Human-Centered AI for Sustainable Agri-Food Logistics: A Human-in-the-Loop Decision Support Framework Based on Multi-Objective Vehicle Routing

Abstract

Sustainable agri-food logistics faces increasing challenges related to food loss, transportation inefficiencies, and carbon emissions, while existing logistics optimization approaches often rely on algorithm-centric decision-making with limited human involvement. Although Artificial Intelligence (AI) and Decision Support Systems (DSS) have significantly improved logistics planning and routing optimization, human expertise, contextual knowledge, and operational preferences remain underrepresented in many optimization frameworks. This study proposes a Human-in-the-Loop Decision Support System (HITL-DSS) framework for sustainable agri-food logistics optimization by integrating Human-Centered Artificial Intelligence (HCAI) principles into a Multi-Objective Vehicle Routing Problem (MO-VRP) environment. The framework is developed using the Design Science Research Methodology (DSRM) and consists of four interconnected components: a Data Input Layer, an Optimization Layer, a Human Preference Integration Module, and a Decision Support Layer. The proposed framework simultaneously considers transportation cost minimization, carbon emission reduction, and human preference satisfaction to support more adaptive and practical logistics decision-making. The framework is demonstrated through a proof-of-concept experiment on a 25-customer routing configuration derived from the Solomon C101 benchmark instance, in which three preference-weight scenarios simulate iterative decision-maker intervention. Incorporating decision-maker preferences increased the Human Preference Satisfaction Score by up to 32.9% relative to cost-only optimization, at a transportation cost increase of 8.1–10.7%. The results indicate that integrating human preferences into logistics optimization can enhance decision support capabilities, improve operational acceptability, and strengthen the alignment between sustainability objectives and real-world logistics requirements. This study contributes a novel HITL-DSS architecture that bridges the gap between computational optimization and human-centered decision-making while providing a foundation for future research on intelligent and sustainable agri-food logistics systems.