EEG reads emotions triggered by AI-generated biodigital architecture

A new study published in the SIGraDi 2024 proceedings and posted on arXiv (2607.24808) demonstrates that electroencephalography (EEG) can reliably decode emotional responses to AI-generated biodigital architecture. Researchers Hongye Yang and Eva Guttmann-Flury first conducted a large-scale pre-experiment with 336 participants who rated 600 AI-generated images. They narrowed the set to 60 images that consistently elicited strong emotions—specifically awe, disgust, or contentment.

In the main experiment, 52 volunteers viewed these curated images while their brain activity was recorded. The team analyzed multiple frequency bands and found that gamma and delta waves provided the highest classification accuracy. Gamma band reached 77.07% ± 13.8% accuracy for identifying awe. The analysis further revealed that design elements such as greenery (plants, natural vegetation) and non-uniform granularity (varied surface textures) were strongly linked to positive emotions, while signs of dampness reliably triggered negative responses.

The results have significant implications for architecture and human-computer interaction. By proving that EEG can objectively measure aesthetic preferences toward biodigital designs, the study offers architects a data-driven tool to optimize buildings for emotional well-being. The authors emphasize that incorporating natural elements and varied textures can enhance both appeal and environmental sustainability of future structures.