Using Metacognitive Information and Objective Features to Predict Word Pair Learning Success

Abstract

There is a variety of metacognitive information that can be used to model learning success, such as judgements of learning (JOLs) and feeling of difficulty (FOD). The latter is not widely used, and part of this study is to collect FODs through crowed annotation and demonstrate its potential as a predictor of learning success. While objective features related to task difficulty provide valuable information for the modeling task, we show evidence that FOD can provide similar insight. We examine and compare the use of objective and subjective features as predictors for learning success in second language word pair learning. The results indicate that metacognitive information is transferable across different groups of subjects. They also show that crowed annotation is a useful method for enriching datasets with FODs and potentially other metacognitive information.