TLDR: A pilot study surveyed 120 Computer Science students to assess the prevalence of AI-assisted cheating in programming courses. Over 25% of respondents admitted to using AI tools like ChatGPT for assignments when not allowed, primarily for bug fixing and code structure. While anonymous surveys proved effective for data collection, interviews were not. The study highlights the significant presence of AI plagiarism and suggests educators adopt transparent policies and creative assignments rather than outright bans.
The rise of artificial intelligence (AI) tools capable of generating computer code, such as ChatGPT, presents a significant challenge to traditional Computer Science education. These tools allow students to complete assignments with minimal effort, raising concerns about a new form of plagiarism. A recent pilot study aimed to understand the extent of this issue and effective ways to measure it.
Conducted by Kaléu Delphino, a researcher at Georgia Tech, the study focused on assessing the feasibility of estimating AI plagiarism through anonymous surveys and interviews. The research involved a large Computer Science class with approximately 120 students from the Online Master of Science in Computer Science (OMSCS) program at Georgia Tech. The goal was to determine if students would admit to AI-assisted cheating in a confidential setting and to explore the best methods for gathering such sensitive data.
The methodology involved distributing a short, self-administered, and anonymous survey. To encourage honest responses, the survey avoided direct “yes/no” questions about cheating, instead asking about the frequency of AI tool use for assignments when not permitted. It also included questions about knowing other students who had cheated with AI, aiming to reduce discomfort. Following the survey, students who admitted to cheating were invited for a short, voice-only interview to gather more qualitative insights.
The survey yielded 66 responses, with 49 deemed valid for analysis, resulting in a response rate of over 50%. The findings were quite revealing: 22% of respondents reported knowing another student who had committed AI plagiarism. More strikingly, 28% of all surveyed students admitted to using AI tools for programming assignments when it was not allowed. This figure rose to 35% among those who had previously used AI tools for code generation. Despite the sensitive nature of the questions, 66% of respondents felt neutral to comfortable taking the survey, suggesting that anonymous surveys are an effective method for this type of research.
Conversely, the interview component faced significant challenges. Out of 15 students given the opportunity to be interviewed, only one ultimately participated. This indicates that while students might be willing to admit to misconduct in anonymous surveys, they are much less inclined to discuss it in a more direct, even if confidential, interview setting. The single interviewee provided valuable insights, describing ChatGPT as a “good teaching assistant” for understanding concepts and handling “busy work” like code structure and bug fixing. They noted the tool’s limitations for larger projects and the need to validate its output. The interviewee also highlighted the absence of clear AI usage policies as a reason for not disclosing their use and emphasized the importance of transparent guidelines from educators.
The study concludes that anonymous surveys are a viable and effective method for assessing the prevalence of AI-assisted cheating in programming courses. The high rate of admitted misconduct (over one-quarter of students) suggests that AI plagiarism is a significant and widespread issue, comparable to traditional forms of plagiarism observed in previous studies. Students primarily use AI tools for debugging and obtaining code structure, often to meet specific assignment requirements or to streamline what they perceive as “busy work.”
Also Read:
- Classroom Insights: Integrating Generative AI in BIM Education
- Academic Integrity Under Scrutiny: The Growing Trend of AI Use Among Students
Regarding countermeasures, students generally oppose outright bans on AI tools. Instead, they favor adaptive approaches from educators, such as designing more creative assignments that are less susceptible to AI generation. However, measures like increasing the weight of exam scores or reintroducing paper exams were not considered fair options by the students. This research underscores the urgent need for educators to develop clear, transparent policies and adapt their teaching strategies to the reality of AI tools in the classroom, rather than attempting to ignore or strictly prohibit their use. For more details, you can refer to the full research paper: Assessing the Prevalence of AI-assisted Cheating in Programming Courses: A Pilot Study.
