Organized by Diletta Huyskes, Maria Sapignoli, Giuseppe Primiero
After the successful first edition of our Seminar Series Towards a Decolonized Artificial Intelligence, this year’s series starts from a simple but radical premise: AI that dominates today’s technological landscape is the product of a specific cultural and epistemic lineage – one that privileges quantification, optimization, efficiency, classification, and the ideal of objective decision-making. But these logics are not the only possible foundations for artificial intelligence. Across the world, and across history, many other ways of conceptualizing intelligence, information, inference, and relationality have existed. Many have been ignored, marginalized, or overwritten by dominant narratives of technology. This seminar series explores how these alternative epistemologies and overlooked scientific traditions – such as indigenous knowledge systems, non-western cosmologies, feminist and relational theories and computational models that complement and overcome the limitations of statistics – might inspire different ways of constructing, using, and imagining AI.
To receive the Teams links needed to follow the seminars online, it is necessary to register for each individual meeting on Eventbrite: https://www.eventbrite.com/cc/other-cultures-of-ai-online-seminar-series-4809393?just_published=true
PROGRAMME
Scale: The Sociomaterial Shaping of AI
Donald MacKenzie | University of Edinburgh
February 3 | 2:30 pm CET
This keynote by MacKenzie, author of foundational works in the sociology of technology, will explore artificial intelligence (AI) as a set of material practices, belief systems and ‘imaginaries’. The focus will be the question of scale. How and why has today’s AI become a giant sociomaterial enterprise? Why are datacentres on the scale of Manhattan being built and trillions of dollars invested ($7.8 trillion from 2025 to 2030, according to a Citigroup estimate), involving carbon-dioxide emissions measured in hundreds of megatonnes?
The presentation will begin by briefly describing the place in the history of AI of scaling (including its crucial material underpinning: graphics processing chips) and of large language models, such as the models underpinning ChatGPT. It will then examine the scaling of large language models as material practice and belief system, including its relations to imaginaries such as ‘artificial general intelligence’. The talk will end by discussing AI’s influential mathematical ‘scaling laws’, their prominent legitimatory role, but also a less-often-discussed implication: diminishing returns. How far out do you go on a diminishing-returns curve, at what financial and environmental cost, and why?
Science and Knowledge in Late Imperial China
Catherine Jami | French National Centre for Scientific Research (CNRS)
February 13 | 9:30 am CET
The history of science has long been influenced by comparative models that implicitly take European scientific traditions as their point of reference. From this perspective, questions about science in other historical and cultural contexts often treat Europe in a universalistic and “standard” way, leading to narratives of absence, failure or delay where methods, theories or results differ.Drawing on four decades of research on mathematics and science in late imperial China, Jami discusses the importance of research questions – and their reformulation – in enabling historians to better understand what kinds of science existed in China and how they developed according to their own trajectories. Through examples from her work, she will show how this approach helps to identify “other cultures of science” developed in the late imperial period and contributes to a more pluralistic understanding of the history of science.
Imagining AI
Kanta Dihal | Imperial College London
March 17 | 2:30 pm CEST
How societies imagine intelligent machines profoundly shapes how such technologies are designed, governed, and integrated into everyday life. This keynote explores the cultural narratives through which AI has been imagined across different historical and geographical contexts, and how these imaginaries influence contemporary debates on ethics, bias, and responsibility in AI.
Drawing on cross-cultural research and the volume Imagining AI: How the World Sees Intelligent Machines which she co-edited, Dihal examines how dominant Western narratives of AI as a threat or dystopian force have spread globally, often overshadowing or displacing local traditions of imagining intelligent machines. She contrasts these narratives with alternative imaginaries where for example intelligent machines have more often been portrayed as cooperative or benign. By making visible the power of narrative in shaping technological futures, this keynote argues that imagining AI differently is a crucial step toward developing more just and culturally grounded approaches to artificial intelligence.
Feminist Data and Epistemologies
Catherine D’Ignazio | MIT
April 14 | 5:30 pm CEST
Contemporary AI systems are largely built upon epistemic traditions that prioritize quantification, classification, prediction, and the ideal of neutral, objective decision-making. Feminist perspectives in design and data science challenge these assumptions by insisting that data, models, and metrics are always situated, political, and consequential. From this standpoint, AI is not merely a technical artifact, but a socio-technical system that actively shapes what – and who – can be known, valued, and acted upon.
D’Ignazio draws from her book Counting Feminicide: Data Feminism in Action (MIT Press, 2024) and on the participatory research and design project Data Against Feminicide, foregrounding lived experience, collective memory, and accountability as central design requirements. The talk examines how these practices offer concrete alternatives to dominant AI logics by redefining core operations such as counting, categorization, data collection, and evaluation.
Green AI and Sustainable Futures
Verónica Bolón Canedo | Universidade da Coruña
May 22 | 9:30 am CEST
The environmental costs of contemporary artificial intelligence are often framed as an unavoidable consequence of advances in machine learning, particularly in data-intensive and large-scale models. This keynote challenges that assumption by drawing on recent research in Green AI, which shows that alternative technical approaches already exist but remain marginal in mainstream AI development. Building on a systematic analysis of methods aimed at reducing energy consumption, computational resources, and environmental impact, the talk presents Green AI not merely as an optimization strategy, but as a different set of design priorities for AI systems. By foregrounding efficiency, task-specific modeling, and sustainability-aware evaluation practices, the keynote demonstrates that high-performing AI does not necessarily require environmentally extractive infrastructures. In doing so, it contributes to a broader reflection on how rethinking technical choices can foster a more eco-conscious and energy-efficient future for AI systems.
Distributed and Relational Indigenous Intelligences
Jason Edward Lewis | Concordia University
TBC