A preliminary attempt of an intelligent system predicting users' correctness of notifications' sender speculation

Prior interruptibility research has focused on identifying interruptible or opportune moments for users to handle notifications. Yet, users may not want to attend to all notifications even at these moments. Research has shown that users' current practices for selective attendance are through speculating about notification sources. Yet, sometimes the above information is insufficient, making speculations difficult. This paper describes the first research attempt to examine how well a machine learning model can predict the moments when users would incorrectly speculate the sender of a notification. We built a machine learning model that can achieve an recall: 84.39%, precision: 56.78%, and F1-score of 0.68. We also show that important features for predicting these moments.