{"resourceId":"daily-tar-heel-unc-admissions-ends-slate-reader-ai-essay-scoring-20261025","versions":[{"version":"external-95b1bb06a7f36d9cdfce01eca620ec6831c19b4dcbc155f177a9222fef74e2df","resource":{"id":"daily-tar-heel-unc-admissions-ends-slate-reader-ai-essay-scoring-20261025","title":"UNC-Chapel Hill stops AI-assisted essay scoring in first-year admissions because AI-written essays made scores skew higher and lose meaning","organization":"University of North Carolina at Chapel Hill, Office of Undergraduate Admissions","sector":"Higher education admissions and enrollment","geography":"Chapel Hill, North Carolina, United States","publishedAt":"The Daily Tar Heel, 25 October 2026","publicationDate":"2026-10-25","eventDate":null,"sourceName":"The Daily Tar Heel","sourceLabel":"Student newspaper reporting","sourceUrl":"https://dailytarheel.com/488331/university/university-new-admissions-ai-policy/","evidenceClass":"independent-reporting","outcomeClass":"cautionary","topics":["governance-procurement","knowledge-work"],"finding":"UNC-Chapel Hill undergraduate admissions has stopped using AI tools in first-year essay review for the 2026-27 cycle. The office had used Reader AI in Slate to pre-read documents and to assign an AI-generated score reflecting word choice, sentence structure, vocabulary, grammar, punctuation and length. A university spokesperson said the AI-assisted grammar and writing rating was removed because it no longer provided a meaningful distinction, citing widespread use of AI writing tools by applicants: scores increasingly skewed higher, making the rating less useful. The associate provost for admissions said the office must regularly evaluate whether tools serve their purpose. Applicants and students quoted were divided on whether human readers are fairer.","sledRelevance":"Admissions is the entry point to the student-success pipeline; this is a documented institutional decision to retire an AI scoring tool after its signal degraded, a rare reported de-adoption.","evidence":"Institutional decision reported with an on-record spokesperson statement; the claimed score inflation is not quantified.","architectureImplications":"Reader AI is a feature of the Slate CRM; disabling a scoring feature is a configuration change. Interpretation: institutions should monitor score distributions of any automated rating over time.","governanceImplications":"Shows periodic review leading to retirement of an AI feature; the source does not describe how the review was conducted.","securityPrivacyImplications":"Not discussed; pre-reading applicant essays with AI involves sensitive applicant data.","caveats":"No data on the scale of score skew; the effect on admissions outcomes is unknown; reported via spokesperson statement.","streamIds":["student-success"],"roles":{"sales":"Sales framing: this is a documented case of an institution switching off an AI scoring feature because applicant AI use eroded its value. Use it to ask enrollment leaders which automated ratings they rely on and how they check that the ratings still discriminate. A bounded engagement could review AI features in an admissions CRM and set monitoring and retirement criteria. Do not claim AI scoring is useless everywhere; UNC published no data on the size of the score skew.","engineering":"Engineering framing: Reader AI is a built-in feature of the Slate CRM, so the fit question is how institutions monitor output distributions of embedded AI features. Check whether score histories can be exported by cycle, whether ratings are logged with reader decisions, and how features are toggled. Proof of value should track score variance and its relationship to reader decisions over cycles, flagging compression toward high scores as the signal UNC described.","delivery":"Delivery framing: the lesson is to include review and retirement in the operating model for any AI rating tool. Delivery work would define who owns each embedded AI feature, schedule a review each admissions cycle, track score distributions, and document decisions to keep or remove features. Communicate changes to applicants and counselors, since the reporting shows applicants worry about being flagged for AI and about fairness of human versus AI reading."},"roleDetails":{"sales":{"customerProblem":"Embedded AI ratings in admissions can lose discriminating power as applicants use AI writing tools.","stakeholders":["Associate provost for admissions","Admissions operations lead","Slate or CRM administrator","General counsel or compliance"],"discoveryQuestions":["Which embedded AI features score or summarize applicant materials?","How do you check whether a score still distinguishes applicants?","Who decides to retire an AI feature?"],"valueHypothesis":"Interpretation: regular monitoring of AI rating distributions can catch degraded tools before they distort review.","potentialEngagement":"A one-cycle review of AI features in the admissions CRM with monitoring and retirement criteria.","unsupportedClaims":["That AI essay scoring was biased or harmful","The size of the score skew","Any effect on who was admitted"]},"engineering":{"fit":"Institutions using AI features in enrollment CRMs such as Slate.","architectureIntegration":"Vendor CRM feature producing per-document scores alongside human reader workflow.","prerequisites":["Exportable score history by cycle","Logging of reader decisions","Ability to toggle features per workflow"],"constraints":["No published data from UNC","Vendor-controlled model","Applicant data sensitivity"],"security":"Confirm how applicant documents are processed and retained by the vendor AI feature before and after disabling it.","proofOfValue":["Track score variance per cycle","Measure correlation between AI scores and reader ratings","Define a drift threshold that triggers review"]},"delivery":{"work":"Set ownership, monitoring and retirement criteria for AI features used in applicant review.","dependencies":["CRM administrator","Admissions leadership","Compliance review"],"ownership":"Office of undergraduate admissions.","skillsAdoption":"Readers and operations staff learn how AI ratings are produced and when to disregard them.","governanceCheckpoints":["Pre-cycle feature review","Mid-cycle distribution check","Post-cycle keep or retire decision"],"acceptanceCriteria":["Proposed: each AI feature has a named owner","Proposed: score distributions are reviewed every cycle","Proposed: retirement decisions are documented and communicated"],"risks":["Relying on a degraded signal","Increased reader workload after removal","Applicant confusion about AI use in review"]}},"retrievedAt":"2026-10-26T03:16:50Z","enrichedAt":"2026-10-26T03:19:52Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Returns essay evaluation fully to human readers, potentially increasing reader workload; not quantified.","procurementImplications":"AI features bundled in enrollment CRMs should be evaluated for continued value, not assumed to stay useful.","operatingModelImplications":"Interpretation: AI measurement tools need drift monitoring and retirement criteria, as applicant behaviour can erode their value.","sourceVerification":{"openedUrl":"https://dailytarheel.com/488331/university/university-new-admissions-ai-policy/","referenceExcerpt":"Scores increasingly skewed higher, and thus the rating became less useful as an evaluative tool.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}