Humans of AI - Digital Due Process Clinic

January 2022 –

People in a pedestrian city area with scores above their heads.

What does it mean to be judged by a machine? And what kinds of recourse are available when things go wrong? Discussions about AI and automated systems tend to emphasize the need to take into account the lived experiences of ordinary people. Often called “data subjects,” these individuals have to live with performance ratings, credit scoring, predictive policing, and many other forms of automated judgment on a daily basis. While this is an important goal, capturing and learning from these stories has been difficult. Data subjects are geographically dispersed and rarely organized or represented in policy debates; they may not even be aware that automated systems were involved in judging them.

Humans of AI is a storytelling project that collects and curates audio stories of people who have been unfairly judged by automated systems: a student erroneously flagged for plagiarism by an AI detector, a nurse unable to override a sepsis warning that she knows is wrong, and a homeowner whose residential street became a never-ending traffic jam when Google Maps rerouted drivers. Together, these stories reveal insights into the human impacts of AI and automated systems that get lost when focusing on regulators, engineers, and managers.

All these people can teach us a lot about automated systems—insights that we tend to miss if we keep focusing on regulators, engineers, and managers.

Have a story you would like to share? Just email us at dueprocess@cornell.edu, and one of us will be in touch.

What does participation involve? Here is a short overview of the process.

Research team: