Generative Affective Design

Architecture is a field at the crossroads of the arts and sciences and is influenced by countless factors, some of which are difficult to explicitly identify. Beyond concrete economic and technical constraints, architects must also be able to take rather abstract parameters concerning the well-being of the people experiencing a space into account. Generally, architectural perception is considered to be a subjective phenomenon and is therefore rarely analyzed and quantified. This leads architects to make decisions based on their personal experience and intuition, without any explicit evidence-based reasoning. 

However, numerous studies in the fields of environmental psychology and neuroscience have demonstrated the possibility to quantify – to some extent – the psychological and neuronal effects of the built environment on human beings. 

Furthermore, the rapid development of artificial intelligence (AI) is growing beyond purely quantitative tasks and showing some encouraging preliminary results. 

Therefore, with the recent discoveries in neuroscience and artificial intelligence, we wondered if it is possible to predict affective qualities, such as pleasingness or privacy, using Machine Learning (ML) models. We are not only interested in whether rooms can be judged qualitatively by a machine but also whether algorithms could generate architectural spaces according to specific qualitative criteria.