In this report, we investigate the value of common wrong answers. By introducing common wrong answers into model building, can we achieve better models in predicting next problem correctness? To answer our question we build two tabling models. The first model is our control model. This model makes predictions based on three types of student responses, correct responses, hint responses and wrong answer responses. The second model is our experiment model, named Interesting Common Wrong Answers Model (ICWAs Model). The ICWAs model extends upon the control model by giving each common wrong answer its own prediction. We compared the results from both models and verified that common wrong answers do not bring improvements to predicting next problem correctness.
Worcester Polytechnic Institute
Interactive Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.