This project explores the craft behind data work as a grounding point for civic engagement with AI risks.

In a first phase, the research explored data enrichment workers' perspectives on their work. In a series of qualitative interviews, workers individually discussed what brought them to the industry, what skills they see as being essential, and what matters to them in this work

The goal of this phase of research was to un

The process

From qualitative research analysis to embodied illustrations and embroideries

In a context where data work is rendered invisible in the data supply chain, conveying the result of the thematic analysis as a textual summary felt like an incomplete way to share what was learned. With this in mind, each theme that emerged from the qualitative research was transformed in a visual representation.

Catherine D’Ignazio and Lauren F. Klein (2020) discuss in depth the value of embodying and opening up to emotions when presenting data, one of their core principles behind Data Feminism. All drawings represent a certain motion, a fluidity behind the work, with common visuals like the data squares or circles. The image below highlights the process taken from thematic analysis to emrboidery.

The skills & labour behind data work

01

Flexibility and adaptability are shared as key skills to have to do data enrichment work: the capacity to accept new requirements, to move as the data moves. This is partially to support the need to be resilient when the data landscape changes. But it is also survival for an industry where little agency is given to the workers.

02

For workers, their actions carry a responsibility. It goes beyond doing the bounding box and the tagging. Data work comes with a sense of responsibility. In certain cases,  the labellers themselves show clients some blind spots. The duty behind the work can lead to significant changes in the machine development cycle.

03

The longer you work with data, the more you accumulate its sediments. Workers shared how data can be intense and tedious to examine, but also how some types of work may go against their values. The accumulation of data seen becomes almost a sedimented layer in their head.

04

The act of labelling is one of diligence and guidance. It often involves researching much more than what has been provided, to help teams be prepared for different types of data. It is an ongoing act of calibration, where humans are themselves validating their knowledge, and then transferring it to machines.

05

Data workers are lucid on the potential fragility of this type of work, and concerned about what this means for job security. They see the potential drying out of data labour, and consequences of automation on the flow of data coming in. The lack of visibility and the uncertainty make for a challenging industry to evolve in.

06

Data work brings out something different in each participant, in terms of how they see their professional evolution. It is perceived as something that can challenge you, get you out of your comfort zone. Even in recognising the fragility of the industry and its opacity, workers expressed how it also supported the growth of new skills, enabling them to try new things.

07

A final overarching theme relates to the larger scope of data work. Workers are motivated when looking at the impact this may have on society, the why behind their work. They feel pride in seeing the end-product they contributed to. The awareness that the quality you put in can help technology become better.

Transforming data & insights into tangible artefact

The choice of digital embroidery connects with the transformation workers are also seeing in their craft, and the potential threats to their expertise as AI models get better and better.

Parallels have been made between needlework and machines. Transforming the findings of the data conversations in a hybrid artefact, manually drawn but automated during the needlework, repeats this tension.

combined_grid_1.mp4