Evaluating the effects of adaptively presenting worked examples, erroneous examples and problem solving in a constraint-based tutor.

Learning from Problem Solving (PS), Worked Examples (WE) and Erroneous Examples (ErrEx) have all been proven to be effective learning strategies in Intelligent Tutoring Systems. A worked example consists of a problem statement, its solution, and additional explanations, and therefore provides a high level of assistance to students. Many studies have shown the benefits of learning from WEs and PS in ITSs. An erroneous example (ErrEx) presents an incorrect solution and requires students to find and correct errors, therefore helping the student to solve problems. Erroneous examples may also help students become better at evaluating problem solutions. In this project, we aim to investigate how to maximize learning by adaptively providing learning activities for students based on their performance in the domain of Structured Query Language (SQL). The project was conducted in the context of SQL-Tutor, which is a constraint-based tutor that teaches SQL. A series of studies conducted during the project produced promising results. Our first study demonstrated that a fixed sequence of WE/PS pairs and ErrEx/PS pairs (WPEP) resulted in improved problem solving and that it also benefitted students with different levels of prior SQL knowledge. We then introduced an adaptive strategy in the second study, which decided what learning activities (WE, ErrEx with one or two errors, or PS) to provide to the student based on his/her performance on problem solving. We found that students who studied with the adaptive strategy improved their post-test scores on conceptual, procedural, and debugging questions (i.e., analyzing the solution, explaining the errors, and then making appropriate corrections) with significantly fewer learning activities. The final study compared the enhanced adaptive strategy to the self-selection strategy, as well as compared the enhanced adaptive strategy to the original adaptive strategy from the second study. The results show that the enhanced adaptive strategy is superior to the self-selection strategy. However, the original adaptive strategy was the better choice compared to the enhanced adaptive strategy, for students with varying levels of prior knowledge