Last updated on April 27, 2023
Contents
Prompting/General Use/Practice
Perhaps the greatest concern about large language models in higher education is cheating. It is possible for learners to ask ChatGPT to craft entire essays, discussion posts, and other writing assignments for them. Academic integrity always has been and always will be an issue. The relative ease with which students can cheat today proves to be the most disruptive circumstance in higher education since the computer, the pocket calculator, the slide rule, the slate–you get the picture.
(click below to reveal the example)
Try It Out!
- Using ChatGPT, post an assessment prompt (as-is) from one of your courses.
- What is the result? How accurate is it?
- Next, post the same prompt, adding relevant course materials before your question.
- How does the response change?
- Reflect on the accuracy, writing style, and tone of the response.
- How easy would it be to determine when a student’s response was generated by GPT?
Educators have adapted before, and we must do so now if we want to continue delivering quality instruction and ensuring that this instruction is internalized by learners. It simply is no longer possible to teach using methodologies from 1990 (or 1890) and to expect the same results. How can this be accomplished? Here are a few ideas:
Eliminate Regurgitation
The easiest way to assess learners is to test their recall using fact-based assessments that have not progressed further than the bubble sheet. Learners memorize facts, a low-level skill, and re-present them to us objectively. Avoid this by finding ways for learners to analyze and evaluate information. Not only are these higher-order skills, but responses to prompts given in such ways are more difficult to generate digitally. As a result, your students will become better able to think critically and independently. This is a foundational principle of higher education in general, and part of MSU Denver’s core values.
(click below to reveal the example)
Try It Out!
Before: The process that moves the greatest amount of heat between the tropics and the poles is:
- a) Hadley cells
- b) jet stream
- c) thermohaline circulation
- d) convection.
Revised: “Climate patterns play a crucial role in redistributing heat across the planet, but which of the following mechanisms is responsible for transferring the most heat energy between the Earth’s tropics and polar regions? Consider the factors that affect these processes, such as ocean currents, atmospheric pressure, and temperature gradients, before selecting your answer from the following options: Hadley cells, jet stream, thermohaline circulation, or convection.”
If your course leans heavily toward objective assessment like quizzes, try this strategy.
Just the Facts?
In many ways, objective assessment prompts can be reduced to binary answers: yes or no, true or false. Instead, always use open-ended questions. This persuades learners to think more deeply about the topic, as well as to develop their own opinions and ideas.
(click below to reveal example)
Try It Out!
Before: After viewing the video of a student being assessed for name writing and letter recognition, the student is most likely in which stage of literacy development?
Revised: After viewing the video of a student being assessed for name writing and letter recognition, how do you feel about giving these types of literacy assessments? How did you score the student on the letter recognition test? Explain how you would have evaluated the student’s writing ability?
If your course leans toward questions designed to extract simple answers from learners, revise your questions in ways that make them think about the situation and reflect on what they have learned.
Keep it Real
Use case studies that are current, real-world examples of your chosen topics. If at all possible, make scenarios relevant to your learners. In addition to promoting critical thinking in learners, this will increase engagement. As a bonus, most large language models in use today are trained on older data. Keeping your examples fresh makes it more difficult to use these models for writing assignments.
(click below to reveal example)
Try It Out!
Before: Complete the balance sheet for this fictitious corporation.
Revised: Consider the recent case study of Silicon Valley Bank. Critically evaluate the role of accounting in shaping the public’s perception of the bank’s collapse and the impact on individuals, businesses, and society as a whole. What ethical considerations should accountants consider when dealing with such situations, and how can they ensure that their work is transparent and “accountable”? How can individuals and businesses make use of accounting principles to make informed decisions about their own finances and investments, and what challenges and opportunities does this present in today’s economic landscape?”
If your course re-uses data and assessment prompts each semester, it is much more likely that generative AI (and “study” sites) will have access to the prompts and the data.
Follow the Herd
Use group activities that are project-based. These activities should be completed in phases, and instructor feedback is crucial after each. In this way, learners interact with you, with the content, and with each other throughout the process.
Try It Out!
Before: Write a two-page essay about how institutions, policies, and cultural norms perpetuate or challenge gender identity stereotypes and discrimination.
Revised: For your final project, in groups of four, you will conduct a research project that explores gender identity with an intersectional focus. Deliverables will include a research proposal, literature review, data collection and analysis, research paper, and group presentation.
If your course focuses mainly on individual writing assignments, converting to a group format can encourage avoiding generative software to complete assignments, especially if groups agree to set their own standards for the usage of this software.
You Can Use it, Too!
Of course, it is also possible to use ChatGPT to your own benefit when making revisions like those seen above! If you find yourself with writer’s block when thinking up an assessment for your course, feed your prompt to ChatGPT and ask it to create a new prompt for you that includes the qualities you’re looking for.
Try It Out!
Example: “My aerospace course has this assessment prompt: [paste here]. It’s all over Chegg. I need a new prompt for this skill that [makes students think critically, uses different data, makes the assessment relevant to my students, also makes students reflect on their own learning].”
AI Checkers
Inform learners on AI-detecting tools, plagiarism, and academic integrity. Set clear guidelines on plagiarism and which (if any) plagiarism-detecting tools are used in a course. Include an academic integrity statement in the syllabus or in a syllabus quiz to ensure learners are aware of the consequences of using AI to plagiarize and other acts of misconduct related to cheating. Keep in mind, AI checkers will always be playing catch up with the generative AI technology, and as such will likely not be 100% accurate. Be sure to consider ways of addressing potential use of AI as cheating beyond immediate punishment, as you may get false positives from AI detectors.
Try It Out!
Run several student essays through an AI detector, such as OpenAI’s own AI Text Classifier. What parameters does the tool use to determine cheating or plagiarism? Can you spot these things without a plagiarism checker?
Policy
In order to support the use of chatbots and ChatGPT to support the learning environment, instructors should provide clear guidelines to learners on the dos and don’ts of using AI in their respective courses. Instructors should practice feeding different prompts to ChatGPT in order to better identify the language when it is used. Adequate time should be given for assignments so that learners can engage with classmates using peer reviews and ongoing feedback from instructors. Assignments can be chunked to provide tiered feedback and time for learners to correct their mistakes and revisit their work. Some examples of policies are as follows:
- Require learners to submit drafts of their work
- Ask learners to present their work (in class, video recording, slides, etc.)
- Set clear guidelines for use of GPT and other chatbots
- Use proctoring software when needed
- Use project-based assignments