We need to talk about data work for machine learning

On Wednesday 14 June 2023, the 163° Nexa Wednesday will be held, entitled "We need to talk about data work for machine learning", with guest speaker Milagros Miceli, from the Weizenbaum Institute for the Networked Society (The DAIR Institute)

Abstract

Data quality plays a pivotal role in the performance of machine learning (ML) models. Over the past decade, considerable research and industry efforts have focused on addressing biases and minimizing personal subjectivities in data collection, curation, classification, and labeling by data workers. In this talk, will propose a shift of perspective to emphasize the importance of labor in data production and explore power imbalances inherent in data work that significantly shape datasets and systems. The enhancing labor conditions in data work and leveraging data workers’ expertise can improve data quality and help develop ML systems that are more inclusive and just. Starting from the assumption that power imbalances are the problem, not just bias, leads to fundamentally different research questions and methods of inquiry. In this sense, will highlight the need for interdisciplinary dialogue and cooperation in the study of data quality and data work.

Event organised by the Nexa Center for Internet & Society of the Politecnico (Department of Control and Computer Engineering - DAUIN)