Differential effects of feedback in computer-based instruction

A control theory model that describes response and feedback cycles was used to test learners\u27 use of feedback in computer-based instruction. The research was designed to show (1) how learners\u27 judgments moderate their use of feedback, and (2) the effectiveness of feedback in correcting errors. The model includes a function for calculating the discrepancy created when the learners\u27 actual performance does not meet their expectations. The calculation is based on the magnitude of the learner\u27s judgment of response certitude (i.e., the degree of certainty a learner associates with a response). The model predicts that feedback is used most with high-certitude errors and least with high-certitude correct responses. Two hundred and twelve university students were randomly assigned to three feedback conditions (elaborate, simple, or information-seeking) in verbal information and intellectual skills tasks, implemented in an interactive, multimedia CBI application. ANOVA was used to analyze interactions between self-efficacy, response certitude, and discrepancies on information seeking, error correction, and reactions. The results did not fully support the three-cycle model. Students did not select the information-seeking option in the pattern for feedback use as predicted. Students did not seek information very frequently, Mdn = 1, M 2.15, SD = 4.73, for the 28 lesson items. The interaction between self-efficacy and discrepancy was not significant for information seeking or on error correction. A chi square test showed a significant relationship between self-efficacy and response certitude; students with higher self-efficacy had higher response certitude. Students distinguished between the constructs at the task- and item-level, e.g., students with high self-efficacy identified low as well as high certitude responses. Students improved their performance by almost 22% from pre- to posttest, and corrected 49% of their pretest errors. The error correction rate on verbal information items was 77%, whereas the rate for intellectual skills items was only 41%. The results suggest that the model should be expanded to include other factors that affect students\u27 decision to use feedback, especially for higher-level learning (intellectual skills) tasks