USING SYMBOLIC EXECUTION TO CLASSIFY STUDENT PROGRAMS AND TO DEVELOP FAIRER GRADING CRITERIA

Li, Jingshu

Permalink

https://hdl.handle.net/2142/124884 Copy

Description

Title

USING SYMBOLIC EXECUTION TO CLASSIFY STUDENT PROGRAMS AND TO DEVELOP FAIRER GRADING CRITERIA

Author(s)

Li, Jingshu

Issue Date

2022-05-01

Keyword(s)

Symbolic execution; Computer teaching; LC-3 assembly language

Abstract

In computer science education, evaluation and feedback are important parts of the teaching process. Whether the evaluation of students’ efforts is objective and fair affects both students’ learning and their enthusiasm for the material. Programming assignments are a common form of homework in computer education. How to grade programming assignments fairly is a constant challenge for teaching staff. With limited staff time per student, it is difficult for staff to grade students fairly. In the current grading approach, due to problems such as overlapping between test cases, the grading results do not fairly reflect students’ efforts. Also, manual reading of code grading is extremely expensive due to the length and complexity of programming assignments. These are the challenges of fair scoring. In the thesis, we present a grading approach to get fair and stable grades. We start by obtaining a relatively complete and non-overlapping set of test cases, using the KLC3 [1] symbolic execution engine and LC-3 tools to realize the reduction of test cases and classification of student LC-3 codes. Based on the classification, we also use the KLC3 symbolic execution engine to verify the classification results. Furthermore, using the results of classification and reduction, we can compute the functionality scores of student programs, which grade student programs referring to the difficulty of each test case. We tested our grading approach in the ZJUI Fall 2021 ECE220 course LC-3 assignments (machine problems). For these three assignments, student programs with stable grades account for 97.6%, 94.3%, and 94.2%, respectively. Meanwhile, through our approach, 35% of the students in the second assignment received more reasonable grades than they had in the FA21 ECE220 course. We also gained some insights on improving the grading criteria and getting fair grades.

Type of Resource

text

Language

eng

Owning Collections