The research develops novel methods for utilizing physiological signals from the user to build user models and adapt human-information interaction. Previous research has utilized mainly behavioral signals for learning user models, such as clicks, dwell time, or gaze focus, in order to improve various system functions. The proposed project aims at a fundamental breakthrough in this area by utilizing truly implicit physiological signals and modeling these signals unsupervisedly on an uncontrolled natural stimuli when users are perceiving information. This establishes scientific foundations for implicit physiological human-information interaction; jointly modeling human's brain responses, information and adapting the information by generating subsequently improved new information anticipating the user's needs. Moreover, the project aims at demonstrating applications of the models in brain-adaptive recommender systems and image generation.

2019

2024