Tomasz Gross, Julia Szefler
The rapid development of generative artificial intelligence is having a significant impact on contemporary practices in information architecture, UX design and data visualisation. Currently, AI models are used not only to generate visual content, but also to interpret and evaluate it, as well as to support design processes. However, a fundamental question arises: do these systems remain neutral towards the analysed materials, or do they reproduce specific aesthetic, cultural and cognitive patterns derived from the training data?
The aim of the research was to conduct a comparative analysis of popular AI models: ChatGPT, Google Gemini and Groka, with regard to the presence of biases in the processes of generating and interpreting visualisations related to information architecture and UX. As part of the study, the models were given standardised prompts regarding the design of interfaces, dashboards and visual communication. The generated materials were then subjected to comparative analysis and evaluated by all the models under study according to a common set of criteria.
The study is exploratory in nature and constitutes an attempt at a critical analysis of the role of artificial intelligence in the processes of generating and interpreting visualisations. Particular attention was paid to issues of visual perception and the interpretative patterns reproduced by AI models. The research findings may contribute to a better understanding of the limitations of contemporary generative systems and their potential impact on the future of information design and user experience.