Below are strategies based on faculty experiences and current research, to address the use of AI in assessments and assignments. They include approaches for mitigating AI use, as well as ways to leverage generative AI in assessment design. These strategies aim to promote fair, authentic, and inclusive assessment practices, in which students’ knowledge and skills can be accurately assessed.
When considering revising assessments based on the availability of AI tools, it’s important to align the assessment with your learning objectives: what is the assessment activity measuring, and does this fit with the learning outcomes for the course? When deciding between various assessment options, consider whether students have achieved the learning goals of the course.
Designing assignments to mitigate the use of AI writing tools
Some instructors may wish to design assessments that can help mitigate use of generative AI, to promote fairness and help ensure assessments reflect students’ own knowledge and skills. Some assessment strategies that are designed to promote academic integrity more broadly may also be effective for helping to reduce students’ ability to use generative AI on assignments.
Incorporate in-class assignments
One approach to mitigating the use of ChatGPT in assignment and assessment design is to incorporate more in-class or otherwise synchronous assignments, either written or oral, or change your current grade weighting to emphasize these (Weissman 2023). For example, in-class essays could work in larger classes, and in a small class one might meet with students to discuss texts written on their own time.
Bring parts of assignments into the classroom
While completing all assignments in class may not be feasible, integrating specific elements of assignments can help you monitor student work more closely and reduce the use of ChatGPT and other generative AI tools. For instance, you can have students reflect on the development of an assignment or work collaboratively in groups to solve specific problems related to their assignments.
By incorporating these elements into classroom activities, you can ensure that students are developing their critical thinking and writing skills while promoting academic integrity. This approach may also allow you to offer feedback as students are developing their work, fostering a more engaging and collaborative learning environment.
UBC Example – Christina Hendricks: PHIL 102
In PHIL 102 (Introduction to Philosophy), Christina Hendricks has engaged students in peer feedback during class. Students could bring an outline of an essay to class and share these outlines in pairs or threes, reviewing them for a few minutes, and discussing feedback. For longer essay drafts, she has asked them to review drafts ahead of time and provide feedback on a worksheet (Word doc) based on a marking rubric. The worksheet can be filled out before class and students can then discuss the feedback during class. She has also used a version of the same worksheet for students to do a self-assessment of their own work, which can also happen during class.
In the same class, Dr. Hendricks has also designed a discussion summary assignment in which students sign up for a week to raise discussion questions about the texts and topics for the week, in a small group discussion. Students need to submit their questions as well as a summary of the group’s discussion, for a small portion of their course mark.
Design non-textual assessments
Currently, most AI tools that generate text work through textual inputs, though some are developing the capacity to generate texts from images. For example, GPT-4 from OpenAI has the capacity to generate text based on images, though this functionality has not yet been released publicly. One strategy to mitigate their use, at least for now, could be to ask students to respond to non-textual resources, such as images, diagrams, or videos.
On the other hand, you could also design assessments that allow students to express their learning using methods other than writing. For example, depending on the course learning goals students could be asked to create a mind map, a timeline, an infographic, a video, etc. (Note that the landscape is changing quickly, though, and some of these artifacts may be able to be produced by AI tools very soon, if not already.) Students could also be asked to do presentations in class with Q&A, whether individually or in groups. Though they might use an AI text generator to develop part of what they present, they will still need to understand the material enough to effectively present it and answer questions from other students.
Using different assignment modes can help to support inclusive learning environments. Offering students a choice among modes can be helpful for supporting diverse learners, and adheres to Universal Design for Learning guidelines by providing multiple means for students to express their learning. This can also address the issue that such assessments, and platforms used for them, may not be equally accessible to all students.
Incorporate more disciplinary, situational, and individual-based questions
Another strategy that could be explored to address the issue of students using AI is by creating more personalized, contextualized, or authentic assessments. For example, students could be asked to discuss their own individual experiences or views on course topics, or to provide a specifically disciplinary or course-informed response to real or fictional case studies. You could also design assessments that connect to specific points discussed in class, on discussion boards, and the like, or to other courses students may have taken before.
UBC Example – Sarika Bose: ENGL 364 (fully asynchronous, online course)
In ENGL 364 (Nineteenth Century Literature), Sarika Bose incorporates questions on the final exam that require students to respond to specific scholarly articles assigned for the class as well as responses in the online discussion forum.
Continue reading ▼
Scaffolding Assignments
In her English 364 course on Victorian literature, Dr. Bose asks her class of 45 students to demonstrate their learning by writing a series of scaffolded assignments. Only the Term Paper assignment will be described here. Early in term (Week 4), they submit a Term Paper Proposal to the CANVAS assignment page. There are detailed instructions and they get detailed feedback from the instructor. The pedagogical purposes are to ensure students are learning in stages and also understanding how to organize their Term Paper work through the term in order to develop ideas in depth and without last-minute cramming. The final Term Paper must respond to the instructor’s feedback and also stay consistent with the general arguments that were initially proposed, though obviously, they are expected to evolve over the term as they are exposed to different readings. This line of tasks keeps students on track, and makes it harder perhaps to cheat.
About 3 weeks before the Term Paper itself is due (Week 13), students must submit a draft of their Term Paper, and engage in a peer-review workshop (with detailed worksheets that break down skills and content expectations), enabled by the CANVAS peer review function. It is to be completed approximately 2 weeks before the due date for the Term Paper. The pedagogical purposes are to ensure the Term Paper is well-advanced (though not perfect), and to allow students to get peer feedback regarding aspects like meeting the assignment requirements for research, careless errors, bibliographic accuracy, etc. Again, the many detailed requirements for the assignment, as well as the requirement to use some research articles from the course and some from outside, can help to defend against cheating.
In the final Term Paper, they must demonstrate they have taken on board some of the peer feedback, which is unique to them . They will also write a short reflection of what changes they made in response to that feedback and my feedback on their Proposal.
Their draft, peer feedback and Reflection will be checked as feedback and evaluation are written on their Term Paper.
The Final Exam has 2 parts: a short multiple-choice, live invigilated exam, and a longer take-home essay. For both parts, Dr. Bose will incorporate a requirement to respond to an idea pattern that was unique to this class’s weekly discussion forum, and design questions that incorporate requirements to respond to scholarly articles that were assigned for this course.
Example: She has found over the years that students who have been engaging in these discussions as required, have no problem picking out the prominent patterns, and no guesswork is involved. For example, one year, the drama class was talking about madness in Renaissance plays, another year my children’s literature class was talking about religion, etc. This term so far, the students are interested in misfits. Students also have to incorporate ideas from at least 1 research article I have assigned in the course.
ChatGPT is becoming known for making up citations and being slightly off-topic, but is quickly becoming more and more sophisticated. For example, colleagues have tested the function to see how it responds to requests to write papers on made-up theories. The first time, it’ll produce a perfectly plausible paper. The second time such a request is made, it’ll have learned, and will say there is no such theory. When citations are looked up, they either don’t exist or the page numbers are incorrect. Any information online may help ChatGPT; possibly more historical documents that are found physically in the library will need to figure more strongly in assignments.
Something Dr. Bose wants to tell students – who are so often talking about being allowed to have their voice and not being silenced – is that giving your voice and your intellectual activity over to A.I. is a willing surrender of your voice and your intellectual independence and knowledge.
As you can see from the above techniques, all of these strategies make a lot of extra, unpaid work for all instructors, but are a particularly large burden on the contract faculty who do most of the teaching.
Incorporating the use of an AI tool into assignments
Some faculty may wish to allow or promote the use of generative AI tools by students, to help them learn about such tools, understand their strengths and weaknesses, or to use them productively to improve critical thinking and writing skills. Similar to how researchers use other tools like NVIVO to code data and explain their methodology in research reports, students can use generative AI and explain its use as part of their method in writing assignments.
Please note the privacy and other ethical considerations discussed in this resource, if asking students to use such tools.
Have students analyze AI-generated text or other outputs
One way to include the use of generative AI in assignments, that may or may not involve students creating accounts and engaging with the tools themselves, is to guide students to critique responses provided by generative AI tools. For example, for written texts students could evaluate the accuracy and other strengths and weaknesses of the content created, identify gaps where more needs to be added, examine whether summaries appear to match the original texts, examine the credibility of cited sources, etc. It can be useful to provide a rubric to guide such review of AI outputs, to support students’ learning about how to improve their own writing.
Some faculty members are having their students analyze ChatGPT outputs, not just for written work, but for other outputs such as equations or chemical formulas. Despite the fact that the generative AI tools may provide incorrect information at times, this presents an opportunity for students to assess the AI’s answers and learn in the process.
Through active discussion with faculty members and fellow students, students can fine-tune their analytical and critical thinking skills using AI-generated responses. More importantly, such active discussion amongst faculty and students foster a scholarly learning environment in classrooms where students and faculty are communicating and developing ideas together.
UBC Example – Anubhav Pratap-Singh: FNH 303
Anubhav Pratap-Singh, an assistant professor in land and food systems, has incorporated ChatGPT into FNH 303, Food Product Development. As part of the course, students ask ChatGPT to generate formulations for specific food products and then request modifications to make the foods softer, chewier, sweeter.
Continue reading ▼
Using ChatGPT to suggest formulas and replacement ingredients for food product recipes, in an undergraduate course
In FNH 303 (Food Product Development), Dr. Pratap-Singh has modified the traditional “Product Selection” part of the assignment for New Product Development, allowing and encouraging students to use ChatGPT to identify recipes (ingredients and their compositions) for new food product development. Students are expected to use the Internet to modify an existing product in the market to either: a) make it more sustainable by removing problematic ingredients/processes; or b) modify the product to either add a claim (for example gluten-free; plant-based; etc.). Dr. Pratap-Singh is encouraging students to seek replacement ingredients using ChatGPT, as well as use other internet sources to describe known chemical and physical properties of different ingredients in the food product under consideration. To demonstrate their understanding of how each ingredient works, students are expected to address how modifications of the formulations may lead to food products with different traits (sweeter, chewier, etc).
UBC Example – Jennifer Jenson: ETEC 511
In ETEC 511, Foundations of Educational Technology, Jennifer Jenson asks students to provide their own views on several questions about AI and machine learning, and then compare to ChatGPT’s answers to the same questions, to better understand what the tool can and can’t do. In their own answers students must refer to readings provided along with the assignment.
Continue reading ▼
Using ChatGPT to suggest formulas and replacement ingredients for food product recipes, in an undergraduate course
Artificial intelligence is a keystone of educational technology. So it’s extremely important that we are able to think critically about the scope and limits of what machine intelligence can do, and how our uses of it impact our own and our students’ learning and educational development.
Preparing for the assignment
Set aside a good chunk of time, several hours, which needn’t be all at once, (you could easily spend years here, it’s so interesting!), to find/read information that will enable you to formulate your answers to this assignment’s 5 questions. Take some time to watch Kate Crawford introduce a critical technology studies perspective on the “power, politics and planetary costs of AI.” Do not shy away from Wikipedia for this one—it’s a lot to cover, and you are only exploring and sketching out the area. I’ve provided a few links to get you started, but you can consult any resources. Use these 5 questions to organize your search, and keep a tight focus on AI, not on computing, or economics, or pedagogy, etc., except where you need to mention these aspects to provide context.
One useful starting point is Harvard University’s Science in the News Summer 2017 Special Edition on Artificial Intelligence. Note that some of these linked papers are not currently ‘live’ from the nested links, but you can still find them by searching outside this volume directly on the website.
A “word to the wise” before you begin…
Be prepared in advance that you will not be able to answer all these questions in depth or detail: this is a ‘scoping out’ exercise, where in order to get the perspective you need to engage with these questions, and especially the last one, you first need some breadth. You will be doing a lot of skimming and some of what you will need to skim will have highly ‘technical’ sections that you are of course not expected to understand (unless you have prior knowledge in these areas).
Then, to compose your responses, you need to condense to the “nugget” of things, to work within the short-answer word constraints. Simply do what you can, it’s YOUR journey, and this is a first step into vast (and life-altering!) territory. This is not meant to be easy, and there are no right answers. Careful writing matters. Please resist copying and pasting—this lands you firmly in plagiarizing territory. Use your own words. Do not string together a bunch of quotes.
Production Format
In sentences, not point form, respond to the questions below. Do not exceed word limits. Use the APA format when citing sources and listing references. For this, and every assignment in this course, multimodality is encouraged, even when the assignment asks for primarily textual forms. Remember to limit your quotations, and do not copy and paste without properly citing.
The 5 Questions
- First and in your OWN words. Who were these people, and how did/does each contribute to the development of artificial intelligence? How did/does each think “intelligence” could be identified? (~50 words each). Next, using the new AI ChatGPT, ask it the same question/s and edit the responses down to 50 words. You can create a chart to show your work and ChatGPTLinks. side-by-side. Do the same with questions 2-4, and DO make sure in your answers you reference the readings.
- How do “machine (programming) languages” differ from human (natural) ones? (~100 words).
- How does “machine (artificial) intelligence” differ from the human version? (~100 words).
- How does “machine learning” differ from human learning? (~100 words)
- And for your LAST challenge, a version of the Turing Test: how do YOUR answers to these questions differ from what a machine could generate? (~200 words) Make sure you reference the differences between your answers and ChatGPT’s.
Support students using generative AI in their own learning
Some educators are introducing the use of generative AI in written assignments as a new approach in specific academic domains. Students can use this technology and include a reference section to detail their usage. At UCLA, Professor John Villasenor is one of the advocates of this approach and even encourages first-year law students to employ ChatGPT. To create work that combines the strengths of both AI and traditional writing techniques, students must learn to seamlessly blend generative AI into their assignments in a logical and coherent manner.
You could suggest ways for your students to utilize generative AI tools like ChatGPT to enhance their learning experience. Students may use generative AI writing tools to support parts of their ideation, research, and writing processes, while doing other parts themselves grounded in disciplinary and course-specific topics and methods.
In connection with all of the suggestions below, you could require students to submit screenshots of the AI outputs, and describe how they built on that. It’s important to remind students that AI can (and at this point often does) make mistakes and provide false information, and that they need to check the AI outputs and correct them as needed.
Ideation and brainstorming
Students could use AI to generate initial ideas for research topics and questions that they then refine and add to themselves. For example, AI could help them move from broad ideas to more specific questions that they then refine and address according to disciplinary and class context.
Initial research
Students could use AI platforms to do basic research to get an overview of a topic, and to help them focus later work based on concepts and keywords they have learned. For example, they can use perplexity.ai or other AI tools for research and analysis by inputting research questions or topics and finding sources and key terms to explore.
Improving grammar and other aspects of writing
Students can input parts of their writing into ChatGPT or other tools to receive feedback on common grammatical errors and tone. ChatGPT can not only edit writing, but explain what it changed and why, which could be a useful way for students to learn.
Adding creative elements to assignments
Students could use generative AI tools to add more creative elements to their work, such as using image generators like Stable Diffusion to create images that they add to slide presentations, games, apps, portfolios, blog posts, and more. There are also AI tools that can generate music or sound effects that could be used for student-created videos or games. Students could also use AI text generators to create draft scripts for videos that they can edit and refine to provide more details and information to better fit with the course context and learning objectives of assignments.
UBC Example – Halimat Alabi, MET program
In her courses in the Masters of Educational Technology program, Halimat Alabi expects students to use the AI tool AcaWriter to refine their writing before they share their drafts in class.
UBC Example – Anubhav Pratap-Singh: FOOD 524
Using ChatGPT to obtain preliminary outline for term papers, and obtain basic understanding of terms/technologies in the field for graduate courses
In FOOD 524 (Advances in Food Processing), students have to submit a term paper on advances in the last 5 years in any particular area of food processing of their choice (20 pages maximum). Dr. Pratap-Singh is encouraging his students to use ChatGPT to get inspiration and ideas on how to approach this term paper, to identify topics of rapid development in selected fields, to come with an initial outline of the term paper, and to use it along with other avenues on internet, to obtain general information about various new terms they shall encounter while completing this term paper assignment. Dr. Pratap-Singh has illustrated to the class how allowing ChatGPT raises the bar of the standard of term paper submissions expected in the course (akin to how open-book exams are designed intellectually tougher than closed-notes exams). Additionally, Dr. Pratap-Singh used ChatGPT during one of the lectures on refrigeration to examine the merits of the following claim together with students: “Mpemba Effect: Does Hot Water freeze faster than Cold Water”
