The "10th Juror": Open-Set Standpoint Screening for Bureaucratic Bias Detection
We study closed-loop bias governance for Dutch government documents, where a system must detect biased language, ground decisions in legal and contextual evidence, rewrite problematic sentences when intervention is warranted, and verify that the rewrite mitigates harm without distorting meaning.
ProofPaper ↗
Key points
- Presupposing the boundaries of bias is itself a form of bias.
- Existing methods face three challenges: (i) discriminative classifiers capture surface regularities but lack normative grounding; (ii) zero-shot LLMs often adopt generic viewpoints and over-flag ambiguous administrative language; and (iii) fixed taxonomies inherit the Closed-World Assumption, missing emerging local targets.
- We propose MARS-Gov, a standpoint-aware multi-agent framework that combines legal retrieval, open-set target screening, specialized jurors, conservative routing, and rewrite verification.
- On DGDB, MARS-Gov sets a new SOTA with 0.880 F1, outperforming the strongest zero-shot LLM detector by 20.2 points (29.8% relative) and the best supervised Dutch encoder by 6.8 points, while reducing unnecessary interventions to 2.5%.
Sources (1)
- [1]The "10th Juror": Open-Set Standpoint Screening for Bureaucratic Bias DetectionarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 03:05 AM
We study closed-loop bias governance for Dutch government documents, where a system must detect biased language, ground decisions in legal and contextual evidence, rewrite problematic sentences when intervention is warranted, and verify that the rewrite mitigates harm without distorting meaning.
Presupposing the boundaries of bias is itself a form of bias.
Extractive summary: sentences quoted from the sources.