The Critical Lens: Synthesising Theory for Responsible AI

Intellectual honesty requires acknowledging that the four critical traditions in this series do not simply complement each other. They also disagree, sometimes sharply, and these disagreements are productive rather than merely problematic.

Postmodern scepticism about identity categories versus CRT and feminist use of them. CRT and feminist theory both depend on stable identity categories, race, gender, class, as analytical tools. They use these categories to identify patterns of discrimination, measure representation, evaluate equity, and mobilise political action. Postmodern theory, following Foucault and Derrida, is deeply sceptical of such categories, insisting that they are constructed, contested, and historically variable.

This tension is real. If racial categories are social constructions, does that undermine CRT’s use of them to analyse racial discrimination? If gender is performed rather than inherent, does that undermine feminist analysis of gendered harm?

The most productive responses to this tension argue that both things are simultaneously true: race and gender are socially constructed and they have real material effects. The construction does not eliminate the reality. A loan denial based on a variable that correlates with race is a real harm to a real person, regardless of whether race is ultimately a natural category or a social one. Acknowledging the constructed character of categories while attending to their real effects is the epistemologically sophisticated position, one that postmodern theory helped produce.

Marxist emphasis on economic structure versus postmodern emphasis on discourse. Marxist theory tends to treat economic relations as the base from which cultural and ideological forms, including discourse, are derived. Postmodern theory, particularly Foucault’s, treats discourse as constitutive, not derived from an underlying economic reality but one of the systems through which reality is produced.

This is a fundamental theoretical difference with practical implications. If economic relations are primary, AI ethics reform requires structural economic change, redistribution of ownership, transformation of labour relations, democratic control of means of production. If discourse is primary, AI ethics reform requires changing the frameworks through which AI is understood, the narratives, assumptions, and common sense that naturalise existing arrangements.

The most productive position treats both as necessary and interacting: economic structures produce and are sustained by discourses, which in turn shape economic possibilities. Neither base nor superstructure is primary in a simple sense. Both require attention.

Liberal feminist reform versus structural change. Within feminism, there is ongoing tension between liberal feminist approaches, focusing on representation, equal opportunity, and anti-discrimination law within existing institutions, and socialist or radical feminist approaches that argue these reforms are insufficient without structural change to economic and political institutions. This tension plays out directly in AI ethics: diversity initiatives versus collective data governance, bias auditing versus democratic ownership.

Reflection question: Where do you find yourself most drawn to in these theoretical tensions? What does that tell you about the assumptions you bring to AI ethics?

Scroll to Top