Our new open-access paper, “Large language model agents capture heterogeneous occupant behaviors in building energy simulation of shared office spaces”, has been published in Building and Environment.
The study demonstrates that building energy use is influenced not only by technology, but also by personality, social hierarchy and control over shared environments. In the simulated office, one occupant initiated 62.1% of environmental adjustments, while another accounted for only 1.3%.
The findings also show that reinforcement learning increased cooling energy consumption by 8.47% in the benchmark scenario, challenging the assumption that more sophisticated artificial intelligence necessarily produces more energy-efficient buildings.
