“Algorithmic tools embed and perpetuate societal and historical biases and injustice.”
Abeba Birhane’s work on relational ethics challenges the idea that an algorithm can be evaluated apart from the society in which it operates. Data, categories, institutions, and technical objectives all emerge from histories of power. A system that appears neutral at the level of code may reproduce injustice through what it predicts, whom it serves, and which relationships it changes. The wording is compact because it identifies the change the authors wanted readers to notice, rather than turning a complex project into a promise of automatic progress.
Relational ethics begins from interdependence rather than an isolated individual. Harm is not only an incorrect output; it can include damaged trust, surveillance, stigma, exclusion, or a shift in power that makes people less able to shape decisions. Context is therefore part of the technical problem. The same model may have different consequences depending on the institution, community, and alternatives available. A careful reading separates the mechanism from the mythology that later accumulates around it. The quotation becomes more useful when its assumptions remain visible.
AI teams can apply this perspective by involving affected groups before goals and datasets are fixed, not merely after a prototype exists. Evaluation can include qualitative research, histories of the domain, labor conditions, environmental costs, and distribution of benefits. Designers should ask who gains authority, who becomes more visible, and who has the power to refuse or seek redress. The modern relevance lies in method, not nostalgia. Researchers and product teams can preserve the original curiosity while demanding stronger documentation, independent checks, and realistic operating conditions.
Checklists can flatten relational concerns into another compliance score. Participation can also become symbolic if communities lack information, time, compensation, or influence. Organizations must be prepared to change the objective, restrict the use, or decline to build the system. A respectful process does not guarantee agreement, but it makes power and accountability harder to hide behind technical language. A responsible interpretation keeps people in the frame. Technical success is incomplete if the system weakens agency, obscures accountability, or shifts risk onto those with the least power to object. Birhane’s sentence is a warning against historical amnesia. Algorithms participate in relationships, and responsible design must examine how those relationships are transformed. For readers, the practical habit is to return to the evidence: identify the task, examine the conditions of success, and ask who can act when reality differs from the demonstration. The work continues with care.
Models operate inside institutions shaped by history, incentives, authority, and unequal access to resources.
The same technical output can have different meaning and consequences depending on those relationships.
Algorithmic harm may include stigma, surveillance, exclusion, damaged trust, and shifts in who can make or contest decisions.
Qualitative evidence and community knowledge can reveal effects that benchmark scores do not capture.
Consultation is not meaningful when affected people lack information, compensation, time, or influence over the outcome.
Organizations should be willing to change goals, limit deployment, or decide that a system should not be built.
Explore more "Quotes of The Day"
Continue exploring how writers, thinkers, artists, and advocates have described the choices that shape an everyday life.