Analyzing Persona Effects in Generated Explanations from Multimodal LLM Agents in Urban Perception

2026-05-27

AI Summary

This study investigates how role prompting influences the interpretability of multimodal large language models (MLLMs) in urban perception tasks. By constructing 1,200 agent-based prompts with distinct personas alongside role-free baselines, the authors generate scene descriptions, explanatory rationales, and perceptual labels for urban scenes, supported by systematic analysis of 59,808 human-annotated responses. The work presents the first quantitative evidence that while scene descriptions remain highly consistent across roles, explanatory rationales exhibit significant divergence—particularly in sensitivity to socioeconomic and political attributes. Although perceptual labels show no statistically significant differences, discernible trends emerge. Integrating topic modeling with statistical significance testing, the study demonstrates that role-based prompting steers MLLMs toward distinct evaluative dimensions, thereby introducing a novel mechanism for controllable multimodal explanation generation.

Abstract

We study how persona prompting shapes language generated by multimodal large language models in an urban perception setting. Using 59,808 annotations from 1,200 persona-conditioned agents and two no-persona settings, we analyze captions, justifications, and perception tags across personas. Results indicate strong convergence in captions for different personas, whereas justifications display systematic variation associated with socioeconomic and political attributes, while perception tags show no statistically significant persona-related differences, though effect trends are observed. Topic analysis further reveals that personas emphasize different evaluative themes when interpreting the same scenes.

Problem

• persona effects
• multimodal LLM
• urban perception
• generated explanations
• socioeconomic attributes

Innovation

• persona prompting
• multimodal LLM
• urban perception
• explanation generation
• systematic variation