Traditional plagiarism detection tools, which primarily rely on direct comparisons between a user’s input and existing sources, often struggle to identify more sophisticated forms of cheating, such as extensive paraphrasing or the use of external assistance, including generative AI or other individuals.
Thus, this study aims to address academic dishonesty in coding by analyzing typing patterns and examining the differences in typing dynamics when individuals code and code trace compared to when they refer to or copy responses from ChatGPT. These differences are characterized by variations in thinking time, typing speed, and the frequency of editing actions during the programming and code tracing process.
Data Collection Process:
There are four different sessions for collecting data. In each session, participants will respond to six Python coding exercises, which are designed to invoke various cognitive load levels.
You are currently on Paraphrasing ChatGPT Session.
In this session, participants will feed each prompt to ChatGPT, then paraphrase the generated response. Participants should focus on preserving the original structure, functionality, and overall ideas of the GPT response while modifying the generated solution through changes such as renaming variables, rewording or adding comments, adjusting formatting, and making other non-functional edits.
Evaluation Criteria:
Upon submission, participant responses will be evaluated based on several criteria: