Ethical Web and Data Architectures – they give our project its name, but what precisely makes a technical artefact like a web or data architecture, ethical? This is an excellent question and one we receive often.

This is also a question for which the existing web provides some good examples. Wikipedia seems to miraculously negotiate consensus across millions of voices to deliver truth in a mis-and-disinformation-ridden web. The SOLID project promises to enable a new age of web interoperability and personal data control through new protocols. Data trusts, collaboratives, and cooperatives put forward a diverse vision for collective governance in their own respective, brighter, data futures (Brandusescu, Ana and van Geus, Jonathan 2020).

These proposals all latently embody values of fairness or autonomy, but are dotted with intricate voting mechanisms, technical implementations, and esoteric concepts for what the future technologies might look like. Seldom are these designs for data futures relatable to, nor designed by the data subjects who they aim to benefit. For those who will inevitably experience these data futures, design questions are not focused on the lightning rod terms of data governance like quadratic voting, homomorphic encryption, or ephemeral services. They are more along the lines of how do I stay safe, make more money, or connect with community with the help of my data, rather than in spite of it (Zhang et al. 2022)? Though Human Computer Interaction (HCI) values user participation in design, existing research has steered clear of allowing users to participate in data stewardship, labelling data architectures and protocols as too technically complex for the data literacy of the average web user, or simply a matter of implementation details (Bødker and Kyng 2018). We see these architectures differently. They are the product of the socio-technical systems and relations which they create, support, or automate and can themselves be as (un)ethical as those systems too.

In a new paper accepted to this year’s ACM CHI Conference on Human Factors in Computing Systems (CHI23), EWADA researchers challenged this assumption. The paper, led by EWADA PhD student Jake Stein with help from EWADA summer intern Vid Mikucionis, developed new methods to elicit what data architectures should embody based on the direct input of data subjects themselves. In the paper, the EWADA team engaged with gig workers driving and riding with platforms like Uber, Lyft and Deliveroo, whose algorithmic management practices makes workers particularly vulnerable to exploitative information asymmetries. The research provides meaningful strides bridging the gap between novel data infrastructures and the practical everyday challenges created by data stewardship. The EWADA team puts forward a new methodology that retools canonical HCI methods from their typical domain of interface and interaction design to address the fundamentally contingent and socially-bound dynamics of data architecture building.

Contrary to prevailing thinking, the team found data subjects of varying data literacy levels were excited to participate in collective data stewardship. In simulated data stewardship tasks, the participants were able to first internalise the affordances of separate data structures, then provide meaningful input into the design of collective data infrastructures based on their reflection. In our methodology, the EWADA team highlighted the great potential of using existing legal tools including Data Subject Access Requests to augment commonly used participatory design tools like online whiteboards.

Beyond demonstrating new means for participatory data architecture design, the participants expressed important considerations that should be included in designs for collectively governed data architectures. These included the need for dispute, discussion, correction and contextualization mechanisms, granular permissions affordances, and community-facing analytics portals. Participants also contributed a wide variety of ideas about which stakeholders should participate in new data institutions and their respective roles from labour unions and governments, to autonomous collectives, affinity groups based on ethnicity and gender identity, and online communities. The study concludes by providing two points of reflexive consideration. It questions researchers’ role in gig economies, questioning how researchers can intervene to expose information asymmetries without replicating gig-like work conditions and providing valuable information to workers. Finally, it calls for future research to continue developing participatory design methods suitable for data data architecture design pointing to the danger for architectures to repeat patterns they aim to resist.