The SWE colloquium talk on Thursday 10 October will be given by

  Stefan Mayer, University of Tübingen, and

  Jan Landwehr, Goethe University Frankfurt.

Capturing determinants of processing fluency using deep neural networks to predict consumer behaviour

In recent years, the processing fluency framework has been proven to be a very powerful theoretical model to understand a wide variety of phenomena, especially in social and consumer psychology. At the same time, the extraction of low-level visual information from images has made significant progress and now allows the extraction of objective measures of determinants of processing fluency such as symmetry, contrast, simplicity, or prototypicality from any digital image. In the present project, we will show how new algorithms based on deep neural networks can improve the measurement of visual similarity and visual typicality. In particular, we show that our new measure can be employed to predict liking judgments and purchasing behaviour of cars in line with the predictions of the processing fluency framework. On this basis, we discuss to what extent deep neural networks are an adequate model of human visual perception and offer the potential to improve our understanding of perceptual processes.