Over recent years data science has become integral to many sectors across society. From learning analytics and public health to making sense of social media metrics, data science is now thought of as essential for optimizing performance, lifting profits and ensuring success. However, the insights derived from data science are only as good as the processes and principles upon which it is based. While there’s a lot of hype about the power and objectivity of data-driven insights, we now have a long list of cases in which data science has proven decidedly fallible. Think, for example, of the algorithms that classified black people as ‘gorillas’ or data that is used in ‘redlining’ particular neighbourhoods to deny residents various services. The problem is not simply with the ‘science’ per se, but the assumptions that data scientists have about the world which we now know influences virtually every stage of datafication.
In response to these issues, we have seen the emergence of data ethics and data justice. While both are steps in the right direction, there is now a pretty strong argument for advancing data justice over data ethics. As D’Iganzio and Klein (2020) explain in Data Feminism, data ethics might attempt to do good, but because it is mainly focused on technological fixes it never challenges the systems and structures that maintain oppression. In other words, data ethics keeps the roots of the problem in place. For example, rather than alleviating bias, which is a focus of data ethics, data justice seeks to alleviate oppression. Similarly, data ethics calls for transparency of data science processes, whereas data justice aims for critical reflexivity of data scientists. While the concept of data justice lays down a more significant challenge for the research community, it is only through a comprehensive shift in practices and dispositions that genuine change can take place.
One recurring theme throughout D’Ignazio and Klein’s book that is particularly useful to our Data Smart Schools project is the notion of co-liberation. Where data ethics might aim for accountability, data justice seeks co-liberation with those who are subjects to (and of) the data. In many respects, co-liberation provides a useful complement to our methodology of participatory design. However, co-liberation could be seen as more explicitly political in orientation. So, what might it mean to foreground co-liberation in our work with data in schools?
First, a commitment to co-liberation means working with all data stakeholders in schools throughout the project – from design through to data collection and analysis. As D’Ignazio and Klein explain, co-liberation ‘doesn’t mean “free the data,” but rather “free the people”’ (p.63). It is a ‘commitment to and a belief in mutual benefit, from members of dominant groups and minoritized groups’ (p.63). The idea of mutual benefit means not just the people with less privilege (i.e. students), but also those with more privilege (i.e. teachers and principals). Early findings from our project suggest that a wealth of data from numerous sources (automated and manually generated) is routinely collected about students. However, it is often one or two teachers (typically male maths teachers or those in administration), tasked with doing the data processing and analysis. Co-liberation encourages us to consider what would it mean to work with students to find out what sort of data they believe reflects their learning and well-being. How would that change the use of data in schools? Furthermore, it might involve asking individual teachers to collect data on their students in a way that reflects the demands of the subject area, rather than simply via external collection methods (i.e. NAPLAN etc).
Second, data is often used to hold teachers, administrators and principals accountable to those higher up the ‘chain’ than themselves. Here data processing and analysis is often done somewhere else (i.e. central administration) and the results left for staff to ‘speak to’. But what if teachers and Principals collected data about their own practices and their school’s processes? Such processes would ensure the data stays ‘local’ in a way recommend by Yanni Loukissas (2019) in order to create a ‘data settings’ rather than ‘data sets’. Data would be freed up to take a variety of forms with analysis providing an opportunity for critical reflection on practice and process.
D’Ignazio and Klein recommend that data science projects pursue two important outcomes. The first is knowledge transfer– i.e. a two-way exchange that involves ‘technical capacity building within the community so that any data products can be maintained and/or enhanced without requiring external expertise’ as well as ‘enhanced understanding of and respect for local knowledge for external collaboration’. The second outcome is building social infrastructures, which ‘involves an explicit focus on cultivating community solidarity through the project’ (p.142). These two outcomes ensure data processes in the school are empowering to all in the data setting, rather than a select few.
Co-liberation is a useful way to counter the new data power circulating in schools. D’Ignazio and Klein remind us that schools should aim for ‘…a well-designed, data driven, participatory process, one that centres the standpoints of those most marginalised, empowers project participants, and builds new relationships across lines of social difference’ (p.148). Before data-driven hegemony is normalised, co-liberation provides schools with the motivation and process to intervene.
D’Ignazio, C., & Klein, L. (2020). Data Feminism. Cambridge, MA: The MIT Press.