Repairing Errors With Elaborative Feedback in Computerised Learning Environments

ABSTRACT Background Although automatic elaborative feedback (EF) is effective for teaching conceptual learning in science, there is insufficient evidence on how to adapt it in computer‐based question‐answering activities. Objectives This study aims to examine how we can make automatic EF more effective and tailored according to the knowledge revision process proposed in studies with refutative texts. Methods Students were required to read a science text and then answer a series of inferential multiple‐choice questions. After each answer, students received corrective feedback (right/wrong) plus automatic EF, according to their experimental condition, and then had a second attempt to answer. Three types of EFs were compared: one focused on elaborating the correct answer (EF Explicative ), another focused on correcting incorrect ideas (EF Refutative ), and another contained a neutral message (NF Control ). Two studies were conducted, one without text access while responding after EF, and the other with access to the text. Results and Conclusions The results of both studies show that EF Explicative is more difficult to process than EF Refutative , although the effects on performance on a second response attempt varied between studies. When the text was unavailable, EF Refutative produced a significantly higher proportion of correct responses than EF Explicative , and both groups performed better than NF Control . Nevertheless, when the text was available, these results were partially attenuated. After discovering errors in their learning process, learners tend to initiate a revision of their knowledge. Feedback that is congruent with this revision process was found to increase efficiency.