• Fatma Yasemin Bayındır Özkan

  Author

The digital ELearning environment provides for continuous streams of behavioral and performance data, thus offering opportunities for LA to assist in detecting learner needs where traditional assessments at the end of a term come too late. This paper presents a synthesis of empirical evidence from institutional, higher education, secondary school, and MOOC settings to analyze how LA and EDM can be used to predict the performance of learners in English. The paper examines three types of prediction methods: procedural and behavioral modeling focusing on sequence and not only the frequency of behaviors; new multimodal data sources in the form of wearable sensors providing information about the learner’s cognitive load and anxiety while listening and speaking; and real-time dashboards and ranking systems that transform the predictions in classroom-facing feedback. The general LA methods are discussed with regard to specific English settings, such as the academic language gaps among college transfer students, the experiment with flipped-classroom feedback and the use of serious games to practice vocabulary and pronunciation. Based on these findings, the article suggests a theoretical model for developing LA systems for English learners consisting of four layers - data, analytics, pedagogy, and equity - stressing the need for interpretation of the predictions rather than delivering unmediated judgements to the learner. The conclusions made by the article include the following – LA technology is promising for the early identification of at-risk English learners and for customizing instructions to them. However, the value of LA is contingent upon the continued focus on transparency, learner empowerment and equity especially for multilingual and non-conventional learners for whom there is little data in the training set.

• PDF