Trust is a crucial element in successful interactions with social robots. However, few existing methods attempt to measure trust in real time to adjust the robot’s behavior according to the user’s trust level. One of the significant challenges in this area is the scarcity of curated data sets necessary for developing machine learning models to detect trust from sensor data. In this paper, we introduce a structured approach to collecting data sets for trust assessment. We demonstrate this approach using an EEG data set from a study designed to build, break, and subsequently repair trust between the user and the robot. The data is available on Zenodo for development of real-time trust assessment models.