Addressing Misuse of Generative AI

Overview

Student Conduct and the Academic Appeals Board discourage faculty from relying exclusively on the results of an AI detection program because the accuracy of these detection programs is still in debate. The university therefore asks faculty to address any additional information that informed their opinion about a student’s submission.

A faculty member’s familiarity with the student, their analysis of the work the student submitted and the instructor’s professional experience in the classroom are all very important in the evaluation of a case. Below is a list of talking points that have been useful to faculty and the Office of Student Conduct in assessing the likelihood that a submission was AI generated versus the student’s own original composition. The Office of Student Conduct encourages faculty to consider the following in articulating their justification for their report of a suspected violation of academic integrity standards.

Questions About the Submission to Consider

• What is it about the written submission, other than an AI detection program score, that leads you to conclude that it is not the student’s own work?
• Is the work grammatically perfect, no spelling errors but substantively broad, vague and not on point, no specifics or references to class material or examples?
• Is the information factually true and valid or is there fabricated content, made up facts and sources? (eg. characters or incidents that don’t exist in the story, movie, song etc.)
• Does the suspicious submission fail to align with the students “in-class” or other written work? Is this essay written in a different “voice” than the student’s other work?
• Are the suspicious submissions different in tone, or level of sophistication and expression from the student’s previous responses?
• Are there significant differences in the vocabulary, grammar, sentence structure etc. in this submission versus the student’s previous work?
• Were the responses the student submitted similar in form and content to results generated when you submitted the same prompt to ChatGPT or other generative AI program?
• Was the student unable to explain key concepts, vocabulary, analysis or other elements of the work they submitted? (If so, that fact alone could suffice to establish our “preponderance of the evidence.”)
• Check the citations in the submission. Are they valid? false/fabricated?
• Are there changes in voices (that would suggest that the student had imbedded AI or other non-original material into something they had written in part)?
• Are there errors in reasoning and analysis?

Examples of Faculty Notes in Reports

• “The student submitted an essay with no incremental development, and it was very similar in examples and turn of phrase to two other suspicious essays, which were very similar to an essay ChatGPT generated for me in response to my exam prompt.”
• “Some student papers were discussing the same character that didn’t exist in the film.” Faculty detected a pattern of an AI generating tool generating the same misinformation/false information, phrased the same way.
• Faculty detected similar phrasing over and over again in several student papers.
• “They are using AI paraphrasing tools that paraphrase into something that doesn’t make sense.” Example: The film title, “Who Killed Vincent Chin” appeared in the student’s paper as, “Who Killed Vincent Jaw?” and the film title, “The Good Earth,” became, “The Great Soil.”
• “On exam of short answers, the student’s responses where like what I ‘d expect and have seen from students, but then the last response was a long and sweeping discussion unlike all the other answers the student provided. It stuck out as very different.”
• “I entered the question prompt into ChatGPT and got an answer nearly identical to what the student submitted, structure, format, progression of ideas, concepts etc..”
• “The student didn’t recall much about the paper and couldn’t explain terms or vocabulary used.”