Signal Co-constructing sociotechnical AI governance through participatory system mapping

Summary

Researchers examined what a municipal algorithm register reveals and hides about the sociotechnical structure of algorithmic governance, using a Dutch city's register for a welfare-eligibility decision-support tool as a case. Eight participants, including municipal staff, civil society organizations and ombudsmen, took part in interviews, surveys and participatory system-mapping workshops. This allowed the researchers to assess how far the register alone lets stakeholders map the system. The workshop findings fed into a System-Theoretic Process Analysis. That analysis identified hazards not visible from the register alone, including wrongful denial of welfare benefits, deterioration of system performance and an inability to contest wrongful decisions. The authors reflect on the normative dimensions of algorithm governance and how politics shapes the practice of system safety analysis.