Adaptive feedback and student behaviour in computer-assisted instruction

Two packages of adaptive computer-assisted instruction (CAIs), both offering questions and subsequent feedback, were compared in terms of amount of feedback offered, learning efficiency and appreciation. Feedback was either barely more than knowledge of result ('minimal') or consisted of a complete additional learning path ('elaborate'). The CAIs differed in the way the type of feedback given was triggered.

A total of 97 Bachelor of Science students were stratified based on a pre-test before the experiment and were allocated randomly to two groups to receive either a 'programme-assessed' or a 'student-assessed' treatment. In the former, the feedback provided by the CAI (either elaborate or minimal) was completely determined by the objective correctness of the student's response. In the student-assessed treatment, elaborate feedback was provided only to students who stated explicitly that they did not know the answer. Afterwards, students completed a post-test and an appreciation questionnaire.

Both CAIs resulted in a significant learning effect, but overall the effect was significantly higher in the programme-assessed treatment. Students using the student-assessed CAI hardly ever used the 'I don't know' option and therefore received mostly minimal feedback. There was no difference in students' appreciation between treatments. An interaction between the learner's prior knowledge and treatment was found: for learners with a high level of prior knowledge the programme-assessed treatment resulted in significantly higher learning effectiveness compared with the student-assessed treatment, whereas for learners with a low level of prior knowledge there was no significant difference.

Our results demonstrate that learners achieved a significantly higher learning effectiveness in the programme-assessed treatment in which the received feedback was fully controlled by the correctness of the answer, compared with the student-assessed treatment. In the latter, students hardly ever admitted to not knowing the answer. Therefore, student-initiated use of the tool in a student-assessed CAI requires to be improved.