Invisible, but Audible? Insights from the Voices of AI Data Workers

CRIS SCIENTIFIC SEMINAR 2024-2025

Talk by Paola Tubaro

Senior Researcher CNRS,  Economic Sociologist at ENSAE-CREST

The production of artificial intelligence (AI) relies on a largely concealed pool of ‘data workers’ who prepare and verify data, often under poor conditions and with little recognition. Various organisational and economic factors invisibilise the contribution of these workers and limit opportunities for improvement. Efforts to raise awareness may be ineffective if they fail to account for the diverse realities of data work in distinct regions of the world. 

I leverage multilingual questionnaires and interviews to highlight some of these differences. In wealthier nations, data work serves as a supplementary income, while elsewhere, it is a primary yet precarious livelihood. Migrants and women face the greatest disadvantages. 

In the last part of the talk, I harness an innovative methodology borrowed from large-corpus phonetics and speech recognition technologies for a finer analysis of socio-economic differences among data workers. Findings suggest that invisibility affects workers unevenly, reinforcing existing inequalities. 

Addressing these challenges requires recognising workforce heterogeneity and devising policies that target the most vulnerable.

To find out more: Data Big and Small website - Registration is mandatory, thanks