Practical Responses to ChatGPT and Other Generative AI

“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, its rivals, and its cousins (e.g., Google’s Bard) 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 thought to the tool’s capacities to generate readable prose and calculate problems, among other tasks.

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?

• The response received from ChatGPT on 1/3/2023: “CHATGPT is a variant of the GPT (Generative Pre-training Transformer) language model, which was developed by OpenAI. GPT models are trained to generate human-like text by predicting the next word in a sequence based on the words that come before it. CHATGPT is specifically designed to be used in chatbots and conversational systems, and it is trained on a large dataset of human conversations to learn how to generate appropriate responses in a variety of contexts.”
• Youtube video from How? (9 minutes)
• Read about it from OpenAI.
• Sign up for a free account and experiment for yourself.https://chat.openai.com/auth/login.
• Learn about ChatGPTs’s visual creation sister, Dall-E 2.
• ChatGPT Experiments with Sample Assignments — read through OFE faculty and students’ experiments with ChatGPT — query, output, detection by AI-checkers, and notations about output. (Google doc; netID required).
• Generative AI — OFE Resource and Reading List 
• AI for Teaching and Learning — OFE-run group where instructors exchange ideas and experiences.

  • To join the group, go to the link and click “Ask to Join the group.” 3) Select settings and again click “Ask to Join.” Moderator will approve you shortly. For questions, reach out to Vera Senina at seninav@montclair.edu

Implications for Teaching

Generative AI can be a tool for short-changing learning and getting a passable grade on an assignment. So can Wikipedia, pasting in answers from a digital text, and copying from a peer. Generative AI is also a tool for learning, as anyone who gets into the tool and starts inputting queries and studying the output knows. Your synapses are firing as you write and read the rapidly generated text. It’s fun and you’re likely wide awake, judging, speculating, disagreeing, agreeing, and doing all those things that happen when an engaged reader encounters text.

Generative AI pushes us to focus on the challenges of teaching.

“Teachers teach someone something, in that order.” Samuel Natale

Natale reminds us that our enduring value as instructors is in teaching students how to learn — how to ask and answer questions. How best to learn in our disciplines and sub-disciplines varies, and thus students need to continue to develop and hone their learning skills.

With every new group of students, we thus need to foster self-reflection on the work of learning in our course. Through exam wrappers, reflection questions, in-class quick writes, think-pair-share activities, and large class discussions, we help students come to know their own best methods for learning. With Generative AI the lure of skipping by earning may be greater, so we may need to double down on our attention to attending to students’ growth as learners.  Some specific suggestions:

• Assignment Strategies: Authentic assessment, formative assessment, assessment add-ons like problem-solving logs, exam wrappers, minute papers, “muddiest point” questions.
• Foster Self-Reflection and Planning: Make time for reflection and sharing about learning after each unit.
• Employ Simple Active Learning Strategies: In-class and in homework, active learning assignments inspire learning.
• Collaborative Learning: Sometimes called team or group learning, collaborative learning can be designed to accelerate learning.

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 it’s 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, and 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

• 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 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.
• Try requiring 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 some research shows that we actually remember better when we write by hand. Varying the way we engage with thinking has value as it plays to different students’ preferences, and stretches all of us to try new ways to help us focus on the task of thinking.

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

Updated 2.8.23 EJI

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