This study aims to analyze how artificial intelligence (AI)-generated spaces reproduce and transform the visual and contextual characteristics of original spaces, and to explore their educational implications. For this purpose, eight sites were selected—two urban and two natural environments from both domestic and international contexts. AI-generated images of each site were compared with their corresponding original images. Image generation was conducted using both DALL·E 3 and Midjourney, with each prompt applied three times per site to ensure consistency of results. The analysis involved coding contextual elements into four categories— social, cultural, economic, and environmental—and calculating two metrics: Context Preservation Rate (CPR) and Saturation Change Rate (SCR). The findings revealed a tendency for higher CPR and lower SCR in natural environments, and the opposite in urban environments. This suggests that AI tends to more accurately reproduce repetitive and stable visual patterns, whereas in complex socio-cultural settings, it often simplifies or omits certain elements while intensifying color saturation. These results indicate that AI-generated spaces should be viewed not as mere replicas but as ‘reconstructed spaces’ within the context of geography education. In particular, analytical activities using quantitative indicators such as CPR and SCR may be effective in fostering students’ critical spatial literacy.
Ji-Su Park (Tue,) studied this question.