{"resourceId":"noble-division-ii-diving-qualifier-ai-case-2025","versions":[{"version":"external-f87f9a706ec0e207a5087b7fd6253423a2e2ee96ea736a7125bb2ccec3478c08","resource":{"id":"noble-division-ii-diving-qualifier-ai-case-2025","title":"Division II qualifier case exposes exception-handling errors before corrected results","organization":"Wichita State University; Jeff Noble","sector":"Collegiate athletics","geography":"United States; NCAA Division II","publishedAt":"December 2025; exact day unknown","publicationDate":null,"eventDate":null,"sourceName":"Journal of Contemporary Issues in Sport","sourceLabel":"Practice-driven technical case study","sourceUrl":"https://journals.wichita.edu/index.php/jcis/article/download/509/451/1030","evidenceClass":"academic-research","outcomeClass":"mixed","topics":["knowledge-work","developers-agents","governance-procurement","operating-model"],"finding":"ChatGPT initially omitted a diver in the men's qualification exercise; revised prompts produced the reported correct list. This is feasibility evidence, not independent reliability validation.","sledRelevance":"Direct Division II event-administration evidence adds a smaller-program rules-processing example.","evidence":"The author compared outputs with manually verified answers using an NCAA example and 2024 men's and women's prequalification data. Women's qualifiers were reportedly correct on the first attempt. No reproducible multi-run error rate, dataset row count or controlled labor baseline is supplied.","architectureImplications":"Interpretation: use a deterministic rules implementation as the oracle; keep any language interface outside the authoritative results pipeline.","governanceImplications":"Interpretation: freeze the applicable rule version and require an official to approve every final list.","securityPrivacyImplications":"Interpretation: use authorized competition records and minimize identifiers in model inputs; do not add medical or eligibility files.","caveats":"Practice-driven author has event-official experience; not an independent audit. Prompt tuning on known answers limits generalization. Methods cite 2024 datasets while a reference describes unpublished 2025 results. Exact event date remains unknown.","streamIds":["college-athletics"],"roles":{"sales":"Interpretation: The problem is dependable qualification processing under event deadlines. Engage event directors, officials and athletics IT. Ask how ties, duplicates and appeals are handled today and who certifies the final list. A bounded engagement could reconstruct one historical event with an independent rules engine and an optional language interface. The value hypothesis is quicker preparation of reviewable work. Do not promise fewer errors or staffing reductions from this case; include correction time and appeals in the local baseline.","engineering":"Interpretation: Fit is supervised administrative assistance. Prerequisites include authoritative rules, structured results and an independently approved expected answer. Test unseen tie combinations, duplicated names, missing scores and rule revisions, then repeat identical inputs to measure variation. Isolate document instructions as untrusted data and prohibit automatic result publication. Cloud use requires approved retention and identity controls; no source evidence establishes a need for specialized on-premises infrastructure. Compare the language workflow with conventional code, including total review effort.","delivery":"Interpretation: An event director should sponsor a shadow pilot with an experienced official and a developer. Build the independent test corpus before prompt tuning, train reviewers and rehearse fallback and appeal handling. Dependencies include licensed or authorized results and protected reviewer time. Proposed acceptance criteria are complete agreement with the approved oracle on all critical qualification cases, traceable sign-off and successful manual recovery. These are proposed gates, not observed outcomes. Stop expansion when unanticipated rule exceptions appear."},"retrievedAt":"2026-09-11T03:00:44Z","enrichedAt":"2026-09-11T03:01:33Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: teach officials to detect omissions and explain exceptions; preserve an accessible manual process.","procurementImplications":"Interpretation: require exportable inputs, version logs and reproducibility tests before buying an automated qualification service.","operatingModelImplications":"Interpretation: event director owns the decision; technical staff maintain tests and a rollback route.","updateExplanation":"Absent from 182 archive records scanned at offsets 0 and 100. Historical evidence newly adds a concrete administrative failure mode; no new September release is claimed.","sourceVerification":{"openedUrl":"https://journals.wichita.edu/index.php/jcis/article/download/509/451/1030","referenceExcerpt":"Initial outputs revealed minor discrepancies, including one omitted diver.","promptVersion":"sled-research-v3.1","model":null,"basis":"agent-reported inspection"}}}]}