Practical Responses to ChatGPT

“But even when the essays are a good synthesis of other essays, written by humans, they are not human. Frankly, they creep me out precisely because they are so competent and yet so very empty. ChatGPT impersonates sentiment with sophisticated word choice but still there’s no élan. The essay does not invoke curiosity or any other emotion. There is a voice, but it is mechanical. It does not incite, offend or seduce. That’s because real voice is more than grammatical patternmaking.”  Tressie McMillan Cotton, “Humans This Christmas,” NYTimes, 12.20.22.

“[…] ChatGPT feels different. Smarter. Weirder. More flexible. It can write jokes (some of which are actually funny), working computer code and college-level essays. It can also guess at medical diagnoses, create text-based Harry Potter games and explain scientific concepts at multiple levels of difficulty.” Kevin Roose, “The Brilliance and Weirdness of ChatGPT,” NYTimes, 12.05.22.

The Latest Technology: ChatGPT and other Generative AI bots

Late in 2022, OpenAI’s ChatGPT prompted a flurry of commentary from doomsday predictions to enthusiasm for creative and educational possibilities. Few instructors have not heard about the tool as stories of it abound in the popular press. It’s unclear to most readers whether or not this generation of AI has really passed the Turing test, whether we’re at a watershed moment, but instructors will benefit from giving the tool’s capacities to generate readable prose and calculate problems, among other tasks, some thought as they prepare their courses and assignments.

ChatGPT is not without precedent or competitors (such as Jasper, Sudowrite, QuillBot, Katteb, etc). Souped-up spell-checkers such as Grammarly, Hemingway, and Word and Google-doc word-processing tools precede ChatGPT and are often used by students to review and correct their writing. Like spellcheck, these tools are useful, addressing spelling, usage, and grammar problems, and some compositional stylistic issues (like overreliance on passive voice). However, they can also be misused when writers accept suggestions quickly and thus run the danger of accepting a poor suggestion. Automation bias is in effect — we often trust an automated suggestion more than we trust ourselves. Further, over-reliance can mean students simply miss opportunities to grow and develop as writers.

What is ChatGPT?

Practical Suggestions to Mitigate Non-Learning/Cheating

Course Design and Pedagogy

  • Talk to students about your expectations for academic honesty, and remind them of the University’s policies. Tell them that using ChatGPT is academically dishonest, akin to paying a person to write your paper, take your test, or complete your assignment. Give examples and be specific and frank about your concerns. For most faculty, academic dishonesty is galling because honest students are disadvantaged, but also because it is an offense to our commitment and passion for learning. What’s the point of pursuing a degree, of taking a class, if you don’t learn? Raise these questions to stimulate reflection. Most students who engage in academic dishonesty are doing so impulsively or without reflection. Anticipate this human behavior and engage students in an open discussion about academic dishonesty. See Academic Dishonesty and Student Cheating for additional guidance.
    • Add a clarifying statement, such as, “Use of an AI text generator when an assignment does not explicitly call or allow for it without proper attribution or authorization is plagiarism.”
  • Provide incentives for the behaviors and habits that are associated with strong learning — for trying — as well as producing. If a perfect product (test, paper) is the only way to receive an A, students are more likely to consider cheating. In your effort to fully assess student learning, make sure you assess (and reward) the processes that are needed to be a strong learner in your course: reading, viewing, speaking, improving, reflecting on one’s learning, etc. Review your grading criteria and rubrics to make sure you’re setting your students up to adopt strong learning strategies. See Grading for Learning under Plan for Grading.
  • Run your assignment through ChatGPT, review that answer, and tell your students about your experience (and that you’ve saved the output). Note: ChatGPT does not produce the same answer each time its posed.

Assignment Design

ChatGPT and similar tools rely on available text — text that is not behind firewalls or so obscure or recent that it has not yet been added to the tool’s corpus. These suggestions are built on this design characteristic.

  • In written assignments, reference class materials and notes, or sources that are not available on the free internet (books or articles that are recent, behind firewalls). For example, “Refer to two of the theorists discussed in class.”
  • Include visuals — images or videos that students need to respond to — in your assignment. Be sure to include alt-text for accessibility.
  • Reference or connect to current events or conversations in your field.
  • Ask for application or engagement between personal knowledge/experience and course concepts or topics.
  • For short reading responses, instead of using open-ended questions in Canvas, try social annotation tools that require students to engage with a text along with their classmates. Try Hypothes.is or Perusall, both of which are supported by the University.
  • Replace an essay or short-answer writing assignment with one that requires students to submit an audio file, podcast, video, speech, drawing, diagram, or multimedia project. That is, mix up the assignment in ways that make running to ChatGPT more work than it’s worth.
  • Chunk your written assignments with due dates for individual elements that precede the final submission: an outline, notes on research articles, drafts.

Finally, incorporate ChatGPT in your assignments. For example, ask students who choose to open an account to generate a ChatGPT response to a question of their own choosing, and then write an analysis of the strengths and weaknesses of the ChatGPT response. ChatGPT is interesting! Engage with the tool and discover with your students what it can and cannot do.

Note: Individuals may have reasonable concerns about privacy — ChatGPT acknowledges that they may share account holders’ personal information with third parties, including vendors and service providers. See Privacy Policy.

Extend Flipped Learning: Class Time Can Be for Writing / Creating

  • Ask students to read, view, and digest material at home, and then apply, demonstrate, and perform in class.  For example:
    • Have students write responses in class. If students have 20 minutes to write brief responses to the kinds of questions you might have provided as homework, they will learn a great deal, and as a bonus, your subsequent class discussion will benefit from that engaged individual work.
    • Have students respond orally, requiring each student to respond to a different question.
    • Have students work in small groups in class to present on topics in class.
    • Incorporate brief in-class quizzes, tests and other assessments. The key is to make these short, frequent, and possibly even unannounced. They serve assessment purposes, reward attendance, and provide useful immediate feedback about learning. Small point values for individual assessments allow poor performance to be informative to students rather than disastrous.
  • Learn more about Flipped Learning

When All Else Fails

  • Require handwritten responses. Students will groan, and you may too as you attempt to read student handwriting again, but not only will this deter the use of ChatGPT, but research shows that we actually remember better when we write by hand.
  • Be on the lookout for AI-produced texts. Foreign language instructors have been coping with the problem of Google Translate for years, and so perhaps must the rest of us learn to cope with AI-produced texts. Play around with the tool and get an idea of what kind of prose is produced to the questions you typically ask. Not only will you gain insights on how to better write your assignments, but you may get a sense of the “voice” — or lack of voice — of the tools. Some users of ChatGPT describe ChatGPT text as follows:
    • It’s atypically correct in grammar, usage, and editing.
    • It’s voiceless — correct and easy to read, but without any sense of a human person — fallible, uneven, passionate, awkward — present.
    • It follows predictable formations: strong topic sentences at the top of paragraphs; summary sentences at the end of paragraphs; even treatment of topics that reads a bit like patter: On the one hand, many people believe X is terrible; on the other hand, many people believe X is wonderful.”
    • Note: ChatGPT can be instructed to take on a voice. For example, write like a senior in high school, or write like a marketing executive working in financial services.
  • Take advantage of tools for detecting AI. There are several out there, including openai-detector, found on HuggingFace, a platform for natural language processing based on machine learning. The tool is simple. You drop text you find and suspect is AI generated into a textbox and receive a probability reading. Remember that no tool is perfect – and it can take a few minutes to run. Others include Gltr.io and GPTZero, among others. Turnitin and OpenAI, the company that owns ChatGPT, are promising tools that can detect AI.
Resources and References
  • Alby, C. (2022, December 17). ChatGPT: Understanding the new landscape and short-term solutions. Google Docs. Retrieved January 4, 2023, from https://docs.google.com/document/d/1ERCgdylG2LyOeL93aWrK6Jf97N_m1qaueN9W4kzO0Rk/edit#.
  • Barbaro, M. (Host). (2022, December 16). Did artificial intelligence just get too smart? [Audio podcast episode]. In The Daily. The New York Times Company. Guest: Kevin Roose.
  • Graham, S. S. (2022, October 24). AI-generated essays are nothing to worry about. Inside Higher Ed. Graham had students in his Fall 2022 undergraduate Rhetoric and Algorithms class use AI to write an essay based on a prompt he provided. The results were not good, despite encouraging students to edit the papers before submitting them.
  • Jarry, J. (2022, December 16). I chatted with an artificial intelligence about quackery. McGill University Office for Science and Society. Jarry tests out ChatGPT to see how it handles answering questions about unproven medical interventions.
  • Marchese, D. (2022, December 26). An A.I. pioneer on what we should really fear. The New York Times. Retrieved January 4, 2023, from https://www.nytimes.com/interactive/2022/12/26/magazine/yejin-choi-interview.html?smid=url-share. Interview with computer scientist Yejin Choi.
  • Mondschein, K. (2023, January 3). Avoiding cheating by AI: lessons from medieval history. Medievalists.net. Retrieved January 4, 2023, from https://www.medievalists.net/2023/01/chatgpt-medieval-history/. Medievalist Ken Mondschein experiments with ChatGPT and observes strengths, but also clear weaknesses.
  • Schmitz, R. (Host). (2022, December 16). Has AI reached the point where a software program can do better work than you? [Audio podcast episode]. In: Morning Edition. National Public Radio. Guest: Ethan Mollick of the University of Pennsylvania. Mollick gives an overview and specific examples of how ChatGPT can be used. For example, he tried it out by asking the program to write a syllabus for him, along with a lecture and a final assignment with grading rubric.

For additional resources, view the ongoing Google Doc, AI Writing and Creating Bots (Montclair netID required).

For more information or help, please email the Office for Faculty Excellence or make an appointment with a consultant.

Updated 1.20.23 EJI

Creative Commons License
Teaching Resources by Montclair State University Office for Faculty Excellence is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License

Third-party content is not covered under the Creative Commons license and may be subject to additional intellectual property notices, information, or restrictions. You are solely responsible for obtaining permission to use third-party content or determining whether your use is fair use and for responding to any claims that may arise.

Creative Commons CC BY-NC-4.0