Intelligent Computer Feedback for Second Language Instruction

IN THE PAST FEW DECADES, COMPUTER-ASSISTED LANGUAGE Instruction (CALI), or Computer-Assisted Language Learning (CALL), has been of interest because of its potential for providing individualized instruction. Interactivity between the computer and the learner is one of the important elements in CALI. Compared to a traditional textbook or workbook, a CALI program can provide immediate feedback on the correctness of the learner's response. One obstacle, however, has been the relatively primitive state of the existing CALI systems, many of which possess only limited capabilities for analyzing learners' responses, diagnosing their errors, and providing feedback. Garrett (p. 174) addresses the problem with existing CALI programs as follows: "Most students learn no more from grammar on the computer than they do from grammar in the workbooks, because much of the commercially available offerings are of the 'wrong, try again' model, which only indicates whether student-produced bits of language match the target language model stored in computer memory (sometimes also showing the correct answer) without indicating how or why the student input does not match." If the potential of computer-assisted language instruction as individualized supervised learning is to be realized, we must develop programs that support detailed error analysis and feedback targeted to specific deficiencies in the student’s performance.