Using Auto Tutor in strategic reading

Using Auto Tutor in strategic reading

First Author: Ju-Ling Chen -- National Academy for Educational Research
Additional authors/chairs: 
Jih-Ho Cha; Hou-Chiang Tseng; Min-ying Tsai; Berlin Chen; Yao-Ting Sung
Abstract / Summary: 

This study aims to develop for elementary schoolers an autotutor learning system, called Automated Reading Elaboration System (ARES) using latent semantic analysis and deep learning approach. The ARES, as a learning companion and instructor, encourage learners to master strategic reading and enhances the exercise of these strategies. The ARES can be improved and adjusted in accordance with the empirical evidence acquired from classroom experiments. In the experiment design of this current study, the participants were 40 elementary students of 5th grade, half of whom were low reading ability and the other half were high reading ability. Students’ learning performances were observed and evaluated based on their interactions with the auto tutor, exercise of strategies, and post-tests. Their data of process suggested that most of the participants with low reading ability had lower overall accuracy rate, longer learning session, extended answering path, longer response time to each question than those with high reading ability. They also required more scaffolding, such as prompt, hint, implicit question, explicit question, cloze inference and hints for choice questions, so as to correctly answer content-related knowledge questions. Furthermore, they often gave content-unrelated non-knowledge answers. In general, to possibly engender elaboration inference, teachers or instructors should deliver them supporting knowledge, and the skills of connecting strategies to prior knowledge prior to strategic teaching when designing reading strategies for students with low reading ability.