Data cooperatives & AI for the public good

This project is exploring co-creation and participatory approaches in dataset and model development that better serve the public. At a high level, this project examines how we might build datasets and models that are more democratic, including meaningful public participation. This includes exploring how to reshape the ways datasets and AI technologies are developed and used globally, grounded in values of democratic participation, diversity and empowerment. 

At a more granular level, the project is, first, developing a co-created dataset that utilizes novel methods for community governance in the form of a data cooperative. Second, we will conduct fine-tuning approaches (using the datasets developed) to examine if and how models can perform better for people globally, and more accurately reflect diverse lived experiences and realities. Finally, the dataset will be available and maintained over time through a data cooperative model.

This work is in partnership with Technical University of Munich, Télécom Paris at the Institut Polytechnique de Paris, D-Lab, and the Mozilla Data Collective. This work is made possible with support from the HS Foundation. 
data cooperative

Source: Astralis