SMaRT: Online Reusable Resource Assignment and an Application to Mediation in the Kenyan Judiciary

A multi-university research team has introduced SMaRT (Selecting Mediators that are Right for the Task), a novel AI algorithm designed to solve the complex, high-stakes scheduling problem within the Kenyan judiciary. The system addresses the challenge of assigning over 2,000 mediators—each with varying, often unknown, quality and specific geographic and case-type qualifications—to incoming cases in real-time across 87 locations and 12 case types. This scale and complexity, featuring soft capacity constraints and a high-dimensional state space, rendered existing scheduling algorithms inefficient. The team formalized the problem using a tractable quadratic program for assignment and a multi-agent bandit framework to learn mediator quality on the fly, creating a system that can make immediate, optimized assignments as cases arrive.

The algorithm's performance was validated on both stylized instances and real-world Kenyan judiciary data. SMaRT consistently outperformed baseline methods by providing superior control over the critical trade-off between adhering to mediator capacity limits and maximizing overall case resolution rates. This holds true both in settings where mediator quality is known and in more realistic 'bandit' scenarios where it must be learned through interaction. On the strength of these computational results, the researchers plan to conduct a randomized controlled trial (RCT) with SMaRT in the Kenyan judicial system, moving from simulation to real-world deployment. This represents a significant step in applying advanced, learning-based operations research to improve the efficiency and accessibility of public sector services at a national scale.