simulating complex social systems.11

The master’s presentation stresses training people able to formulate the right questions and to apply the relevant techniques for understanding social phenomena shaped by data.10 The name of one of its electives also points to a tension that anthropology knows well. In the anthropological tradition, the word densos, ‘thick’, points to Clifford Geertz’s thick description, that is, to the distance between recording many behaviours and understanding what they mean to those who live them.

Madrid, between computational training and heritage

In Madrid, Universidad Carlos III teaches the Master’s Degree in Computational Social Science in English, a 60-credit, in-person programme at the Puerta de Toledo campus. The Instituto Juan Linz, the UC3M-Santander Big Data Institute and the departments of Social Sciences, Statistics and Computer Science take part. Its aim is to combine training in the social sciences with predictive models, text analysis and network science, and its subjects include text mining, social network analysis, causal inference and survey methodology.12

Digital infrastructure for the humanities has another centre of gravity. Since October 2023, Universidad Complutense de Madrid has hosted the CLARIAH-CM node, funded by the Community of Madrid with €750,000 over five years and designed to support the digital transformation of research in the humanities and social sciences at six public universities in Madrid (Alcalá, Autónoma, Carlos III, Politécnica, Rey Juan Carlos and the Complutense itself). The node is part of the national CLARIAH-ES platform, which brings together 35 universities and research centres and more than 150 groups, among them the BSC, UNED, the CSIC and the National Library of Spain.13

That detail connects the poles. The BSC appears on the Barcelona map as a supercomputer and on the Madrid map as part of the network that sustains digital infrastructure for the humanities.

The National Library of Spain (BNE) adds the heritage dimension. Its laboratory, BNElab, has set out to analyse the possibilities of applying AI to generate new knowledge from the BNE’s collections and digital resources. The lines it lists range from querying databases and text corpora in natural language to improving OCR and manuscript transcription, extracting entities and image elements, recognising musical scores, assistance chatbots and automated cataloguing and classification, with a stated commitment to ethical, sustainable and transparent use.14

Cataloguing and classifying are old operations, and they are not neutral. On Friday, drawing on Aparecida Vilaça, we saw that a stable category can transform what it seeks to recognise. When cataloguing is automated, part of the decision about what counts as what shifts to the model doing the classifying.

Anthropology comes in through method

As far as we have been able to verify, anthropology does not yet appear on this map as an institutional pole comparable to the BSC or U-CHASS. It appears through method.

In the Granada master’s, Etnografía digital is one of ten electives, of which students choose five, and accounts for 3 of the 60 credits.1011 Disparidades. Revista de Antropología, published by the CSIC, devoted a special section to digital ethnographies in its July–December 2025 issue. The introduction, signed by Elisa García-Mingo, Héctor Puente-Bienvenido, Aida Gallego-Márquez, Silvia Díaz-Fernández and Jacinto Lorca, proposes researching violence, affect and embodiment in digital society. It encourages those who study the sociotechnical to adopt nuanced, diverse and dynamic methods such as digital ethnography, and invites ethnographers to explore neglected digital spaces and obsolete technologies too.15

From Ibero-American anthropology, the yearbook ARIES, published by the Madrid-based Asociación de Antropólogos Iberoamericanos en Red (AIBR), carried in its 2024 volume a reflection on ethnography and artificial intelligence by Jorge Mantilla of the Universidad de Otavalo (Ecuador). Mantilla argues that ethnographic studies make it possible to analyse the sociocultural configurations underlying the design and implementation of AI systems, and that they offer a qualitative understanding often missing from a literature dominated by quantitative approaches. He also examines how AI may transform classical ethnographic methods.16

The three pieces share a shift. AI does not appear only as a tool for doing ethnography at greater scale. It also appears as an object that can be studied ethnographically, together with the people who design it, use it or live with its effects. That movement, studying AI as a social fact, is the one that carries least weight on today’s institutional map.

From Mexico, the question turned into a working group

The question in this radar’s title has an institutional form of its own in Mexico. In May 2025, the Mexican Council of Social Sciences (COMECSO) opened the call for the working group Horizontes de investigación sobre la Inteligencia Artificial desde las Ciencias Sociales (Research horizons on artificial intelligence from the social sciences), with the purpose of gathering theoretical reflections and empirical research from all social science disciplines and organising them within a framework that shows the field’s landscape, concerns and relevant themes.17

It is coordinated by Juan Pablo Duque Parra of UNAM’s Centre for Interdisciplinary Research in Sciences and Humanities (CEIICH), and by Khemvirg Puente Martínez and Jacqueline Torres Ruiz of the Faculty of Political and Social Sciences at the same university. This year the group convened the virtual Jornadas nacionales de inteligencia artificial y ciencias sociales (National days on artificial intelligence and the social sciences) from 28 September to 1 October, with two keynote lectures announced for 5 and 9 October within the National Social Sciences Week. Ricardo Arechavala Vargas of the University of Guadalajara and Danay Quintana Nedelcu of FLACSO-Mexico take part as invited coordinators.18

The call seeks to consolidate a space for dialogue and exchange among specialists from all social science disciplines on AI, sociodigital networks and big data in Mexico, Latin America and Ibero-America.18 The first of those keynote lectures falls on the day we publish this radar.

The comparison is not meant to set two models against each other. In Spain, the map is drawn mainly from infrastructures, units and degrees. In Mexico, part of the social sciences has made its own questions about AI an object of work. And it has done so looking towards Ibero-America, which opens that conversation to the Spanish poles as well.

Two directions with different weights

This brings back the double question from the start. In the cases examined here, the first direction, AI as an instrument for studying society, concentrates the supercomputing, the European funding and the credits in the curricula. The second, the social sciences intervening in how AI is designed, governed and interpreted, appears in the objectives of some projects, in elective subjects, in special sections and in working groups.

The anthropologist Néstor García Canclini, Distinguished Professor in the Department of Anthropology at the Universidad Autónoma Metropolitana, Iztapalapa campus (Mexico), who holds a doctorate in philosophy from the universities of Paris and La Plata, offers a key for reading that asymmetry. In Ciudadanos reemplazados por algoritmos (Citizens replaced by algorithms, 2020) he asks what alternatives remain for citizens when, as the book’s presentation puts it, ‘our opinions and behaviour, captured by algorithms, end up subordinated to globalised corporations’.19

WHAT-IF allows us to observe that dispossession from within research. If publicly funded social science needs people to donate their data, it is because platform data are proprietary.4 To see what a platform sees by default, research has to ask each person for part of their own trace.

The researcher Paola Ricaurte adds an epistemological layer. In an article published in 2019 in Television & New Media, written from the Tecnológico de Monterrey (Mexico) and Harvard University, she argues that data-centric epistemologies should be understood as an expression of the coloniality of power, an imposition of ways of being, thinking and feeling that denies the existence of alternative worlds and epistemologies. In response, she suggests possible alternative data epistemologies that are respectful of populations, cultural diversity and environments.20

Read this way, the Spanish map stops being only a question of leadership. It becomes a question about what idea of society is encoded when society enters the supercomputer, and about who takes part in defining it.

Ethnography does not resolve that question. But it remains the method that asks people what the data they produce mean.

Spain already has several centres for studying society with AI. What we have not found, at least with comparable institutional stability, is an equivalent place devoted to studying AI itself as a social fact.

References

[1] Plan de Recuperación, Transformación y Resiliencia. ‘El Barcelona Supercomputing Center celebra 20 años como el gran cerebro digital de España y referente europeo en innovación’. 12 November 2025. In Spanish.
https://planderecuperacion.gob.es/noticias/barcelona-supercomputing-center-celebra-20-anos-gran-cerebro-digital-espana-prtr

[2] HPCwire. ‘Mercè Crosas Spearheads New Computational Social Sciences Program at BSC’. 31 May 2023.
https://www.hpcwire.com/off-the-wire/merce-crosas-spearheads-new-computational-social-sciences-program-at-bsc/

[3] ICA Barcelona 2025. ‘Mercè Crosas, pionera en supercomputación y ciencia de datos, primera ponente confirmada del ICA Barcelona 2025’. In Spanish.
https://icabarcelona2025.cat/es/merce-crosas-pionera-en-supercomputacion-y-ciencia-de-datos-primera-ponente-confirmada-del-ica-barcelona-2025/

[4] CORDIS, European Commission. ‘Advanced Simulations for Testing the Effect of the Information Environment on the Functioning of Democracy (WHAT-IF)’. Project 101177574, call HORIZON-CL2-2024-DEMOCRACY-01-06.
https://cordis.europa.eu/project/id/101177574

[5] WHAT-IF. ‘About’.
https://what-if-horizon.eu/about/

[6] U-CHASS, University of Granada. ‘About us’.
https://uchass.ugr.es/about-us/

[7] U-CHASS, University of Granada. ‘Nicolás Robinson-García’.
https://uchass.ugr.es/miembro/nicolas-robinson-garcia/

[8] University of Granada, Research Output Portal. ‘Unidad de excelencia: Centre for Computational Humanities and Social Sciences C-CHASS’. In Spanish.
https://produccioncientifica.ugr.es/proyectos/796458/detalle

[9] Faculty of Political Science and Sociology, University of Granada. ‘Inauguración del Máster Universitario en Ciencia de Datos Aplicada a las Ciencias Sociales’. In Spanish.
https://polisocio.ugr.es/facultad/noticias/inauguracion-del-master-universitario-ciencia-datos

[10] University of Granada. ‘Máster Universitario en Ciencia de Datos aplicada a las Ciencias Sociales (DaSoc). Presentación’. In Spanish.
https://masteres.ugr.es/dasoc/informacion/presentacion

[11] Boletín Oficial del Estado. ‘Resolución de 11 de diciembre de 2024, de la Universidad de Granada, por la que se publica el plan de estudios de Máster Universitario en Ciencia de Datos Aplicada a las Ciencias Sociales’. BOE-A-2024-27283. In Spanish.
https://www.boe.es/diario_boe/txt.php?id=BOE-A-2024-27283

[12] Universidad Carlos III de Madrid. ‘Máster Universitario en Ciencias Sociales Computacionales’. In Spanish.
https://www.uc3m.es/master/ciencias-sociales-computacionales

[13] Universidad Complutense de Madrid. ‘La Complutense albergará el nodo CLARIAH-Comunidad de Madrid (CLARIAH-CM) para la transformación digital de la investigación en Humanidades y Ciencias Sociales’. 3 October 2023. In Spanish.
https://www.ucm.es/la-complutense-albergara-el-nodo-clariah

[14] National Library of Spain, BNElab. ‘Artificial intelligence’.
https://bnelab.bne.es/en/project/artificial-intelligence/

[15] García-Mingo, Elisa; Puente-Bienvenido, Héctor; Gallego-Márquez, Aida; Díaz-Fernández, Silvia; Lorca, Jacinto. ‘Etnografía(s) digital(es): investigar violencias, afectos y corporalidades en la sociedad digital’. Disparidades. Revista de Antropología, 80(2), e1109, 2025. In Spanish.
https://doi.org/10.3989/dra.2025.1109

[16] Mantilla, Jorge. ‘Etnografía e inteligencia artificial: potencialidades, retos metodológicos y desafíos futuros’. ARIES. Anuario de Antropología Iberoamericana, 19, 2024. In Spanish.
https://doi.org/10.11156/aries/2024.AR0002408

[17] Consejo Mexicano de Ciencias Sociales (COMECSO). ‘Horizontes de investigación sobre la Inteligencia Artificial desde las Ciencias Sociales’. Call for participation, 28 May 2025. In Spanish.
https://www.comecso.com/convocatorias/horizontes-de-investigacion-sobre-la-inteligencia-artificial-desde-las-ciencias-sociales

[18] Consejo Mexicano de Ciencias Sociales (COMECSO). ‘Jornadas nacionales de inteligencia artificial y ciencias sociales’. 2026. In Spanish.
https://www.comecso.com/grupos-de-trabajo/inteligencia-artificial/jornadas-nacionales

[19] García Canclini, Néstor. Ciudadanos reemplazados por algoritmos. Bielefeld University Press, 2020. Open access. In Spanish.
https://doi.org/10.14361/9783839448915

[20] Ricaurte, Paola. ‘Data Epistemologies, The Coloniality of Power, and Resistance’. Television & New Media, 20(4), 350–365, 2019.
https://doi.org/10.1177/1527476419831640