Preserving Privacy in Multimodal Learning Analytics with Visual Animation of Kinematic Data

Protecting personal data in research has gained significant importance in recent years. Data-intensive research fields that require extensive data analysis, such as Multimodal Learning Analytics, rely on data collected through various sensory devices, such as wearable devices or visual sensors. Depending on the type of sensor used, the data collected may contain personal information about the users, such as full video recordings to examine their kinematic behaviour. This can become controversial due to ethical considerations and international regulations like the General Data Protection Regulation of the European Union. In this study, we propose using animations instead of videos to analyse multimodal data, such as learning analytics, in educational settings. Visual animations can preserve the participants’ privacy while maintaining the data analysis’s quality. We test the method by annotating the performance of a cardiopulmonary resuscitation (CPR) procedure using either video or animation. We find out that the agreement between the raters is similar for both methods, suggesting that animation is a feasible alternative to video.