{"resourceId":"bird-castleman-embedded-supports-pvcc-2026","versions":[{"version":"external-ccbaf4be76e07c0a0ce9cd7c8a5509a787322a6d4bf8016fee375570b49b57d4","resource":{"id":"bird-castleman-embedded-supports-pvcc-2026","title":"Community-college RCT: classroom-embedded tutoring/advising raises take-up, pass rates, and persistence","organization":"University of Virginia / Piedmont Virginia Community College / EdWorkingPapers (Annenberg)","sector":"Higher education student success / community college advising & tutoring","geography":"United States (Piedmont Virginia Community College, VCCS)","publishedAt":"EdWorkingPaper 26-1485, May 2026","publicationDate":"2026-05-01","eventDate":null,"sourceName":"EdWorkingPapers working paper (open PDF)","sourceLabel":"Bird & Castleman (2026) — Meeting People Where They Are","sourceUrl":"https://edworkingpapers.com/sites/default/files/ai26-1485.pdf","evidenceClass":"academic-research","outcomeClass":"effective","topics":["operating-model","knowledge-work","accessibility-workforce"],"finding":"Multi-semester RCT randomizing 52 gateway course sections at PVCC (Fall 2023–Fall 2024) embedded tutors and advisors in treatment classrooms while keeping identical centralized on-demand services for controls. Treated students were ~31pp more likely to meet a tutor and ~29pp more likely to meet an advisor; +5pp more likely to pass and 4.6pp less likely to withdraw. Spillovers included higher outside-course help-seeking, semester GPA gains, and about 10pp higher subsequent-term persistence (10.1pp next semester; 10.8pp two-term). Larger pass gains among new students (+31.5pp).","sledRelevance":"Unarchived community-college causal evidence directly on tutoring/advising take-up and persistence—the stream’s brief—after recent AI early-alert coverage (e.g., ACC Digital Twin) that lacked outcome RCTs. Why now: shows friction reduction and default access points, not information or chatbots alone, move durable success metrics.","evidence":"Source findings: section-randomized ITT; Navigate meeting logs; VCCS admin outcomes; first-gen heterogeneity. Limitations: single campus; three terms; some withdrawal visibility limits; intervention is human embedding, not GenAI.","architectureImplications":"Interpretation: student-success architecture should co-locate human supports with instructional defaults; AI referral layers should route into embedded capacity with measurable handoffs.","governanceImplications":"Interpretation: evaluate early-alert/AI investments against utilization and persistence, not engagement vanity metrics; preserve centralized equity access while testing embedding.","securityPrivacyImplications":"Meeting and early-alert logs are FERPA-sensitive; interpretation: require least-privilege Navigate/SIS roles and no consumer-model training on advising notes.","caveats":"Not an AI chatbot RCT. Single Virginia CC limits external validity. Persistence gains are promising but campus-specific.","streamIds":["student-success"],"roles":{"sales":"Sales framing: cite this May 2026 EdWorkingPaper multi-semester RCT (52 PVCC gateway sections) showing large tutoring (+31pp) and advising (+29pp) take-up plus +5pp course pass, −4.6pp withdraw, and ~10pp next-term persistence when human supports are embedded in class—while control students kept the same centralized offices. Position AI early-alert tools as referral accelerators into embedded human capacity, not substitutes. Bound proposals to gateway-course staffing models with persistence endpoints.","engineering":"Engineering framing: treat embedding as a delivery-architecture change—put staff where students already are—rather than a new prediction model. Prerequisites include section rostering, meeting logs, and IR joins. Constraints: single-campus RCT; not an AI efficacy test. Proof of value: utilization deltas plus persistence, not chatbot engagement alone.","delivery":"Delivery framing: delivery work is staffing and classroom logistics, not software install. Ownership sits with student success and academic affairs. Acceptance requires utilization and persistence metrics. Risks include declaring victory on alerts without embedding capacity and eroding tutoring centers."},"roleDetails":{"sales":{"customerProblem":"Community colleges need evidence that reallocating tutoring/advising into classrooms beats centralized office access for take-up and persistence.","stakeholders":["Student success VP","Advising leadership","Tutoring center","IR/assessment","Faculty senate"],"discoveryQuestions":["Where do at-risk students actually encounter tutors/advisors today?","Will you randomize or phased-rollout embedded sections?","How will you measure pass, withdraw, GPA, and next-term persistence?","What staffing model funds in-classroom presence without cutting centralized capacity?"],"valueHypothesis":"Embedding targeted tutoring and advising in gateway sections can sharply raise utilization and improve pass, withdrawal, GPA, and persistence versus identical centralized services alone.","potentialEngagement":"Multi-section embedded-support pilot in high-DFW gateway courses with shared outcome dashboard—not a chatbot seat license.","unsupportedClaims":["That chatbots alone replicate embedded human-support gains","Guaranteed 10pp persistence at every campus","That information campaigns equal embedding"]},"engineering":{"fit":"Fits student-success stacks that already have tutoring/advising staff and need higher take-up among at-risk gateway students.","architectureIntegration":"Section-level assignment flags in SIS; Navigate/EAB or equivalent meeting logs; LMS early-alert feeds to embedded staff; IR joins for pass/withdraw/GPA/persistence.","prerequisites":["Staffing plan for in-section tutors/advisors","Faculty partnership protocols","Shared meeting taxonomy","Outcome warehouse"],"constraints":["Single Virginia community college","Three-term window","Administrative data miss some withdrawn students","Not an AI intervention"],"security":"Student meeting and early-alert data are education records; minimize vendor training use and enforce role-based access.","proofOfValue":["Pre-register pass/withdraw/persistence primaries","Report first-gen heterogeneity","Compare embedded vs centralized utilization, not tool clicks"]},"delivery":{"work":"Schedule embedded tutors/advisors into high-DFW sections, train faculty on warm handoffs, and instrument Navigate/meeting logs.","dependencies":["Tutoring/advising FTE","Faculty buy-in","SIS section flags","IR reporting cadence"],"ownership":"Student success operations jointly with academic affairs and tutoring/advising directors.","skillsAdoption":"Staff need classroom facilitation and proactive outreach skills; faculty need referral norms.","governanceCheckpoints":["Pre-term staffing confirmation","Midterm utilization review","Post-term persistence readout with equity cuts"],"acceptanceCriteria":["≥25pp tutoring take-up lift in treated sections","Pass/withdraw deltas reported","Next-term persistence tracked for first-gen subgroup"],"risks":["Understaffing embedded roles","Cutting centralized services prematurely","Substituting AI outreach for in-person presence"]}},"retrievedAt":"2026-09-20T03:08:33Z","enrichedAt":"2026-09-20T03:15:04Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: first-gen and new-student heterogeneity suggests embedding may reduce navigation barriers for underserved students, but disability-specific effects are unmeasured here.","procurementImplications":"Interpretation: procure staffing/time for embedded roles and logging integrations before buying additional outreach bots.","operatingModelImplications":"Interpretation: shift default access from office hours students must seek to in-section presence; keep centralized services as backstop.","sourceVerification":{"responseId":"curl-edwp-ai26-1485-2026-09-19","openedUrl":"https://edworkingpapers.com/sites/default/files/ai26-1485.pdf","referenceExcerpt":"treated students were 31 percentage points more likely to meet with a tutor, 29 percentage points more","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}