Automating Essential Work

Millions of people were deemed “essential workers” during the Covid-19 pandemic. To mitigate potentially dangerous exposure to the virus — for both workers and the public — critical sectors rapidly integrated artificial intelligence (AI) in the hopes that technology could perform this material, in-person work. ‘The Transformation of Essential Work’ is a National Science Foundation-funded project that investigated how AI was adopted and adapted by essential workers. As Co-PI’s, Sarah E. Fox and I conducted a multi-sited field study in two critical sectors: recycling sorting and janitorial work.

Our research revealed the way that the imperfect operation of AI increased the amount of labor required of essential workers, and made their daily duties more complex and more technical. Essential workers responded with ingenuity, modifying AI infrastructure and their practices to improve AI on-the-ground. You can read more in our article “Patchwork: The Hidden, Human Labor of AI Integration in Essential Work,” published in Computer Supported Cooperative Work.

Representing the Invisible Work of Integration

Read the pictorial (PDF) in the proceedings of the 2021 conference on Designing Interactive Systems. Or read the shortened magazine-article version for students in XRDS: Crossroads (PDF).

The accomplishments of essential workers are everywhere — clean floors, sanitized tables, objects made from recycled plastics — though the activities themselves often occur behind the scenes. This illustrated article identifies a set of visual patterns (or “tropes”) that are used in news reporting on automated technologies being implemented in essential workplaces, and ultimately obscure workers.

We then created a set of counter-visuals that don’t just focus on the features of new technology and those who build them, but instead represent the unacknowledged workers who integrate, reconfigure and repair AI.

In collaboration with Franchesca Spektor and Sarah E. Fox (Carnegie Mellon University) and Estefania Rodriguez (UT Austin).

Journalistic Resource: Tips for Reporting on AI

Given the urgent context of the Covid-19 pandemic, we used our initial finding to create and distribute a set of accessible documents that provided actionable tips for reporters covering AI in essential work sectors. Our 1-sheet “Don’t Be A Drone” (PDF) encourages journalists to include workers as sources, to visit sites after the moment of introduction, and to be wary of narratives that are produced technology companies. A shorter, blog-post version written for newsrooms is available at the Center for Media Engagement.

These tips are based on our exhaustive analysis of five years of news reporting about AI in essential work sectors and detailed in our Executive Summary (PDF). We found that the experiences of on-the-ground essential workers were mostly absent from news articles. In our data-set of nearly 50 articles, none featured on-the-ground workers as a quoted source. Instead, robotics companies, business executives, and other industry leaders dominated the reporting and set the agenda for how automation and AI technologies were portrayed. Without workers’ voices to discuss the reality of robotics in the workplace, the articles overhyped the actual capabilities of the automated machines and erased the additional responsibilities and labor placed on workers.

This project was lead by journalism major and undergraduate research assistant Estefania Rodriguez.

The communications team at UT wrote a profile about our research project for International Women’s Month. You can read the full story “Inside Recycling Facilities and the Robots that ‘Run’ Them: The Untold Story of Women and AI”.