CLARITY: An Exploratory Scaffold for Educational Engagement with AI-Assisted Evidence Synthesis in Rehabilitation.

Abstract

To describe and evaluate CLARITY, an AI-assisted literature synthesis scaffold designed for structured learning and hypothesis generation in rehabilitation medicine. Proof-of-concept case series using ScholarAI, which integrates GPT-assisted triage and fairness-aware reweighting. All outputs are exploratory and not for clinical application. Open-access rehabilitation literature applied in academic and clinical training environments. Synthesis speed, adjudicator concordance, demographic representation, and fairness-adjusted effect sizes. CLARITY reduced review timelines from months to days, achieving 84.7% concordance in clinician adjudication of borderline studies. Fairness adjustments (±20%) modestly altered pooled estimates and highlighted underrepresentation trends. SHAP and LIME overlays improved transparency but did not affect inclusion outcomes. CLARITY is an AI-assisted educational scaffold (not yet a clinical decision aid tool), supporting efficient, reproducible, equity-conscious engagement with literature to stimulate hypothesis-generation and critical dialogue. Although limited by single-reviewer design and absence of audit logs, it provides a valuable entry point for structured AI-assisted synthesis in early-stage or educational contexts. Outputs are strictly exploratory and require expert interpretation.