Dr. Paul Bouman
Period: November 2025 – October 2026
Funded by: NWO
The "Prompting for Personality" NWO-XS project explores how Large Language Models (LLMs) can be integrated into computational models for simulation and decision-making, with a focus on transportation and optimization research. While LLMs offer notable flexibility in processing diverse prompts, their “black-box” behavior, opaque training data, and reliance on restricted commercial APIs pose challenges for transparent, reproducible research.
In this project, we investigate locally deployable, open-weight models—including Meta’s Llama, Alibaba’s Qwen, OpenAI’s gpt-oss, and Google’s Gemma. We aim to assess their value compared to traditional demand and decision modeling approaches. The project aims to establish reproducible, methodologically sound workflows for incorporating local LLMs into scientific research, ultimately enhancing both the transparency and cost-effectiveness of AI-assisted decision modeling in transportation systems.
During this project we gain experience in the use of local LLMs. The goal is to share this experience with colleagues who are for reasonable, compliant and controlled ways to utilize AI in other fields of research and application areas. This way, the project has impact in transportation research and beyond.
