School of Social and Political Science

Description

Controversies in the Data Society 2026 series - Session 7

How do we responsibly domesticate new forms of computing? In this session we explore the options and challenges for building responsible AI. Alex Taylor explores the supply chains of labour and finance that deliver ‘AI’ to our screens, and Lara Dal Molin discusses her findings on how to align AI models to human values.

About this session

Dr Alex Taylor will give a talk called AI is a supply chain: 

Thinking of safer, more responsible AI, we’re often drawn into a technological language - that is, a language where the technology takes centre stage, whether it be an algorithm, model or some hardware system with embedded ‘smarts’ (e.g. a phone or even a robot). In my research, I’m interested in seeing AI not as a technological ‘thing’ - an algorithm, model, etc.- but as a knotted web of actors, entities, events, and systems, and the always emerging connections, flows and frictions between them. For me, the claim ‘AI is a supply chain’ has helped to activate this way of thinking. It has invited greater attention to AI’s dependencies on outsourced labour, supply chains, financial markets, etc. and the economic and political conditional that make AI possible. In this talk, I want to consider what this position means for safer, more responsible AI. I’ll share some ongoing research tracing the outsourcing of labour involved in tuning AI models so that they generate less toxic, unsafe content. I’ll aim to show how the operationalising of this ‘trust and safety’ work invites harsh criticisms of a sector that adopts a technological langue of responsibility to arguably avoid a fuller accountability for the mess we are in. 

Lara Dal Molin describes her talk: 

This session interrogates methods to operationalise responsibility and alignment in large language models. Focusing on an example of participatory fine-tuning, conducted as part of my PhD project, I explore both empirical and ethical possibilities and pitfalls of hybrid qualitative-quantitative methods for capturing human preferences in LLMs.

About the speakers

Dr Alex Taylor is Reader in Design Informatics, and AHRC BRAID Fellow. Alex is a sociologist with a fascination for the relations between machines and social life, and what possibilities technoscientific entanglements might create for fundamental transformations in society. He is a Reader in Design Informatics at the University of Edinburgh, and currently an AHRC BRAID research fellow focusing on the operationalising of responsibility. He is also a fellow of the RSA and holds visiting roles at the University of Sweden and City, University of London.

Lara Dal Molin is a PhD student in Science, Technology and Innovation Studies, also part of a joint programme in Social Data Science with the University of Copenhagen. In her research, Lara develops creative participatory methods to resist algorithmic bias through community building. At present, her PhD project employs co-design workshops, survey-based methods and participatory fine-tuning techniques to explore how large language models reproduce and perpetuate gender bias.

Further reading

• Miceli, Milagros, and Julian Posada. "The data-production dispositif." Proceedings of the ACM on human-computer interaction 6.CSCW2 (2022): 1-37. https://doi.org/10.1145/3555561 

Longer Readings

• Fourcade, Marion, and Kieran Healy. The ordinal society. Harvard University Press-T, 2024. 
• Hao, Karen. Empire of AI: Inside the reckless race for total domination. Random House, 2025. 
• Mezzadra, Sandro, and Brett Neilson. The politics of operations: Excavating contemporary capitalism. Duke University Press, 2019. 
• Tsing, Anna Lowenhaupt. The mushroom at the end of the world: On the possibility of life in capitalist ruins. Princeton University Press, 2015