Our bi-monthly DiPLab online seminar will welcome Maxime Cornet and Clément Le Ludec (both PhD Candidates at Telecom Paris) for a session on “The globalisation of service jobs: what’s the place of data-labour in AI’s global value chain network? Findings from two Franco-Malagasy case studies.“.

The seminar will be held in hybrid form, both at Telecom Paris and online on the platform Big Blue Button. To receive the link, please register by sending an email with your full name and affiliation at: contact@diplab.eu.

Artificial intelligence is presented as a technology of structural change for society and especially for work. Beyond the fears related to the end of work (Frey and Osborne, 2017), these techniques are also used in systems that transform the recruitment or control of workers (i.e algorithmic management). At the intersection of these two issues, the literature on micro-work has studied both the platformization of work and the use of human data work to produce AI systems (Gray and Suri, 2018).

A few studies on micro-work analyze the position and contribution of all actors in the chain. In our presentation, we analyze the whole French AI production process investigating what is the place of data labour in this value chain? Who does this activity? And above all, how is it organized? Our work rely on two case studies focused on two AI systems which we drew up on the basis of 26 interviews with data workers, 8 interviews with French start-up employees and secondary data sources. One case study describes the production of an automated canteen terminal, while the others concerns a video-surveillance algorithm sold to supermarkets to automatically detect and report theft.

By reconstructing product-wide production process, accounting for the numerous actors involved in the chain, we found that what’s ultimately sold and deployed as an ‘AI product’ is the production process itself rather than an embedded ready-to-use algorithm.
This include the captation of end-users’ unremunerated labour, as well as the transformation of workers role in France to include AI-related tasks.

In this sense, we argue that AI can be seen as a “last-mile automation” socio-technical device. A logistic tool used to facilitate and articulate various labour process that ultimately produce a service. The characteristics of commercial AI is therefore not to automate service jobs but to displace labour, lengthening externalization chains. Through its production chain, AI both reproduce itself and enable labour displacement. Ultimately, involving AI in the services’ production processes is not about automation, but allow the outsourcing of labour toward both ends of the value chains.

This study shows the scale of data production processes by shedding light on the distribution of data labour in the value chain. It also questions how to regulate AI biases, notably in the light of EU legislation obligation for companies designing certain AI systems to make public the steps in the production process of these devices.

Maxime Cornet is a PhD student at Telecom Paris – Polytechnic Institute of Paris, working on HUSH. He is particularly interested in social network analysis, sociology of organization, and economic sociology. He’s working on the impact of labor organization on the production of AI.

Clément Le Ludec is a PhD researcher at Telecom Paris – Polytechnic Institute of Paris. He works on the HUSH project, exploring data supply chains in French-speaking African countries.