Co-constructing sociotechnical AI governance: participatory system mapping using algorithm registers

The research highlights challenges in current algorithm registers designed for transparency in public services. It notes a disconnect between the expectations of diverse publics regarding transparency and their ability to interpret existing information. A key finding is that these registers struggle to represent the complex sociotechnical systems where algorithms are embedded, thus limiting their effectiveness in facilitating accountability. The paper explores how algorithm registers reveal or obscure governing sociotechnical systems and seeks to integrate diverse stakeholder perspectives for a more comprehensive, system-theoretic approach to safety.

Why it matters

This research is strategically important because it addresses the foundational challenges in ensuring effective governance and accountability of AI and algorithmic systems, particularly in public services. A robust framework for transparency and accountability is crucial for maintaining public trust, mitigating risks, and fostering responsible innovation in AI development and deployment across various sectors.

Key insights

• Algorithm registers are championed for transparency in public service algorithm use.
• Public expectations for transparency and ability to parse information in registers vary.
• Current registers struggle to represent the sociotechnical systems algorithms are embedded in.
• The effectiveness of registers in facilitating accountability due to system-level representation is unclear.
• The research aims to understand what registers reveal or occlude about governing sociotechnical systems.
• It seeks to incorporate diverse stakeholder perspectives for a more pluralistic, system-theoretic safety approach.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.12166