Paper Type

Short

Paper Number

PACIS2026-2157

Description

Tourism is one of the world’s largest industries but generates significant environmental and social impacts, including carbon emissions, overtourism, and pressure on local communities and cultural heritage. Most travel recommendation systems optimize for user preferences while overlooking sustainability considerations, a challenge further complicated by the lack of sustainability certifications for many destinations. This paper presents SMTRec (Sustainable Multi-Agent Travel Recommender), a system that integrates personalized travel recommendations with automated sustainability assessment. SMTRec introduces a Sustainable Place Score (SPS) that evaluates Points of Interest across four dimensions: environmental impact, community and economic benefit, cultural heritage, and governance. The system employs a multi-agent architecture powered by Large Language Models to perform conversational preference elicitation, semantic retrieval, sustainability-aware ranking, and itinerary generation. Demonstrations on the TourPedia dataset illustrate the feasibility of integrating sustainability-aware scoring within LLM-driven multi-agent travel recommendation systems.

Comments

01-AIML

Download

Share

COinS
 
Jul 5th, 12:00 AM

SMTRec: A Multi-Agent System for Sustainable Travel Recommendations

Tourism is one of the world’s largest industries but generates significant environmental and social impacts, including carbon emissions, overtourism, and pressure on local communities and cultural heritage. Most travel recommendation systems optimize for user preferences while overlooking sustainability considerations, a challenge further complicated by the lack of sustainability certifications for many destinations. This paper presents SMTRec (Sustainable Multi-Agent Travel Recommender), a system that integrates personalized travel recommendations with automated sustainability assessment. SMTRec introduces a Sustainable Place Score (SPS) that evaluates Points of Interest across four dimensions: environmental impact, community and economic benefit, cultural heritage, and governance. The system employs a multi-agent architecture powered by Large Language Models to perform conversational preference elicitation, semantic retrieval, sustainability-aware ranking, and itinerary generation. Demonstrations on the TourPedia dataset illustrate the feasibility of integrating sustainability-aware scoring within LLM-driven multi-agent travel recommendation systems.