Abstract:

Learning analytics has attracted attention from many educators since its birth. In its early development stage, LA’s focus was more on large-scale, institution-level applications. Business intelligence and data mining, the two root disciplines from which LA has sprouted, provided frames of reference for the infant discipline with the macroscopic approaches. Development of prediction models for large educational institutes to detect signs of dropping out was popular research theme then. This macroscopic approach is useful for the decision makers in identifying who is to fail, and for giving preventive warnings for learners, instructors, and managements of the institute. Information adopted for the research and practice of this early stage was from between-subject variance observed from click-events in LMSs.

However, this approach may not effectively help neither the at-risk learners nor the stakeholders as to what to do next. Without detailed guides for what to do ‘when and where’ in the context of instructional events throughout the course, the at-risk learners and their instructors may not effectively and efficiently act upon the warning. Recent trend in LA is more centered on microscopic, time-series approaches to understand individual learners’ interactions with the series of instructional events, and to provide nudges for how to engage in that moment. This new approach focuses on the “process” of learning, and, thus, can provide the instructor where and when the individual learners experience difficulties, and ways to create interventions to ease them. To be able to do this, within-subject variance of individual actions and psycho-physiological indicators collected by multi-modal sensors are actively in use today.

Recent educational innovation comes from the flipped learning movement. By taking full advantages of self-paced, digitalized pre-study and interactive face-to-face main instruction, flipped learning aims to achieve instructional objectives of higher-order in efficient and effective way. The critical premise of flipped learning is to make sure learners master prerequisites of face-to-face session during the online pre-study. By allowing more study time for individuals falling short of the requirements for the face to face sessions, learners can make themselves, at least theoretically, be equipped with enough skills and knowledge for the following interactive instructional phase such as project- or problem-based learning activities. In reality, however, instructors often get into trouble when they discover large variances from the learners in the degree of readiness of the prerequisites. And without detailed information of individual learner’s learning behaviors, instructors may not be able to construct collaborative teams with maximum diversity and productivity, which are critical success factors in highly interactive team-based learning such as PBL.

Learning analytics can be a crucial solution to the issues of flipped learning in two ways, by providing online nudges to help at-risk learners to catch-up, and by providing individual learners’ behaviors and psychological characteristics to inform the instructor of face-to-face session. Related theories and hands-on research case studies with ethical and technical issues are to be discussed.

Keywords:

Learning Analytics, Flipped learning, problem-based learning, ill-structured tool-design capability.