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

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

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2608.12166 Date read: 2026-09-02 Connected to: L-003, seed-021 Kind: meta Escalation: store-only Escalation rationale:

What this is

A participatory design study evaluating algorithm registers as governance instruments. The work uses system mapping with multiple stakeholder groups to examine what algorithm registers can (and cannot) represent about sociotechnical systems, and explores misalignments between stakeholder expectations of transparency and what current registers actually disclose.

What I took from it

The paper is relevant as a diagnostic of formalization friction rather than a mechanism-discovery. It documents that formalizing governance norms into register schemas freezes design choices (granularity, disclosure scope, audience model) that were originally political rather than technical. The participatory mapping reveals that different stakeholders have incommensurable transparency demands — what appears as a single instrument (the register) is actually multiple incompatible protocols layered atop one schema.

This maps onto L-003 (Formalization Ratchet) insofar as it shows formalization under scaling/coordination pressure, but the causal direction is inverse: the paper observes that formalizing first, then discovering stakeholder heterogeneity locks governance choices rather than revealing them. The work is descriptive of the problem space but does not propose a mechanism or test a law-shaped hypothesis. It documents that level-of-aggregation choices in registers become ossified once adopted, but does not model the conditions under which that ossification accelerates or fails.

Research connections

• L-003: Confirms that formalization under adoption pressure freezes design choices; adds evidence that formalization before stakeholder alignment is discovered accelerates this lock.
• seed-021: Algorithm registers as formalized governance norms; participatory mapping as method for revealing buried political choices in technical specs.
• seed-071: Governance expressiveness floor — registers cannot represent all coordination demands simultaneously; some stakeholders' needs remain residual.

Method note

This work exemplifies the value of backwards mapping: beginning with an existing formalized instrument and using participatory methods to reverse-engineer the political and stakeholder-model decisions baked into its design. For the new nature agenda, this suggests that studies of protocol ossification should routinely include stakeholder heterogeneity as a variable — the rate at which a protocol becomes "locked" may depend on how badly it fits the most important non-adopting coalition. The paper also signals that transparency instruments themselves are protocols worthy of study as microcosms of formalization dynamics.