The Iconicity of the Generated Image

2025-09-19

AI Summary

This study investigates whether visual generative AI models can learn and reproduce the semantic and formal features of iconic images—widely recognized cultural symbols—as humans do. Method: We systematically evaluate the reconstruction capability of mainstream diffusion models on 128 iconic images using data attribution analysis, cross-modal semantic similarity metrics, and a three-stage user study. Contribution/Results: Despite abundant related training data, model outputs exhibit significantly lower structural fidelity and public recognizability than human expectations. Data attribution reveals minimal influence of iconic images on generation, indicating failure to establish stable semantic anchors. This work provides the first empirical evidence of a fundamental limitation in generative models’ capacity to encode “cultural memory.” It introduces a novel methodology for assessing AI’s visual cultural understanding and establishes a benchmark dataset for evaluating cultural semantics in generative vision models.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Symbolic Representations
• Natural Language Processing: Generation

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

How humans interpret and produce images is influenced by the images we have been exposed to. Similarly, visual generative AI models are exposed to many training images and learn to generate new images based on this. Given the importance of iconic images in human visual communication, as they are widely seen, reproduced, and used as inspiration, we may expect that they may similarly have a proportionally large influence within the generative AI process. In this work we explore this question through a three-part analysis, involving data attribution, semantic similarity analysis, and a user-study. Our findings indicate that iconic images do not have an obvious influence on the generative process, and that for many icons it is challenging to reproduce an image which resembles it closely. This highlights an important difference in how humans and visual generative AI models draw on and learn from prior visual communication.

Problem

• Investigating iconic images' influence on AI generation process
• Comparing human and AI learning from visual communication
• Assessing reproducibility of iconic images by generative models

Innovation

• Data attribution analysis for image influence
• Semantic similarity analysis of generated images
• User study comparing human and AI interpretation