Explainable AI-Assisted Multi-Criteria Decision Support for Tutor Recommendation

Abstract

Manual tutor assignment in online tutoring marketplaces
becomes increasingly time-consuming, inconsistent,
and challenging to scale as the number of available tutors
grows. This paper proposes an Explainable AI-assisted Multi-
Criteria Decision Support framework for tutor recommendation,
developed and evaluated on the Hamro Tutor platform. The
framework combines a deterministic Multi-Criteria Decision
Making (MCDM) engine, which first applies grade-level, subject,
and gender eligibility filtering and then ranks eligible candidate
tutors using weighted scoring across location, availability, budget,
qualification, and teaching experience, with a Large Language
Model (LLM)-based explainability layer that generates humanreadable
justifications for each recommendation. Unlike fully
autonomous recommendation systems, the proposed framework
keeps the human administrator in the decision loop, using AI to
support rather than replace human judgment. The system was
evaluated on 10 real student tuition requests against a pool of 30
tutors, comparing manual and AI-assisted assignment workflows.
Results show that AI-assisted selection reduced average decision
time from 240.4 seconds to 22.6 seconds per request, a reduction
of approximately 90.6%. Three platform administrators rated
the suitability of AI-recommended tutors at an average of 4.57
out of 5, and rated the clarity of generated explanations at 4.77
out of 5. These findings suggest that explainable, human-in-theloop
AI decision support can meaningfully reduce administrative
effort in tutor allocation while achieving positively perceived
recommendation suitability and explanation quality.