One of feminist theory’s most important contributions to epistemology, the study of knowledge, is its critique of claims to objectivity.

The dominant model of knowledge in science and technology holds that valid knowledge is objective: it stands apart from the knower, free from the distortions of perspective, identity, and interest. The ideal scientist is a neutral observer. The ideal algorithm is a neutral processor. Both are supposed to see the world as it is, not as any particular person experiences it.

Feminist philosophers of science, most notably Sandra Harding and Donna Haraway, have subjected this ideal to devastating critique. All knowledge, they argue, is situated. It is produced from a particular standpoint, shaped by particular experiences, interests, and power relations. The claim to objectivity, to the view from nowhere, is not a description of how knowledge is actually produced. It is an ideological move that conceals whose perspective is being treated as universal.

This is not relativism, the claim that all perspectives are equally valid. Feminist standpoint epistemology does not deny that some claims are better supported by evidence than others. It argues that recognising the situated character of knowledge production makes us better at it, more attentive to whose experiences are being generalised from, whose are being excluded, what assumptions are built into the questions we ask.

Applied to AI, this critique is immediately practical. AI systems are built by people with particular standpoints, predominantly male, predominantly white, predominantly from wealthy countries, predominantly employed in elite technology companies. These standpoints shape every design decision: what problem is worth solving, what data is worth collecting, what counts as an error, what counts as success.

When these systems claim objectivity, when they present their outputs as the result of neutral mathematical processing rather than human choices made from particular standpoints, they are making a claim that feminist epistemology has systematically undermined.

Donna Haraway’s concept of situated knowledge argues for what she calls a feminist objectivity, one that is more rigorous precisely because it acknowledges its own situatedness. A facial recognition system that acknowledges it was trained predominantly on lighter-skinned faces, and that its performance degrades for darker-skinned faces, is more epistemically honest than one that claims accuracy without qualification. Situated knowledge does not weaken AI. It strengthens it, by forcing accountability for what the system actually knows and does not know.

Reflection question: Whose standpoint is reflected in an AI system you work with or know about? What assumptions does that standpoint bring? What does it make invisible?