Campaigns on Data: Reflecting on Algorithms and Infrastructure
In working on resources for our Community Campaigns on Data cohort I’ve grappled a lot with the question of what, if anything, makes a campaign about data distinct?
When I’ve first talked to potential collaborators for the programme, they’ve often interpreted it as about using data within campaigning, rather than campaigning to change how data is used (or collected, or shared). This can be complicated by the fact that sometimes data is a tool, as well as target focus, of many campaigns.
However, even once we establish that some dataset or data system is the direct focus of campaign actions, this often begs the question: why? And in the vast majority of cases data systems are only part of an intermediate campaign goal. In our upcoming design lab we’re planning to explore this through a simple workshop activity creating ‘User story placards’:
“What do we want?” “Access to landlord registry data!”
“Why do we want it?” “To improve rental property”.
Data can be the enabler or infrastructure for further campaign actions, or the lever with which to shift the behaviour of a system. But from the perspective of campaigns other than ideological open data type campaigns, changing the governance of data is valued because of the consequential changes it can enable or effect.
I’ve been reading David Robinson’s Voices in the Code this week, which presents a powerful account of public engagement around an organ donation algorithm. It’s notable that algorithms, as clear sites of decision making, appear more tractable as a focus of campaigning, than the datasets that feed them. Yet, while it might be easier to hold up an algorithm-campaign focussed placard saying something like “Fair housing allocation now”, not all data campaigns relate to a specific algorithm, and even where algorithmic decision-making is involved, seeking to act upon the data supply-chain of analysis and algorithmic decision making can have greater reach and impact over the long-term than fighting the fight decision-by-decision, and algorithm-by-algorithm.
This comes though, at the cost of moving ‘upstream’ into data infrastructures that can, at first, be hard to make legible to the public, and that are often sitting at the intersection of many different communities and interests. Just as a transit line used by morning commuters, evening clubbers, weekend tourists and so-on is a common infrastructure, serving many different communities who might be almost unaware of each other, a data infrastructure might have multiple stakeholders who rarely, if ever, meet. In addition then, to virtual invisibility, data infrastructures can be tightly embedded and interconnected. This can be a challenge when the change a campaign wants to see creates incidental costs or impacts for orthogonal actors, who might therefore oppose it. Or it could be an opportunity, when other ‘travellers’ using the same datasets and systems can be enlisted in the campaign for change.
In this week’s design lab we’re planning explorehow activities based around the idea of a data cycle can assist campaigners in identifying the implication of these ‘infrastructural’ opportunities and challenges for their campaign strategies.
But do we need to go into this depth in community campaigns? Can we not just focus tightly on the ultimate outcome, and leave the details of data schemes, systems and sharing as technical questions to be addressed by officials or narrow advocacy? In Voices in the Code, Robinson suggests not. Policy measures resulting from higher-level goal-oriented campaigns are almost always underspecified when it comes to software or dataset design, and “technical experts are left to decide central moral questions” (p 28). Robinson shows this happening in two ways: then technical experts “imperiously take over moral choices”, or, perhaps more commonly, when “the rest of us fall silent”, allowing that “Technical intricacy can act as an escape hatch from moral challenges.”(p 29).
While Voices in the Code goes on to set out institutional measures of transparency, participation, forecasting and audit that organisations can adopt to mitigate the problems that might come from delegating algorithmic implementation to technical experts alone, an argument can also be found here for building campaigns from outside the institution that are able to engage with the detail of policy implementation within data: linking grassroots publics and affected communities, with in-depth discussions over data fields, codelists and data sharing agreements.
Which leaves the question of what this linkage might look like? Today we had a Community Campaigns Cohort session focussed on co-production: exploring one set of strategies for working with communities as co-producers of research, knowledge, priorities and advocacy strategies, often starting with literacy building, and making systems of data power more legible to communities. I’m also curious to explore strategies that use campaign actions to reveal the system and build literacy: such as inviting people to resist filling in particular form fields, or to join collective actions about data sharing agreements and so-on. Hopefully after Thursday’s workshop we’ll have a lot more on this.