Active Academic Integrity

Alex Kassil

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2021-157
May 21, 2021

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-157.pdf

The response to academic misconduct in introductory computer science courses can potentially affect the behavior and culture of students throughout their time as undergraduates. Moss, the most commonly used system for detecting excessive collaboration, is not well suited to assignments whose solutions are only a few lines long. This paper describes an active approach to detecting the use of solutions from prior semesters in which instructors change the names of various identifiers, change constants, and change logic used in the problem. We developed a tool that searches student submission history to detect the use of identifiers, constants, and logic from prior semesters. As with TMOSS, the misconduct is detected across the entire history of partially completed student work, not just their final submission. Our developed tool was first used in CS 61A which had a Spring 2020 enrollment of around 1,800 students. There were 164 students notified and 142 penalties enforced (86.6% of notified students) for the first homework. Over the Spring 2020 semester, there were 452 enforced cases of misconduct on homework, labs, and projects out of 598 accusations (75.6% of cases). In Summer 2020, improvements to the tool sped up response time and reduced the false positive rate from 24.4% to 21.9%. In Fall 2020, the false positive rate was reduced to 20.5%. In Spring 2021, an automated early warning system further reduced the false positive rate to 6%. The tool has also been adapted for use by another similar course, CS 88. The paper finally discusses what was learned using this tool and communicating with hundreds of students about academic integrity, as well as best practices for instructors so students are caught breaking the rules before they develop bad habits that hinder their learning and lead to persistent cheating.

Advisor: John DeNero


BibTeX citation:

@mastersthesis{Kassil:EECS-2021-157,
    Author = {Kassil, Alex},
    Title = {Active Academic Integrity},
    School = {EECS Department, University of California, Berkeley},
    Year = {2021},
    Month = {May},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-157.html},
    Number = {UCB/EECS-2021-157},
    Abstract = {The response to academic misconduct in introductory computer science courses can potentially affect the behavior and culture of students throughout their time as undergraduates. Moss, the most commonly used system for detecting excessive collaboration, is not well suited to assignments whose solutions are only a few lines long. This paper describes an active approach to detecting the use of solutions from prior semesters in which instructors change the names of various identifiers, change constants, and change logic used in the problem. We developed a tool that searches student submission history to detect the use of identifiers, constants, and logic from prior semesters. As with TMOSS, the misconduct is detected across the entire history of partially completed student work, not just their final submission. Our developed tool was first used in CS 61A which had a Spring 2020 enrollment of around 1,800 students. There were 164 students notified and 142 penalties enforced (86.6% of notified students) for the first homework. Over the Spring 2020 semester, there were 452 enforced cases of misconduct on homework, labs, and projects out of 598 accusations (75.6% of cases). In Summer 2020, improvements to the tool sped up response time and reduced the false positive rate from 24.4% to 21.9%. In Fall 2020, the false positive rate was reduced to 20.5%. In Spring 2021, an automated early warning system further reduced the false positive rate to 6%. The tool has also been adapted for use by another similar course, CS 88. The paper finally discusses what was learned using this tool and communicating with hundreds of students about academic integrity, as well as best practices for instructors so students are caught breaking the rules before they develop bad habits that hinder their learning and lead to persistent cheating.}
}

EndNote citation:

%0 Thesis
%A Kassil, Alex
%T Active Academic Integrity
%I EECS Department, University of California, Berkeley
%D 2021
%8 May 21
%@ UCB/EECS-2021-157
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-157.html
%F Kassil:EECS-2021-157