Seeing Candidates at Scale: Multimodal LLMs for Visual Political Communication on Instagram

arXiv cs.CV / 4/22/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The study evaluates traditional computer vision models versus a multimodal LLM (GPT-4o) for Visual Political Communication (VPC) analysis on Instagram during the 2021 German federal election campaign.
It focuses on practical tasks such as identifying prominent (“front-runner”) politicians and counting people in Instagram stories and posts.
GPT-4o significantly outperforms the compared vision systems, reaching a macro F1-score of 0.89 for face recognition and 0.86 for person counting in stories.
The results suggest that multimodal LLMs can better scale and improve visual content analysis for political communication, while also pointing to methodological issues for future work.

Abstract

This paper presents a computational case study that evaluates the capabilities of specialized machine learning models and emerging multimodal large language models for Visual Political Communication (VPC) analysis. Focusing on concentrated visibility in Instagram stories and posts during the 2021 German federal election campaign, we compare the performance of traditional computer vision models (FaceNet512, RetinaFace, Google Cloud Vision) with a multimodal large language model (GPT-4o) in identifying front-runner politicians and counting individuals in images. GPT-4o outperformed the other models, achieving a macro F1-score of 0.89 for face recognition and 0.86 for person counting in stories. These findings demonstrate the potential of advanced AI systems to scale and refine visual content analysis in political communication while highlighting methodological considerations for future research.