In her latest book If…Then: Algorithmic Power and Politics, Taina Bucher uses the notion of ‘relational ontology’ to analyse the power and politics of algorithms. As she explains relational ontology derives from actor network theory, and conceptualises action as arising from a ‘heterogeneity of actors, including non-humans or the more-than-human’ (p.49). In this blog post I outline Bucher’s approach to sociomateriality to complement the notion of materiality that was outlined in an earlier post. I want to extend on this to explore how sociomateriality and relational ontology can shape the way we think about and approach accountability and agency in schools. After all, any attempt to enhance the use of digital data in schools will depend on teachers, students, school administrators and parents believing that what they do will be of benefit to themselves and/or others – in other words, to have some impact on the world.
A core tenet of relational ontology is sociomateriality. A sociomaterial approach does not consider the social and technological separate, but instead as dependent on each other. Indeed, this suggests ‘a radical symmetry between human and nonhuman actors’ (p.50). As such the social and the technical are always ‘engaging in symbiotic relationships organised in networks, assemblages, or hybrids’ (p.50). This can lead to new and novel arrangements (or ‘agencements’ in French) of composite entities that are capable of expressing an ‘agential force’ or even, according to Muller (2015), ‘new realities’. For schools, the assemblage of heterogenous actors would include digital devices and data infrastructure, as well as the system developers, software engineers and platform operators, as well as the practices and discourses that datafy schooling.
While these theoretical constructs are useful, perhaps more important for our methodology is how the idea of sociomateriality relates to accountability and agency in the datafied school. If nothing else, then sociomateriality draws attention to the fact that agency itself is difficult to define. As Bruno Latour (2005) writes agency is ‘the most difficult problem there is in philosophy’ (p.51) because agency no longer lies with the individual user but is distributed acrosshuman and nonhuman actors. Not only does this make the notion of ‘developing agency’ in individuals more complex, but it also makes it more difficult to hold someone or something accountable when things go wrong. Bucher draws on the ‘racist’ Google algorithm launched in their Photos app that tagged two black people as ‘gorillas’. Who should be held to account for this act? If there is a ‘radical symmetry’ between human and nonhuman actors, then is it the algorithm that is racist?
Rather than focusing on where agency is located and to whom it belongs, a relational ontology encourages us to analyse ‘when agency is mobilised and to what end’ (p.151). In this sense, realities are never ‘given’. Instead, realities emerge from, and are shaped by, interactions between different actors in the assemblage. The politics of algorithms are not so much determining who acts or discriminates, but how entities and interactions coalesce to produce particular outcomes. A relational approach to studying the politics of algorithms (or the politics of data in schools for that matter) requires a methodology of tracing and mapping, discursive analysis and reverse engineering, interviewing and observation to reveal the intersections of actors and the decisions that are made in those moments.
Developing agency in the students, staff and parents who participate in our research is another matter. Once the data infrastructures (both material and immaterial) have been traced, mapped and analysed, then exploring this new ‘landscape’ with participants will certainly help to reveal moments of decisive data, as well as the analytical frames individuals bring to data and data-work. These are the pressure points that can then be explored in terms of how data might be ‘done differently’ … with both human and non-human actors seen as having potential to change.