Paper short abstract:

STS scholars and critics of big data know that data-intensive and algorithmically mediated systems can reify and reproduce inequities; sometimes, so do designers and analysts. I present two case studies of "data science for social good" teams confronting bias in their urban mobility projects.

Paper long abstract:

Critiques of big data commonly highlight the potential for data-intensive decision-making and algorithmic mediation to reify and reproduce inequities. Interrogations of systems ranging from automatic ad placement to algorithmic recidivism prediction show that biased data collected about a biased world inevitably produce biased applications and models -- a high-stakes instantiation of the computer science aphorism, "garbage in, garbage out." Such critiques of data-intensive and algorithmic systems have most often highlighted the outputs of those systems rather than focusing on the practices and sense-making that inform their creation. Based on immersive participant observation of collaborations formed under the banner of "data science for social good," I present two cases of big data urbanism-in-the-making. The first involves efforts to make transactional data generated by a public transit payment system useful for the purposes of transportation planning. The second involves efforts to create a digital routing application designed specifically for the informational needs of people with mobility impairments. In both projects, analysts and designers acknowledge the biases baked into digital data infrastructures and place those biases at the center of their work, either by trying to mathematically correct for them, or by trying to subvert the very systems that produced them. I explore how each of the teams' implicit understandings of sociomateriality shaped ongoing deliberation about the ethical implications of their work, and how the decisions they made in designing and implementing their projects have material consequences for moving about and accessing the city.