Exploring the capabilities of large language models in seat recommendation systems for hot-desking offices

Hiroaki Murakami, Keiichiro Taniguchi, Yosuke Kamiya, Katsuya Koike, Yoshihisa Toshima, Yasunori Akashi, Yoshihiro Kawahara
Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, November 2024
https://doi.org/10.1145/3671127.3698179

Abstract
Hot-desking offices, where users move to seats according to their preferences, have garnered attention for the efficient use of office space and cost-saving benefits, highlighting the need for an effective seat recommendation system. Traditional systems focus on specific parameters such as temperature or illuminance; however, user preferences vary widely, making sensor-dependent systems limited and impractical due to the high cost and complexity of deploying various sensors. This paper explores the potential of an approach where large language models (LLMs) process and utilize linguistic feedback, such as workplace complaints, for seat recommendations. We developed an LLM-powered system and evaluated it in a living lab environment, examining both the capabilities and limitations of LLMs in enhancing seat recommendation systems.