CIRO: Agentic UDL Tutoring for At-Risk Students

At-risk students (first‑generation, low‑income, and historically underserved learners) withdraw from college at higher rates than their peers. Research shows that early intervention, personalized feedback, and multimodal engagement improve retention, yet universities often lack the human resources to scale such support. Conversational generative‑AI (GenAI) tutors offer a promising, cost‑effective alternative, but existing systems rarely align with validated learning frameworks or undergo rigorous evaluation with at‑risk populations. We introduce CIRO (Conversational Intelligence for Retention Optimization), a GenAI‑powered tutoring platform grounded in Universal Design for Learning (UDL) principles. CIRO employs retrieval‑augmented generation (RAG) and intelligent agent architecture to deliver personalized academic support through three UDL-aligned mechanisms: 1. Engagement: Real-time scaffolding and gamified progress dashboards to sustain motivation 2. Representation: Multimodal content delivery to adapt to diverse learning preferences 3. Action & Expression: Flexible assessment options to accommodate various demonstration methods A controlled experiment in Fall 2025 will enroll 80 first‑year at‑risk students across four seminar course sections (two CIRO-enabled, two control) to test three hypotheses: H1-CIRO increases academic engagement (time‑on‑task, interaction counts); H2-CIRO improves representation (content‑comprehension scores); H3-CIRO enhances action & expression (quality of submitted work). Using a mixed-methods approach combining quantitative metrics with qualitative insights from focus groups and instructor observations, we will isolate CIRO’s impact from existing support services. Expected contributions include (1) empirically validated UDL design guidelines for GenAI tutors targeting at-risk populations, (2) evidence regarding GenAI Agentic efficacy in improving specific learning dimensions with at‑risk students, and (3) a scalable, ethical reference architecture for institutions seeking affordable AI‑driven academic interventions.