Abstract:

Learning analytics has been actively pursued in the field of learning assistance. The spread of learning management systems (LMSs) such as Moodle and Canvas LMS has realized the acquisition of a large amount of learning history. Furthermore, an educational assistance system does not consist of only a single LMS. Multiple educational assistance services generate their own learning histories.

A Learning Record Store (LRS) exists to collect and analyze these learning logs in an integrated manner. OpenLRW and Learning Locker exist as LRS implementations. In order to avoid losing the large number of learning records generated by LMSs, key value stores (KVS) such as MongoDB are generally used for data persistence in LRSs. For architectural reasons, KVS allows to duplicate registration of stored data. Therefore, unlike SQL-based databases, KVS persistence does not guarantee consistency. If consistency is not maintained in the persistence of learning history, it is difficult to ensure the reliability of the stored learning history. If the stored learning history is not reliable, the reliability of the analysis results using learning analytics cannot be ensured.

These LRSs are operated by public cloud services such as Amazon Web Services (AWS) and Google Cloud (GCP), etc. A region in a public cloud such as AWS generally consists of multiple availability zones (AZs). However, in order to minimize latency, it is important to minimize the number of AZs. However, from the perspective of minimizing latency, it is necessary to minimize the area where all AZs are located within a region. All AZs are located within 100 km.

Large-scale conflicts and changes in the global environment can disrupt the stable operation of systems, whether public cloud services or on-premises, and cause loss of learning logs. the massive Russian invasion of Ukraine, which began on February 24, 2022, is still ongoing two years and six months later. The Copernicus Climate Change Service (C3S) announced that the global average temperature on July 22, 2024 was the highest ever recorded. Heavy rainfall disasters and large-scale flooding have been reported not only in Japan but also around the world. A system configuration that relies on a single region of public cloud services is not sufficient against large-scale conflicts and intensification of disasters. We must ensure that the learning history of learners, which cannot be recovered once it is lost, is stored and maintained even in multi-hazard situations.

In this study, we implemented a learning history retention framework using blockchain technology to ensure the reliability of learning history. To store and maintain the learning history, we applied a decentralized and autonomous blockchain technology and attempted to detect inconsistencies in the learning history in the LRS. We implemented a framework that guarantees the availability of the learning history against factors that can cause the loss of the learning history, such as man-made or natural disasters. This paper outlines the implemented learning history retention framework and describes the mechanism of learning history retention in conjunction with an existing LMS. We also describe the implementation and demonstration of a prototype system implemented to verify the effectiveness of the framework, and clarify its effectiveness.

Keywords:

e-learning, learning record, blockchain infrastructure, disaster recovery.