Introduction of the LAUREATE dataset: the Longitudinal multimodAl stUdent expeRience datasEt for AffecT and mEmory research

Abstract

Modern wearable devices enable the recording of physiological data from users in a continuous and unobtrusive manner. Using individual wearable devices to monitor students' psycho-physiological states could be helpful for personalised learning interventions to improve their understanding and memory recall of challenging topics. Working towards personalised interventions for students, we conducted a longitudinal study throughout a university semester with 44 participants (including two lecturers) in two courses. We recorded their physiological signals using a wristband device. We also administered pre-experiment, post-experiment and daily surveys, asking the students about their daily behaviour (e.g. study hours, smoking habits, physical activity, caffeine and food intake) and their perceived engagement, attention and emotions after class. Data was collected during the whole semester, totalling 26 sessions for each course, including their examinations (in-class quizzes, midterm and final exams). Hereby, we present the LAUREATE dataset, a LongitudinAl mUltimodal student expeRiencE for AffecT and mEmory dataset. The full dataset contains over 1400 hours of physiological data from 44 individuals, 3600 completed daily surveys from students, 70 hours of audiovisual recordings from the classes, and students' grades for quizzes, assignments and midterm and final exams, among others. To the best of our knowledge, this is the largest dataset for physiological data from the Empatica E4 device. We plan to make the dataset available for scientific purposes under a sharing agreement. Besides the initial intention of the dataset, i.e., to design post-class interventions to improve students' understanding and memory recall, the dataset is quite versatile and enables applications in many domains, including but not limited to the areas of affective computing, privacy, behaviour and learning performance modelling, context recognition, multimedia signal analysis, and the development of multi-modal machine learning algorithms.