One of CRT’s most distinctive methodological contributions is counter-storytelling, the deliberate foregrounding of the experiences of marginalised people as a form of knowledge that dominant narratives systematically exclude.

The dominant narrative in AI development is told by engineers, product managers, executives, and regulators. It is a narrative of capability, efficiency, and progress. The counter-narrative is told by the people whose lives are shaped by AI systems, the job applicant rejected by a hiring algorithm, the benefits claimant surveilled by a welfare system, the defendant assessed by a risk scoring tool, the community member photographed by a predictive policing camera. These voices are rarely in the room when AI systems are designed, evaluated, or regulated.

Counter-storytelling in AI governance takes several forms.

Community-led AI auditing, organisations like the Algorithmic Justice League, founded by Joy Buolamwini, document the experiences of people harmed by AI systems and make those experiences legible to policymakers and the public. This is counter-storytelling as political practice.

Participatory design, involving affected communities in the design of AI systems from the earliest stages, not as users to be consulted but as co-designers with genuine decision-making power. This is rare in practice but essential in principle.

Impact assessment with community participation, AI impact assessments that include structured collection of community experiences, not just technical audits of algorithmic outputs.

Community oversight boards, mechanisms that give affected communities ongoing oversight of AI systems that affect them, not just one-time consultation.

The resistance to these approaches is instructive. Counter-storytelling challenges the epistemic authority of technical expertise. It insists that the people most affected by AI systems have knowledge that the people building those systems lack, and that this knowledge is not merely anecdotal, not merely emotional, but analytically important.

CRT would note that the exclusion of community voice from AI governance is not an oversight. It reflects existing power relations, the same power relations that produce the discriminatory AI systems in the first place. Addressing algorithmic harm requires more than better algorithms. It requires a redistribution of epistemic and political power.

Reflection question: In your organisation, whose voices are included in decisions about AI systems? Whose are absent? What would it take to include them?