FairTutor: Equity-Aware Pedagogical LLM Routing for Budget-Constrained AI Tutoring

FairTutor addresses educational inequity in AI-powered tutoring by introducing an equity-aware routing framework that maintains 97.1% of premium pedagogical quality while reducing costs by 71.6%. The framework uses multi-agent orchestration with selective escalation to premium models, introducing metrics to measure AI Education Advantage Gap between premium and budget-constrained services.

FairTutor tackles a critical emerging problem in educational technology: the widening gap between students who can afford premium AI tutoring services and those limited to budget alternatives. The research recognizes that as generative AI becomes central to personalized learning, unequal access to high-quality models risks amplifying existing educational disparities. This matters because AI tutoring systems are rapidly deploying in schools and learning platforms globally, making equity concerns immediately relevant.

The framework's innovation lies in pedagogically-motivated model routing rather than simple cost-cutting. By analyzing queries, planning appropriate pedagogical interventions, leveraging low-cost models for initial responses, and strategically escalating complex cases to premium models, FairTutor creates a cost-quality optimization problem rather than a binary choice. The 71.6% cost reduction while maintaining 97.1% quality represents a compelling efficiency gain that could democratize access to quality AI tutoring.

For EdTech developers and institutions, this research provides a practical template for scaling AI tutoring without proportional cost increases. The tunable Pareto frontier enables organizations to calibrate spending based on their constraints and student populations. The introduction of the AIED Advantage Gap metric and TutorAccessEval benchmark establishes new standards for measuring educational equity in AI systems.

The broader implication extends beyond tutoring: this model-routing approach demonstrates how to maintain quality while reducing costs across AI applications serving resource-constrained users. As educational AI deployment accelerates, frameworks that balance accessibility with pedagogical effectiveness become increasingly valuable for policymakers and institutions committed to equitable outcomes.

• →FairTutor achieves 97.1% of premium tutoring quality while reducing serving costs by 71.6% through intelligent model routing
• →The framework introduces AIED Advantage Gap metric to quantify educational inequity created by AI service tier disparities
• →Multi-agent orchestration with selective premium model escalation enables cost-effective pedagogical quality optimization
• →TutorAccessEval benchmark spans five subject areas, providing standardized equity measurement for AI tutoring systems
• →Tunable cost-quality Pareto frontier allows institutions to tailor AI tutoring investment to diverse student population needs