Skip to navigation Skip to main content



Can computers anticipate personality inferences from clothes?

Feb 22 - 18h Hannes ROSENBUSCH (UvA)

The fashion industry is rich with applications of machine learning and AI tools. However, some parts of the fashion universe appear to be intrinsically human, such as the psychological process of trying to infer personality based on clothes. Much research has shown that people try to make these inferences (although they often have limited success/accuracy). We tested whether AI models could replicate such human inferences. In other words, human personality inferences from clothes might be wrong, but they might still be predictable. Our project firstly collected the most common psychological inferences made from clothes. Next, it quantified how much the clothes steer impressions vs how much impressions are due to the onlooker's own tendencies. And lastly, it measured how accurately an AI model can predict the psychological associations to specific clothing articles from images. The newly generated database of thousands of clothing items including psychological ratings is obtainable online.

Hannes Rosenbusch

Link to paper