[2604.03881] Enhancing behavioral nudges with large language model-based iterative personalization: A field experiment on electricity and hot-water conservation

Computer Science > Computers and Society arXiv:2604.03881 (cs) [Submitted on 4 Apr 2026] Title:Enhancing behavioral nudges with large language model-based iterative personalization: A field experiment on electricity and hot-water conservation Authors:Zonghan Li, Yi Liu, Chunyan Wang, Song Tong, Kaiping Peng, Feng Ji View a PDF of the paper titled Enhancing behavioral nudges with large language model-based iterative personalization: A field experiment on electricity and hot-water conservation, by Zonghan Li and 5 other authors View PDF HTML (experimental) Abstract:Nudging is widely used to promote behavioral change, but its effectiveness is often limited when recipients must repeatedly translate feedback into workable next steps under changing circumstances. Large language models (LLMs) may help reduce part of this cognitive work by generating personalized guidance and updating it iteratively across intervention rounds. We developed an LLM agent for iterative personalization and tested it in a three-arm randomized experiment among 233 university residents in China, using daily electricity and shower hot-water conservation as objectively measured cases differing in friction. LLM-personalized nudges (T2) produced the largest conservation effects, while image-enhanced conventional nudges (T1) and text-based conventional nudges (C) showed similar outcomes (omnibus p = 0.009). Relative to C, T2 reduced electricity consumption by 0.56 kWh per room-day (p = 0.014), corresponding t...