Data Science as Political Action: Grounding Data Science in a Politics of Justice

AI for Not Bad

“…He thoroughly argues that data science should be seen of as political and, responding to the frequent practitioner argument that “We should not let the perfect be the enemy of the good,” states, “data science lacks any theories or discourse regarding what “perfect” and “good” actually entail” (Green, 2018 , p. 19). The pro-technology argument takes “for granted that technology-centric incremental reform is an appropriate strategy for social progress” (Green, 2018 , p. 19) without having to worry about how (or whether) this actually occurs. This belief that the introduction of a technology is sufficient to yield a positive end is often, like by Dalton and Thatcher ( 2014 ), called technological determinism.…”

“…For example, Green questions the focus on crime prediction systems at USC's AI for Social Good initiative. He argues that the initiative bolsters racist and oppressive policing instead of working to address the structural problems which lead to police action (Green, 2018 ). Similarly, Palantir, a big data company that produces crime prediction systems for clients like the U.S. government, recently partnered with the United Nation's World Food Program (WFP) (World Food Program, 2019 ).…”

“…First, I offer a functional definition of “good” for an AI system using the capability approach and social justice. Green ( 2018 , p. 4) cites (Collins, 2002 ) in defining a social justice project as “an organized, long-term effort to eliminate oppression and empower individuals and groups within a just society” and advocates for such projects in data science. Such a project can work in complement with the capability approach, a theoretical framework predicated on context-dependent individual freedom and well-being as defined by people's capabilities or real opportunities to act.…”

“…In this paper, I offer a critical perspective on the use of language of AI practitioners like myself who, from practice to theory, apply their work to some definition of “good.” I use discursive analysis to explore the space between the notion of such projects and their actuality. In so doing, I follow Green ( 2018 ) in identifying AI systems as inherently political. Vague terms are the wagons of a modern gold rush into the promised riches of a mythic AI frontier.…”

Recovering critique in an age of datafication

“…In what follows, I will use as exemplar of the problematic 'gap' in critical social thought, not affect theory, but Latour's influence on social science through ANT and STS. Other influences on contemporary social analysis could perhaps have been chosen, but I choose ANT/STS because it is precisely that tradition which even scholars looking to critique contemporary data practices (Green, 2019) often see as the required toolkit for the critical data scientist. There are good reasons, however, for believing that ANT/STS, given its general philosophical biases, cannot provide, or sustain, a sufficient toolkit for addressing critically what is going on with data in society today.…”