{"resourceId":"manikonda-uneven-genai-reliance-survey-20260916","versions":[{"version":"external-02a157c2128f3a608e8a897775d571fdcfeed5b2246aac4354f3f6354587b4c0","resource":{"id":"manikonda-uneven-genai-reliance-survey-20260916","title":"RPI survey: GenAI learning benefits uneven across reliance style, access, and course policy","organization":"Rensselaer Polytechnic Institute","sector":"Higher education / academic research","geography":"United States (single R1 university)","publishedAt":"2026-09-16","publicationDate":"2026-09-16","eventDate":null,"sourceName":"arXiv","sourceLabel":"Preprint (arXiv:2609.18676)","sourceUrl":"https://arxiv.org/pdf/2609.18676","evidenceClass":"academic-research","outcomeClass":"mixed","topics":["knowledge-work","accessibility-workforce","governance-procurement","operating-model"],"finding":"Among 118 students in 12 AI-related courses, four usage clusters emerged; self-reported academic benefit associated with early reliance and task support, while negative impact associated with early reliance and attitudinal change, strengthening as evaluation literacy rose. Premium and multi-tool users reported more benefits; instructor policy related to confidence and perceived impact. Why now: posted 2026-09-16 and absent from the student-success archive after the 2026-09-19 edition.","sledRelevance":"Directly informs U.S. higher-ed student-success programs deploying GenAI in courses: durable learning and equity depend on how students rely on tools and whether access/policy are equitable—not on adoption counts alone.","evidence":"Cross-sectional survey (N=118) with hierarchical clustering, FDR-controlled group comparisons, and multiple regressions with HC3 and course-clustered SEs. Constructs include cognitive reliance, early reliance, evaluation literacy, and policy clarity (1–4 Likert). Self-report outcomes only; single R1.","architectureImplications":"Prefer institution-provisioned GenAI with comparable capability tiers; instrument reliance features and policy metadata rather than raw query counts.","governanceImplications":"Move beyond permit/restrict binaries toward clarifying appropriate uses, coaching cognitive reliance, and reviewing differential access to premium tools.","securityPrivacyImplications":"Survey and interaction data are student education records; minimize collection and keep on approved platforms with retention limits.","caveats":"Self-report, single R1, AI-course sample, cross-sectional associations—not causal learning effects or multi-institution generalization.","streamIds":["student-success"],"roles":{"sales":"For enrollment and student-success teams, this RPI survey shows GenAI benefit is uneven across reliance style, tool access, and instructor policy—not a single adoption curve. Premium and multi-tool users report more benefits, while early reliance links to negative impact especially among higher evaluation-literacy students. Position discovery around how students actually use tools before promising equity or learning gains. Bounded value is clarifying appropriate uses and access gaps, not guaranteeing GPA or retention lifts from GenAI availability alone.","engineering":"Engineering takeaway: instrument cognitive versus early reliance, evaluation literacy, and policy clarity alongside LMS usage logs. Architecture should prefer campus-approved models with equitable access tiers rather than unmanaged consumer mixes that recreate free-versus-premium splits. Prerequisites include survey or telemetry definitions for reliance constructs and course-clustered analytics. Constraints: single-institution self-report; no causal learning design. Security: keep student GenAI interaction and survey PII in FERPA-aligned systems. Proof of value: replicate cluster and regression patterns on local AI-course cohorts before scaling.","delivery":"Delivery work is a mixed-methods pilot in AI-related courses: field a short reliance/evaluation survey, map instructor policies, and compare free versus institution-provisioned access. Dependencies include IR/ethics approval, teaching-center partners, and SIS/LMS extracts. Ownership sits with teaching & learning plus student-success analytics. Skills: survey design, multilevel regression literacy, and faculty development on productive GenAI use. Governance checkpoints: policy clarity review and equity access check. Acceptance: documented cluster profiles and adjusted associations; risks include overinterpreting self-report as causal learning effects."},"roleDetails":{"sales":{"customerProblem":"Campuses fear GenAI widens gaps or harms learning without knowing which student practices drive benefit versus harm.","stakeholders":["provost / teaching & learning","student success / advising leads","faculty in AI-related courses","IT / academic technology"],"discoveryQuestions":["How do your AI-course policies distinguish permitted uses?","Do students have equitable access to premium tools?","Do you measure early versus cognitive reliance, not just usage volume?"],"valueHypothesis":"Bounded advisory value: surface reliance and access disparities that correlate with self-reported benefit/harm before large GenAI purchases.","potentialEngagement":"4–6 week discovery: survey AI-course students, map policies/access, and brief leaders on local cluster patterns—not an outcome guarantee.","unsupportedClaims":["Guaranteed GPA or retention gains from GenAI access","Causal proof that early reliance harms learning","Transfer of RPI AI-course patterns to all majors"]},"engineering":{"fit":"Fits as measurement and policy-instrumentation layer around existing GenAI tools in AI-related courses.","architectureIntegration":"Combine course-policy metadata, optional approved-tool SSO, and survey or interaction-feature logging; avoid training on student prompts without DPA.","prerequisites":["IRB/ethics path for student survey","Defined constructs for early vs cognitive reliance","Instructor policy taxonomy"],"constraints":["Single R1 self-report sample","Cross-sectional associations only","AI-course population may not generalize"],"security":"Store survey and chat telemetry as education records with least-privilege access and retention limits.","proofOfValue":["Reproduce four-cluster structure locally","Test early-reliance × evaluation-literacy interaction","Compare free vs provisioned access subgroups"]},"delivery":{"work":"Stand up survey instrument, policy inventory, and analysis notebook; facilitate faculty workshop on productive reliance.","dependencies":["Faculty volunteers in AI courses","Analytics capacity","Approved GenAI tooling decision"],"ownership":"Center for teaching & learning jointly with student-success research lead.","skillsAdoption":"Train instructors to coach cognitive reliance and discourage premature GenAI consultation.","governanceCheckpoints":["Survey consent language","Equity review of premium access","Publication of policy clarity updates"],"acceptanceCriteria":["N and response rate documented","Cluster and regression outputs reviewed by IR/analytics","Action list on policy and access—not grade promises"],"risks":["Students misread research as surveillance","Policy changes without faculty buy-in","Overfitting to AI majors"]}},"retrievedAt":"2026-09-21T03:05:53Z","enrichedAt":"2026-09-21T03:18:00Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Free-versus-premium and single-versus-multi-tool gaps can recreate unequal learning support; evaluation literacy without access equity may intensify perceived harms of early reliance.","procurementImplications":"Procure campus-approved GenAI with equity of access and analytics for reliance patterns; do not treat consumer ChatGPT uptake as the deployment model.","operatingModelImplications":"Teaching centers and student success should jointly own GenAI literacy that targets when to use tools, not only whether use is allowed.","sourceVerification":{"responseId":"web-fetch-curl-arxiv-2609.18676","openedUrl":"https://arxiv.org/pdf/2609.18676","referenceExcerpt":"early reliance was associated with greater negative impact primarily among students with average or higher evaluation literacy.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}