Beyond Abstract Compliance: Operationalising trust in AI as a moral relationship

Trust requires ongoing relationships, not just technical fixes: The paper challenges the common approach of treating trust as a property that can be "designed for" through compliance frameworks alone. Instead, trust must be cultivated over time through sustained, meaningful engagement between AI developers and the communities affected by their systems.

Cultural context and community participation are essential: Drawing on African communitarian philosophies, the research emphasizes that trust is "culturally embedded" and cannot be standardized globally. Involving communities throughout the entire AI lifecycle—from design through deployment—helps ensure AI systems are equitable and responsive to local contexts, as demonstrated in their healthcare and education case studies.

Trust is a two-way street requiring mutual respect: The paper reframes trust as a "dynamic, temporal relationship" involving transparency and mutual respect between developers and users. This means communities should have genuine input into decisions, not just be informed after choices are made.

Current compliance frameworks are incomplete: While frameworks like the EU's "Trustworthy AI" provide important normative and technical standards, they miss how trust is "subjectively cultivated and experienced" by actual people. Responsible AI requires going beyond abstract compliance to address the human and relational dimensions of trust.

Long-term commitment builds equitable AI systems: The research suggests that incremental trust-building through participatory processes leads to "more equitable and context-sensitive AI systems," implying that shortcuts to trust undermine both fairness and effectiveness.