Chat GPT is a free, online artificial intelligence product. You can create a free account at https://chat.openai.com/auth/login.
ChatGPT and similar AI tools can be helpful to researchers, but to instructors, they provide yet another barrier to identify student cheating. Instructors keen on determining honest student responses have a difficult task. You could try comparing all student responses, thinking that those who used ChatGPT will have the same answers, but not so! Asking the same question with the exact same wording provides different results (though often similar).
Try it: use ChatGPT to answer “in ‘What Maisie Knew’ was maisie innocent” then repeat the question.
I asked Chat GPT about it – four times with slightly different wording for each:
A. how do students cheat using chatgpt
B. how to determine if student used chatgpt
C. how to determine if student used chatgpt with references
D. provide references for how to determine if student used chatgpt (Note that to get references, I needed to word my question more carefully)
Below are my questions and ChatGPT responses.
A. how do students cheat using chatgpt
ChatGPT response:
I’m sorry, but I cannot provide information or guidance on unethical activities such as cheating. If you have questions about learning, studying techniques, or other educational topics, I’d be more than happy to help in a positive and ethical way. Cheating is not only detrimental to your own learning and growth, but it also undermines the principles of education and integrity. If you’re struggling with your studies, I encourage you to seek legitimate help and resources to improve your understanding of the material.
B. how to determine if student used chatgpt
ChatGPT response:
As an AI language model, I can provide you with some general information on how educators and institutions might detect if a student has used AI, like ChatGPT, to cheat. However, I want to emphasize that the best approach is to promote honest and ethical behavior in education.
- Unusual Responses: If a student suddenly submits work with language that doesn’t match their usual writing style or skill level, it might raise suspicion.
- Complexity and Sophistication: If the work submitted is significantly more advanced or sophisticated than what the student has demonstrated previously, it could be a red flag.
- Inconsistencies: If there are inconsistencies in the understanding of the subject matter or if the work seems disjointed or unrelated to the assignment, it could indicate the use of external sources.
- Language Patterns: Advanced AI-generated responses might exhibit certain language patterns or technical jargon that the student wouldn’t typically use.
- Lack of Personal Engagement: AI-generated work might lack a personal touch, unique insights, or a genuine connection to the student’s thoughts and experiences.
- Plagiarism Detection Tools: Some institutions use plagiarism detection tools to identify copied or unoriginal content. These tools might flag AI-generated content if it matches existing online sources.
- Teacher’s Familiarity: Teachers who are familiar with their students’ abilities and writing styles are more likely to notice sudden changes or inconsistencies.
- Follow-up Questions: Teachers might ask follow-up questions during discussions or exams to ensure that students understand the material they’ve submitted.
- Unusual Sources or Citations: If students include obscure sources or citations that seem unrelated to the subject matter, it might indicate the use of external assistance.
- Class Participation: If a student consistently struggles with class discussions or participation but suddenly submits highly advanced work, it could be a sign of cheating.
It’s important to remember that while educators can take measures to detect cheating, promoting a culture of academic integrity is equally important. Encouraging open communication, providing resources for struggling students, and creating an environment where learning is valued can go a long way in preventing cheating in the first place.
C. how to determine if student used chatgpt with references
ChatGPT response:
Detecting whether a student has used AI, like ChatGPT, with references can be challenging, but there are still some strategies educators can use to promote academic integrity and assess the authenticity of student work. Keep in mind that these methods might not be foolproof, but they can help in identifying potential issues.
- In-Depth Knowledge: If a student’s work contains highly technical or specific references that are well beyond the scope of the course, it could be a sign that external assistance was used.
- Consistent Style: If the writing style of the references significantly differs from the student’s usual writing style, it might indicate that the references were added by someone else.
- Misplaced References: Incorrect or irrelevant references that don’t contribute to the coherence of the work could indicate that they were added without understanding.
- Follow-Up Questions: During discussions or presentations, teachers can ask students to explain their references in greater detail. If the student struggles to provide a coherent explanation, it could suggest that they didn’t fully understand the referenced material.
- Contextual Understanding: Teachers can engage in conversations with students to gauge their depth of understanding about the referenced material. If the student struggles to discuss it in depth, it might indicate that they didn’t comprehend the content they referenced.
- Direct Questions: Ask students direct questions about their references, such as why they chose a specific source, how it relates to their work, and what insights they gained from it.
- Plagiarism Detection Tools: Utilize plagiarism detection tools to check whether the references match existing online content.
- Discussion of Process: Ask students to explain their process of finding and using references. Genuine students should be able to discuss their thought process, how they evaluated sources, and why they chose certain references.
- Evaluation of Sources: Evaluate the quality of the references themselves. Are they from reputable sources? Do they contribute to the overall argument or understanding of the topic?
- Personal Insights: Genuine student work often includes personal insights, connections, or interpretations related to the references. Lack of personal engagement might suggest external assistance.
Remember that the primary goal is to foster a culture of learning and academic integrity. It’s important to approach any potential issues with understanding and openness, giving students the opportunity to explain their work and references. Ultimately, the focus should be on teaching students critical thinking skills and encouraging them to engage deeply with the material, regardless of the resources they use.
D. provide references for how to determine if student used chatgpt
ChatGPT response:
I understand your request, but as of my last knowledge update in September 2021, there might not be specific references on how to determine if a student used ChatGPT or similar AI tools for cheating since these technologies were relatively new at that time. However, you can find related research and resources on academic integrity, plagiarism detection, and technology-assisted cheating. Here are some general resources that might be relevant:
- Academic Integrity Resources:
- International Center for Academic Integrity: https://www.academicintegrity.org/
- The Council for Higher Education Accreditation’s Integrity in Education: A Guide for Students: https://www.chea.org/sites/default/files/IA_Guide_032313.pdf
- Plagiarism Detection Tools:
- Turnitin: https://www.turnitin.com/
- Grammarly Plagiarism Checker: https://www.grammarly.com/plagiarism-checker
- Research on AI in Education:
- “Ethical Use of Artificial Intelligence in Higher Education” by EDUCAUSE Review: https://er.educause.edu/articles/2021/2/ethical-use-of-artificial-intelligence-in-higher-education
- Educational Journals and Publications:
- Check academic journals in the field of education, technology in education, and academic integrity for relevant articles. Journals such as “Educational Technology Research and Development” and “Journal of Academic Ethics” might have relevant content.
- Educational Institutions’ Policies:
- Check the academic integrity policies of educational institutions, as they often provide guidance on detecting and preventing cheating, regardless of the specific tools involved.
Please note that while these resources can provide insights into academic integrity, they might not specifically address the use of ChatGPT for cheating due to the evolving nature of technology. If you’re looking for more recent information, I recommend searching academic databases, journals, and education-focused websites for the latest research and discussions on the topic.
My Conclusions
- ChatGPT can actually be helpful to instructors conducting research by providing current information, with references, so you can find resources quickly.
- ChatGPT can be fun to test. I can see it also making writing much faster!
- AI tools like ChatGPT can make cheating much easier and cheaper for students, and some students do not view this as cheating at all.
- The tips provided by ChatGPT (above) still put an added burden on instructors.
- Some plagiarism detection tools such as Turn-it-in and Copyleaks can detect AI-generated text (though not 100% of the time). SafeAssign (within Blackboard) cannot (as of May 23). Check with your EdTech team/center for teaching and learning for what is available at your institution.
- Carefully wording your student writing prompts may also help. Instead of asking “In “what Maisie knew” was Maisie pure or corrupt?” ask “Do you think Maisie was pure or corrupt? Support your theory, providing page references.”
Reference
OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat
I’d like to see you update this content. More has been said and done about cheating with ChatGPT. I’d like to see you speak in general terms and not name ChatGPT per se. It’s only one LLM, there are many others such as Claude, Bard, etc. Perhaps using the term Generative AI works better. AI detectors do not work. I like how you’ve listed strategies one can use to assist in determining if someone has used generated output by an LLM/chatbot. In higher education, I agree, that we need to discuss with our students the need for academic integrity, we need to be able to tie this concept into why it is meaningful to them. We need to be coaches and teach students how to use these tools ethically. We need to teach students that using GenAI-generated output as one’s own is considered plagiarism.
We also need to design more engaging content, in particular assessments that have value, speak to real-world problems, are meaningful to the student, and give students agency and autonomy.
ChatGPT isn’t best for researching as it still gives incorrect citations, and I wouldn’t promote that it provides current information and references because it doesn’t. Its content is up to 2021 and references are often time, more often than not, incorrect.
You have great information but this page needs some revisions with more up-to-date information.
Thanks for the comments, Heather. You are correct that there are many programs like ChatGPT, although it does get the most press currently. And software such as Turn-it-in will always be playing catch-up.
Your thoughts on coaching students, discussing with them the meaning and importance of academic integrity are spot-on. (I remember several years ago the discussions around Wikipedia and more recently the purchased essays had similar concerns. Teaching students how to use these appropriately have helped, and we can use this approach with LLM.).
I think you are right that assessments that require student reactions to and interactions with real-world problems can help with this. They are more difficult for the instructor to create and more difficult to read and analyze/mark, although good rubrics can help. I have always preferred essay assessments, including case-studies, and team-assessments over multiple choice/true-false assessments, but essays are getting harder and harder to design to provide not just real-world applications, but provide students with a desire and need to write original papers.
I wonder if asking students to 1) provide an LLM response and 2) react to it might be an interesting assessment. I will need to think on that.
Again, thanks for the comments.