The Participation Paradox: When Stakeholders Co-Create AI but Do Not Transition to Use
Participatory and human-in-the-loop (HITL) approaches to artificial intelligence (AI) promise to embed stakeholder knowledge and human judgement in automated systems. Yet participation in designing an AI system does not necessarily translate into meaningful participation through it in practice. This article offers a reflexive methodological account of this gap through SOCYTI, a Norwegian project developing AI-supported approaches to online hate speech. Early stakeholder workshops were well attended, but participation declined sharply during hands-on testing, although those who took part gave detailed, critical feedback. Treating this decline as an epistemic and methodological event rather than resistance, we identify four forms of socio-technical misalignment: temporal, epistemic, interpretive and organisational. Drawing on participatory design, epistemic injustice, technology appropriation and human–AI interaction literature, we argue that participation must create conditions for stakeholders to understand, appropriate and act through AI. Without adequate knowledge, context, authority and manageable cognitive demands, the human “in the loop” may become a responsibility buffer for an uncertain system rather than an empowered decision-maker.
Publisert i Norsk MedieForsker Konferansen 2026, 2026
Les artikkelen her