Critical theory is sometimes treated as an academic exercise, valuable for analysis but not directly applicable to the practical work of AI governance. This lesson challenges that view. Each critical tradition generates a set of questions that can be applied directly to AI governance decisions.

Questions from CRT:

• Who has historically been most harmed by practices similar to what this AI system does? Are those communities represented in its design?
• Does the training data encode historical discrimination? How?
• Who benefits from this system, and does that benefit converge with or diverge from the interests of the most affected communities?
• Are the people most affected by this system’s errors in a position to contest those errors?
• What counter-narratives from affected communities challenge the dominant account of how this system works?

Questions from feminist theory:

• Whose standpoint is reflected in this system’s design? Whose is absent?
• What gendered assumptions are encoded in the system’s categories and evaluations?
• Who performs the labour that makes this system possible, and under what conditions?
• What care responsibilities are invisible to the system’s model of human behaviour?
• Does the system reproduce patriarchal structures, and if so, how?

Questions from Marxist theory:

• Who owns this system? Who profits from it?
• How is it used to extract value from workers or users?
• What ideological functions does the framing of this system serve?
• Whose class interests shaped the regulatory environment in which this system operates?
• What alternative ownership and governance structures could serve broader human interests?

Questions from postmodern theory:

• What claims to objectivity does this system make? Are they warranted?
• What categories does this system construct, and what are the consequences of being classified by it?
• Whose discourse shapes what counts as a problem, a solution, and a success?
• What does the system’s transparency mechanism reveal and conceal?
• Who has the power to contest not just how the system made a decision, but whether the standards it applied are legitimate?

Using the toolkit in practice. These questions are not a checklist to be completed and filed. They are a disposition, a way of approaching AI systems with critical attention that notices what other frameworks miss. In practice, this means:

Asking them in design reviews, not just after deployment. The moment of design is when the most consequential decisions are made, and it is the moment when critical questions are most often absent.

Asking them across functions. The questions are not only for AI ethics specialists. They are for L&D professionals designing training programmes, HR professionals setting hiring criteria, IT governance professionals setting standards, legal professionals interpreting regulatory requirements, and executives making deployment decisions.

Asking them with affected communities. The most important answers to these questions come not from internal deliberation but from the people who live with the consequences of the systems in question.

Documenting the questions and the answers. The process of asking and answering critical questions creates an accountability trail, evidence that ethical deliberation occurred, and what it found.

Reflection question: Select three questions from the toolkit above that seem most relevant to your current work. What would it take to ask them consistently in your professional context?