AI's Blind Spots: Geographic Knowledge and Diversity Deficit in Generated Urban Scenario

2025-06-20

AI Summary

This work identifies significant geographical representation biases in image generation models (FLUX 1, Stable Diffusion 3.5) when synthesizing urban scenes: given generic prompts such as “United States,” models exhibit strong urban centrism—systematically overrepresenting major cities while underrepresenting rural states and small towns—and suffer from toponymic ambiguity, misinterpreting European-style place names (e.g., “Frankfort”) as non-U.S. entities. To quantify these biases, we propose the first evaluation framework combining DINO-v2 ViT-S/14 embeddings with Fréchet Inception Distance (FID), enabling controlled, multi-model, multi-location generation experiments. Results confirm that while models implicitly capture partial geographical structure, they consistently exhibit two structural blind spots: urban-centric bias and referential confusion in place-name interpretation. This study establishes a novel methodology and empirical benchmark for assessing geographical fairness in generative AI.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Natural Language Processing: Generation
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Image generation models are revolutionizing many domains, and urban analysis and design is no exception. While such models are widely adopted, there is a limited literature exploring their geographic knowledge, along with the biases they embed. In this work, we generated 150 synthetic images for each state in the USA and related capitals using FLUX 1 and Stable Diffusion 3.5, two state-of-the-art models for image generation. We embed each image using DINO-v2 ViT-S/14 and the Fr'echet Inception Distances to measure the similarity between the generated images. We found that while these models have implicitly learned aspects of USA geography, if we prompt the models to generate an image for"United States"instead of specific cities or states, the models exhibit a strong representative bias toward metropolis-like areas, excluding rural states and smaller cities. {color{black} In addition, we found that models systematically exhibit some entity-disambiguation issues with European-sounding names like Frankfort or Devon.

Problem

• Assessing geographic knowledge bias in urban image generation models
• Identifying representative bias toward metropolis areas in generated images
• Examining entity-disambiguation issues with European-sounding city names

Innovation

• Used FLUX 1 and Stable Diffusion 3.5 models
• Embedded images with DINO-v2 ViT-S/14
• Measured similarity via Fréchet Inception Distances