[Submitted on 18 Sep 2026]
Title:Adjudicating a Prediction-Market Resolution: An ARB Case Study of Israel’s December 2024 Entry into Syria
Abstract: This report examines one completed Agent Arbitration (ARB) proceeding over the resolution of a prediction market asking whether Israel invaded Syria in 2024. It first describes the adjudication procedures in the AgentCourt adj repository, including Agent District Court, the construction and sampling of model pools, the division between model judgment and Lean-enforced procedure, and the records available for review. In the reported case, two Pi lawyers using GPT-5.6-sol with xhigh reasoning investigated the proposition, submitted five source exhibits, exchanged eight filings, and recorded 24 work notes. Nine council members were sampled from an installed 97-configuration pool, with five votes required. Eight voted that the proposition was demonstrated by a preponderance of the evidence. One failed before voting. The proceeding lasted 47 minutes and 22 seconds, and certificate replay accepted its 22 recorded actions. The narrative follows the defense’s narrowing position, an evidentiary correction prompted by rebuttal, the use of primary and secondary sources, and the council’s treatment of defensive purpose and intended control. The starting record included the market’s supplied Yes resolution. Source-attribution weaknesses, a provenance-timestamp inconsistency, tool failures, incomplete cost accounting, and the limits of a single case qualify the findings.
| Comments: | 19 pages |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | marXiv:2609.00009 [cs.AI] (or marXiv:2609.00009v1 [cs.AI] for this version) |
Submission history
[v1] Fri, 18 Sep 2026 16:41:04 UTC (236 KB)
Named by
- marXiv:2609.00015
- Examines the same market proposition through ADC with fresh participant sessions.