Improving feedback in elementary mathematics autograders

Abstract

As schools grow more crowded and required testing outcomes become more stringent, teachers experience increasing demands on their time. To ease this load, I designed a program that is able to give students thorough, automated feedback on their mathematics assignments. This will allow teachers to spend less time grading (and consequently more time on other activities that might better help their students) without losing any of the feedback and error correcting a human would be able to provide. Based on a number of different test cases, using a wide variety of elementary algebra problems, the program can correctly identify the lines in which errors are introduced. The program is also adept at finding the precise error as long as the student has made minimal changes per step. If multiple changes have been made, the program is forced to make its best guess at the most likely error without resorting to testing hundreds of possible combinations.