The Power of Absence: Thinking with Archival Theory in Algorithmic Design

Algorithmic bias often stems from historical absences, not just flawed data: The paper argues that bias in AI systems reflects what's missing from datasets and design processes—absences rooted in historical exclusions and power imbalances—rather than treating bias as merely a technical problem to be "fixed" through better data cleaning.

Traditional bias-mitigation approaches may be insufficient: Rather than focusing solely on reducing bias in existing systems, the paper suggests that archival theory offers a more fundamental reframing that can address the deeper historical and political structures that create bias in the first place.

Absence itself is a form of power that shapes algorithmic outcomes: Drawing on archival scholarship, the paper proposes that what is deliberately or inadvertently left out of training data and design decisions actively shapes algorithmic behavior—making the study of these gaps as important as studying what's included.

Responsible AI design requires interdisciplinary perspectives: The paper demonstrates that fields like history, Black studies, and feminist scholarship offer valuable insights for understanding algorithmic bias that technical approaches alone cannot provide, suggesting AI developers should engage with humanities and social science expertise.

A more "joyful" approach to algorithmic justice is possible: By reframing bias through archival theory, the paper suggests that addressing algorithmic fairness doesn't have to be purely deficit-focused, but can open creative possibilities for designing more equitable systems that acknowledge and learn from historical contexts.