The Big Picture
A single, practical governance recipe lets a community score and adopt proposals across many decision types in polynomial time while avoiding cyclical outcomes; it gives strong honesty incentives in one-dimensional settings and protocol-level protection for the chosen median rule.
ON THIS PAGE
The Evidence
Design a constitution that ties each amendable item (budget, bylaws, board, rates, etc.) to a metric space and an aggregation rule, and require proposals to meet a supermajority support threshold. Scoring each candidate against the standing status quo avoids the classic cycling problem of pairwise majority fights and keeps per-round computation tractable. Using the generalised median as the aggregator yields strategy-proofness in one dimension and a protocol-level guarantee protocol-level guarantee (no misreport weakly dominates sincere voting) when the threshold is set to half. The system supports public proposals (including AI- or optimization-generated ones) and even self-amendment of the constitution under higher thresholds.
Data Highlights
1Per-round computation is polynomial: each round runs in at most O(n^2) time over the active proposal set (so practical for typical community sizes).
2Compromise gap is zero in one-dimensional decisions (exact best outcome reachable) and is provably bounded (Lipschitz) in higher dimensions.
3Protocol-level honesty: for the generalised median with threshold σ = 1/2, no misreport weakly dominates sincere voting (there exists scenarios where honest voting strictly helps).
What This Means
Platform engineers building community governance, multi-agent orchestration, or decentralized decision systems can use this as a practical, audit-friendly contract for making many kinds of decisions. Technical leaders running cooperatives, DAOs, or federated services get a unified, computationally-feasible alternative to stitching together ad-hoc mechanisms for each decision type. Researchers in collective decision-making can test aggregation rules and semantic distances inside a single, composable framework. Also see Constitutional AI for alignment-aware governance considerations.
Not sure where to start?Get personalized recommendations
Ready to evaluate your AI agents?
Learn how ReputAgent helps teams build trustworthy AI through systematic evaluation.
Learn MoreKeep in Mind
The guarantees depend on the chosen metric: textual bylaws need a meaningful semantic distance (not just edit distance) for outcomes to align with intent. semantic distance is an example of a meaningful metric. Strategy-proofness is proven for the median at threshold 1/2 and in one dimension; extending the protocol-level result to higher supermajorities (σ > 1/2) is left open. Because proposals are drawn from the revealed votes plus public proposals, the framework trades global optimality for tractability — unconstrained optima can be missed, though the gap is bounded and often small in practice.
Methodology & More
The framework models each amendable part of a community’s rules as a metric space where every member reports an ideal point (their vote and personal proposal). A constitution assigns, per component, a support threshold and an aggregation rule; rounds lock votes, accept public proposals (from members, coalitions, optimization, or AI mediation), and score each proposal by aggregating member utilities measured as distance improvements over the status quo. A proposal wins only if it has supermajority public support and maximal aggregate score, otherwise the status quo remains. Public proposals must be novel, supported, and strictly improve the winning score to be admitted, and epochs terminate after two-round quiescence. The framework yields several practical properties: every round either chooses a winner or none in polynomial time (bounded by O(n^2) over the proposal set), epochs terminate under mild space conditions, and comparing each proposal to the fixed status quo prevents cyclical majority outcomes. Instantiating the aggregator as the generalised median gives robustness to outliers, exact optimality in one-dimensional settings, and protocol-level incentives against misreporting when the constitutional threshold is 1/2. The approach covers seven canonical governance tasks (rates, budgets, rankings, committees, texts, elections, and constitutional parameters), admits AI-generated proposals, and supports self-amendment at higher supermajorities — offering a unified, implementable governance pattern for digital communities. aggregation rule
Avoid common pitfallsLearn what failures to watch for
Credibility Assessment:
One author (Nimrod Talmon) has a strong h-index (27) indicating an established researcher; arXiv venue and other author has low h-index, but the presence of a high-impact author supports a higher rating.