Dong et al. propose contextual ambiguity measure for law texts

A new EMNLP 2026 paper proposes measuring ambiguity in laws, contracts and policies by watching how much a set of stakeholders disagree about which rule applies in which specific scenario. The work, from Sally Dong, Luke Thorburn, Andrew Konya and Michiel Bakker, was posted to arXiv on 26 October 2026.

The authors frame ambiguity in prescriptive text as something that can cut both ways. "Ambiguity in prescriptive text such as laws, contracts, rules, policies, or recommendations can be both desirable (e.g., if it enables a law to adapt to changing context, such as new technologies) and undesirable (e.g., if it leads to arbitrary discretion or conflict over the law's interpretation)," they write.

Existing measures, they argue, "capture how text is perceived in isolation rather than how it is interpreted in context." Their proposal is to treat ambiguity as "the degree of disagreement among a set of stakeholders ("raters") on how alternative prescriptive texts ("policies") apply in each of a set of relevant scenarios ("cases")."

Preliminary analysis across two policy domains, the paper reports, suggests the measure "correlates with real human disagreement, and (ii) is robust to spurious text features" even when the raters are simulated rather than real people. The authors then apply it to existing deliberation datasets to describe the association between a policy's ambiguity and public support for it. Two researchers in our directory flagged the preprint shortly after it went up.

The abstract publishes no per-domain numbers, does not name the deliberation datasets, and does not say which model stands in for the simulated raters.