Akin to the “organic” label on food that signals how items were sourced, I could imagine an “equity” label that demonstrates how data and machine learning programs were produced. In fact there is a Dataset Nutrition Label project that riffs off nutrition labels found on food, measuring and presenting the key ingredients of a dataset – for instance where, when, and by whom the data were produced. The project team aims to create standard quality measures that can be widely used as a prerequisite to developing more inclusive datasets.
As some observers have pointed out, we need to implement approaches that extend beyond the initial design. With machine learning, systems can become more discriminatory over time, as they learn to interact with humans. Thus avoiding discrimination requires that we scrutinize how “systems operate in practice.”74 This, in turn, requires transparency and accountability, which is why democratic oversight and engagement are vital.
Allied Media Network, mentioned previously, has been at the forefront of collaborating with community-based initiatives, as has the Detroit Community Tech Portal, for twenty years.75 As the organization Stop LAPD Spying Coalition, which is engaged in participatory action research to understand community members’ experiences of intensifying surveillance, the Detroit initiative crafted digital justice principles after surveying its members. Among other important shifts, “Digital justice demystifies technology to the point where we can not only use it, but create our own technologies and participate in the decisions that will shape communications infrastructure.”76 And it is not only concerned with access to technology, however important, but also with participation and common ownership designed to foster healthy communities.77 This is also something I have come to appreciate more in my engagement with Data for Black Lives, a growing collective of organizers, scholars, data scientists, and more.
In the aftermath of the scandal surrounding Russia’s use of social media to steer the 2016 presidential election, Data for Black Lives cofounder, Yeshimabeit Milner, wrote an open letter to Mark Zuckerberg, calling on Facebook to “commit anonymized Facebook data to a Public Data Trust, to work with technologists, advocates, and ethicists to establish a Data Code of Ethics, and to hire Black data scientists and research scientists.”79 A key tenet of the Data for Black Lives movement is that the data justice issues we are dealing with today are predicated on a much longer history of systematic injustice, in which those in power have employed data against Black lives. But not only that.