Exploring Natural and Urban Visual Typography: A Comparison of Letter Perception between AI and Human Observation

Abstract

Letters not only function as symbols in written language systems, but also have high visual potential in conveying messages and images. This study aims to explore the possibility of forming letter forms through a photographic approach to visual objects in natural and urban environments, and to compare the perception of letter forms between humans and artificial intelligence (AI). The method used is the Human-AI Comparative Analysis Framework with the stages of photo collection, annotation by humans, AI testing, and evaluation. A total of 56 photos of objects resembling letters were collected and annotated by participants from visual design backgrounds, then tested with a multimodal AI model (ChatGPT). The results showed a match rate between humans and AI of 76%, where humans tend to use an associative and contextual approach, while AI focuses on geometric structures. This difference indicates the potential for creative collaboration between humans and AI in exploring environment-based typography. This research broadens insights into visual communication design, particularly in the creation of experimental typography inspired by everyday visual elements.