Artificial intelligence may affect diversity: architecture and cultural context reflected through ChatGPT, Midjourney, and Google Maps

Abstract This study aims to understand how widely used Artificial Intelligence (AI) tools reflect the cultural context through the built environment. This research explores how outputs obtained with ChatGPT-4o, Midjourney’s bot on Discord and Google Maps represent the cultural context of Stockholm, Sweden. Cultural context is important because it shapes people’s identity, behaviour, and power dynamics. AI-generated recommendations and images of Stockholm’s cultural context were compared with real photographs, GIS demographic data and socio-economic information about the city. Results show how outputs written with ChatGPT-4o mostly listed museums and other venues popular among visitors, while Midjourney’s bot mostly represented cafes, streets, and furniture, reflecting a cultural context heavily shaped by buildings, consumption and commercial interests. Google Maps shows commercial sites while also enabling users to directly add information about places, like opinions, photographs and the main features of a business. These AI perspectives on cultural context can broaden the understanding of the urban environment and facilitate a deeper insight into the prevailing ideas behind the data that train these algorithms. Results suggest that the generative AI systems analysed convey a narrow view of the cultural context, prioritising buildings and a sense of cultural context that is curated, exhibited and commercialised. Generative AI tools could jeopardise cultural diversity by prioritising some ideas and places as “cultural”, exacerbating power relationships and even aggravating segregation. Consequently, public institutions should promote further discussion and research on AI tools, and help users combine AI tools with other forms of knowledge. The providers of AI systems should ensure more inclusivity in AI training data, facilitate users’ writing of prompts and disclose the limitations of their data sources. Despite the current potential reduction of diversity of the cultural context, AI providers have a unique opportunity to produce more nuanced outputs, which promote more societal diversity and equality.