AI at Work: An Ethnomethodological Study of Data Science Practices

Saha, Dipanjan ORCID: 0000-0003-2430-5055 (2025) AI at Work: An Ethnomethodological Study of Data Science Practices PhD thesis, University of Liverpool.

Text 201433288_Jan2025.pdf - Author Accepted Manuscript Download (18MB) | Preview

Abstract

The objective of this thesis is to explore the practical settings in which the production of Artificial Intelligence (AI) and Machine Learning (ML) algorithms is an everyday, routine activity. Adopting an ethnomethodological approach, I worked alongside data scientists to examine the mundane activities that constitute their everyday work. While studies of algorithms often focus on general or abstract topics such as “opacity” (Burrell, 2016) or the “cultural features” of AI/ML (Seaver, 2017), recent scholarship has increasingly turned to the practical work of producing algorithms through autoethnographic (Mackenzie, 2017), ethnographic (Brooker et al., 2019; Jaton, 2021), and “DIY” experimentation (Sormani, 2020). Differing from classical studies, ethnomethodological studies (Garfinkel, 2002) aim to provide a description of the “technical aspects” of the work being investigated— a description that is grounded in the practices under study (Livingston, 1986; Greiffenhagen and Sharrock, 2019). Such studies seek to make these practices observable to “outsiders,” i.e., those with limited familiarity with the domain, while preserving the constitutive relevancies of the shop floor practice they describe (Ikeya, 2020). This dual focus requires analysts who are competent in both sociology and the technical domains under study (Ribes, 2019; Ribes et al., 2019). Drawing on my experience in both computer science and sociology, this thesis explicates the “praxiological foundations” (Lynch, 1991) of AI and ML algorithms by engaging in the hands-on work of producing such algorithms in two practical settings: a commercial AI company in Manchester and an academic AI research laboratory in Liverpool. Rather than treating AI and data science as abstract domains, this thesis takes up data scientists’ work as situated, methodic, and inherently tied to the organisational and technical contexts of the shop floor. By examining, through the lens of tutorial activities undertaken by myself which also formulate part of the professional practice of data science, how data scientists develop “intelligent” systems and make their understandings observable in that process, this study offers an empirically grounded account of their disciplinary practices and contributes to social studies of AI in the process more broadly.

Item Type:	Thesis (PhD)
Uncontrolled Keywords:	Algorithms, Artificial Intelligence, Data Science, Ethnomethodology, Machine Learning, Social Studies of AI, STS
Divisions:	Faculty of Humanities & Social Sciences Faculty of Humanities & Social Sciences > School of Law and Social Justice
Depositing User:	Symplectic Admin
Date Deposited:	01 Sep 2025 15:59
Last Modified:	01 Sep 2025 16:00
DOI:	10.17638/03193813
Supervisors:	Mair, Michael Brooker, Phillip
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3193813
Disclaimer:	The University of Liverpool is not responsible for content contained on other websites from links within repository metadata. Please contact us if you notice anything that appears incorrect or inappropriate.