FairnessBiasAccountabilitycs.CY
AI From the Margins (AIM): Rethinking Participatory AI Design Through the Lived Experience of Minoritized Communities
Tijs Portegies, Laureanne Willems, Maaike Harbers, Giovanni Sileno, Roland van Dierendonck, Mayesha Tasnim, Lotte Willemsen, Sennay Ghebreab
May 31, 2026
Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society
1 citations
Key Takeaways
-
Early participation matters more than late involvement: Traditional participatory AI approaches often begin after key decisions about problems and success metrics are already made. This study shows that when minoritized communities participate from the earliest stages—helping define what an AI system should address rather than just how it works—they can fundamentally reshape the purpose and design of AI systems, rather than simply providing feedback on predetermined solutions.
-
Lived experience should directly inform AI policy, not just research: The study demonstrates a concrete model where participants' personal narratives and experiences were translated into actual policy recommendations for municipal decision-makers. Participants described this engagement as "substantive," suggesting that when lived experience is genuinely centered in dialogue with policymakers, it can influence real-world AI governance decisions.
-
Participatory AI requires specific preconditions to be meaningful: AIM identifies four key conditions for effective participation: (1) eliciting narratives through structured methods like biographical interviews, (2) allowing participants to co-create rules, (3) letting communities decide whether AI should be used at all, and (4) ensuring policymakers hear directly from affected communities. Without these preconditions, participation risks becoming tokenistic.
-
Communities should have the power to say "no" to AI: A critical finding is that participants determined not just how AI would be designed, but "whether, where, and how AI should be involved." This suggests responsible AI design must include the genuine option for minoritized communities to reject AI solutions entirely, rather than assuming AI is always the answer.
-
Participants want ongoing engagement, not one-time consultation: Study participants called for continuation of the engagement process, indicating that meaningful participatory design is iterative and relational, not a single consultation event.