Search Papers | Poster Sessions | All Posters
Poster B150 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Evolving synthetic images for the control of affective experience
Darius Valevicius1 (), Celine Haddad1, Vincent Taschereau-Dumouchel1; 1University of Montreal
Creating stimuli that reliably produce targeted alterations in subjective emotional experience would be a useful tool for understanding the neural correlates of conscious feelings. In this study, we develop a method to create images that maximize a target emotion (e.g. fear) by using a combination of generative artificial intelligence and an evolutionary algorithm. We show that image evolution conditioned on fear ratings can generate images that strongly drive fear responses, while conditioning image evolution on a physiological proxy of fear (i.e. electrodermal activity) does not create images rated as fearful. However, the latter still converges on generally feared insectoid categories. This method may be used in the context of decoded neurofeedback to study the contribution of different brain regions to subjective fear.
Keywords: fear physiology consciousness artificial intelligence