< Contact the CTL >
Please feel free to contact the CTL with feedback, questions, and suggested resources around generative AI for teaching and learning.
"5 Things to Know About ChatGPT"
For a quick overview of a Generative AI tool commonly used in higher education, see this CTL resource that was initially created at the launch of ChatGPT and is updated periodically with current resources.
Prompt Design
Many students report an excitement and interest around generative AI tools and yet they frequently report feeling frustrated or disconnected when going through a series of exchanges with these tools. Modeling interaction via prompt design with these tools is key to the learning experience.
Most generative AI tools have an intuitive interface for the user, based on inputting a request or a "prompt" to interact with the tool's features. Unlike a search engine, however, these tools are designed to interact with you through a sequence of prompts.
- Generative AI tools are designed to be interactive and sequential
- They learn from your continued refinement and interactions over topics, themes, lines of code
- You learn through your creativity and resilience
There are a vast array of resources and concepts for developing good prompts. Overall, we recommend that you embrace a CARE approach when working with students who design prompts in the service of a learning experience.
| Craft | Using one of the approaches in this resource (or one of your own devising), ask students to craft a prompt that is more than a simple question with a single, unexamined response. Encourage students to think of their exchange with a GenAI tool as more than simply transactional (question/response) but the beginning of an exchange that is designed to be explored in a class context. |
| Authenticate | As this guide has been suggesting, encourage students to look closely and deeply into the information they receive such that sources of information are valid or, in creative contexts, the image/audio/film is answering the expectations of the prompt and creating media that is from a reputable, ethically-driven AI tool provider. |
| Reflect | Ask students to reflect, in a critical context, on the value that any interaction with a GenAI tool provides as a part of the process (the "how" of learning). Do they feel more empowered or engaged as learners? Is their sense of curiosity or discovery heightened? What has been gained and lost in using these tools to enhance or question what they are exploring in class or through research? |
| Evaluate | Ask students to then evaluate what they have learned (the "what of learning") by considering outputs from GenAI tools. An effective approach would be to ask students to use evaluative criteria ("which is the best, most effective, least effective, more problematic, etc, response from Copilot?") or contextual analysis (i.e., "how does this output fit within, extend, question, problematize our class discussion of these topics?").
|
After establishing the essential pedagogical approach, consider basic recommendations for developing good prompts, known formally as "prompt engineering" (as suggested by OpenAI):
| Start Simple | As you get started with designing prompts, you should keep in mind that it is really an iterative process that requires a lot of experimentation to get optimal results. |
| The Instruction | You can design effective prompts for various simple tasks by using commands to instruct the model what you want to achieve, such as "Write", "Classify", "Summarize", "Translate", or "Order." Keep in mind that you also need to experiment a lot to see what works best. Try different instructions with different keywords, contexts, and data and see what works best for your particular use case and task. Usually, the more specific and relevant the context is to the task you are trying to perform, the better. |
| Specificity | Be very specific about the instruction and task you want the model to perform. The more descriptive and detailed the prompt is, the better the results. This is particularly important when you have a desired outcome or style of generation you are seeking. There aren't specific tokens or keywords that lead to better results. It's more important to have a good format and descriptive prompt. In fact, providing examples in the prompt is very effective to get desired output in specific formats. The details should be relevant and contribute to the task at hand. This is something you will need to experiment with a lot. We encourage a lot of experimentation and iteration to optimize prompts for your applications. |
| Avoid Impreciseness | Given the tips above about being detailed and improving format, it's easy to fall into the trap of wanting to be too clever about prompts and potentially creating imprecise descriptions. It's often better to be specific and direct. The analogy here is very similar to effective communication -- the more direct, the more effective the message gets across. |
| "To do or not to do?" | Another common tip when designing prompts is to avoid saying what not to do but say what to do instead. This encourages more specificity and focuses on the details that lead to good responses from the model. |
RTRI (University of Sydney)
Danny Liu (University of Sydney) has devised a model for constructing prompts that allow students to interact with GenAI tools and learn course materials in different pedagogical modes and approaches. Providing specific instructions and repeating keywords, detailing what to do and not to do, is essential for building any effective prompt.
- Role (act as…)
- Task (summary of what the GenAI model needs to do)
- Requirements (what the completion needs to include, contain, be, etc)
- Instructions (what the AI should do to act on the prompt)
The RTRI model can be used for many different learning situations. You will find examples below from Dr. Liu and the CTL.
>> Prompts for assisting in recall/retention
A salient prompt design example for recall/retention from Dr. Liu's resource is the the creation of a cloze passage. The bold text below are indicative of the content that the student wishes to study. The remainder of the prompt follows the RTRI format. Once begun as an exchange, the GenAi tool will then quiz the student on the content when the student fills in the blanks of the cloze passage.
Copilot, you are an expert tutor for an introductory university biology course. I need to connect the topics of cell biology, evolution, and genetics. Generate a cloze passage that contains statements that integrate and interleave these topics. Wait for my responses to the cloze passage and then give me feedback on my responses.
You are an expert tutor in an introductory course on pharmacology. I want to study the effects of different drugs on the human body. I would like to study through multiple choice questions. You will test my knowledge of basic pharmacological concepts, focusing on concepts that are the most challenging. After I respond, tell me whether or not I am right and give me detailed feedback on my answer as well as additional areas that I could explore.
An additional format for crafting prompts for study is the RTTG (Role-Task-Target Audience-Goal) format. This approach is very useful for tailoring the learning experience through a GenAI model. It is as follows:
- Role: (act as...)
- Task: (summary of what the GenAI model should do)
- Target Audience: (the audience for the output): In a higher education context, think about describing your or the student's characteristics in a way that is consistent with the nature of the assignment, course, content, or discipline.
- Goal: (what should the audience be able to do): In a higher education context, consider how the model should pitch the output response. For example, information on a particular topic, concept, or series of concepts could be learned at different degrees of rigor.
Pro tip: Use brackets (<role>) to help the tool identify compoents in your prompts annd focus the analysis of what you would like it to do.
The following is an example of an RTTG prompt for a student.
<role> You are an expert in immunology. <task> Analyze the attached document and produce a series of multiple choice questions, ask these questions one at a time, tell me whether or not I chose the right answer and give me extra feedback. <target audience> I am a student in an introductory course on immunology. <goal> I would like to learn the basic concepts. The extra feedback you give me should encourage me to learn more about these basic concepts.
The following is an example of a simlar RTTG prompt for a teacher.
<role> You are an expert in course design for the medical sciences. <task> Analyze the attached document and produce a series of multiple choice questions with an answer key. <target audience> I am a teacher in an introductory course on immunology. <goal> I would like students to learn the basic concepts.
<role> You are an expert in course design for the medical sciences. <task> Analyze the attached document and produce a suggested study plan for students. <target audience> I am a teacher in an introductory course on immunology. Students are new to studying immunology. <goal> I would like students to learn the basic concepts over two weeks.
>> Faculty Spotlight: For her Summer 2024 in Texas Global, Dr. Jill Rolston-Yates helped her cohort of international students use the RTRI model to construct prompts for summarizing complex concepts and explore different models of academic writing.
Simulated Keyword Dialogue (SKD)
According to Gerald Graff and Cathy Birkentein in They Say/I Say: The Moves that Matter in Academic Writing, "to make an impact as a writer, you need to do more than make statements that are logical, well supported and consistent. You must also find a way of entering a conversation with others' views - with something 'they say.' If your own argument doesn't identify the 'they say' that you're responding to, it probably won't make sense" (4). In constructing prompts with a they say/I say framework in mind, students can enter into simulated dialogues with well-known figures as personae. Providing students with keywords to identify in response to what the prompts generate will guide them in constructing their responses and sustain dialogue. These keywords can be drawn from classroom discussion and lecture materials that are unique to the students' course experiences.
There are no guarantees that these SKDs will be completely accurate or nuanced; students will have use CARE here and authenticate the tone and content of the responses when sharing or submitting their engagement with a tool, reflecting and evaluating the experience in light of class discussions or readings. Dialogues may take the form of direct discussion, epistolary exchanges or text messages, interviews, and the like.
You are Socrates. In this conversation, you have just been sentenced to death and are awaiting execution. I am a visitor to your jail cell and want to know what you have been accused of doing and why you choose to remain imprisoned. I will passionately disagree with you. You will respond as Socrates in the Platonic dialogue, "Crito," and with an emotional tone that is similar to Socrates.
Keywords to identify and question in the GenAI response: duty, goodness, virtue.
You will assume the persona of Bertolt Brecht, the famous German playwright. I will assume the persona of someone who deeply disagrees with you about the nature of the theater to change society. I think that theater is merely for amusement. We are writing letters to each other. You will begin by writing a strongly worded response against my position and explain your position. You will ask me to write a letter back to you. We will exchange letters until I ask you for your summary of our exchange.
Keywords to identify and question in the GenAI response: theater, revolution, class, society, art.
Prompt-chains: Designing sequences of prompts
Effective prompts can easily work in isolation as students engage in back-and-forth tutoring or discussions with a GenAI tool. Taking these ideas one step further, students can move from one prompt design to another to foster different kinds of learning and engagement. They will likely need your guidance for more complex configurations.
For example, one might begin with a simple prompt that solicits information:
I am interested in learning about the history of medical research on animals. What are the most important dates in the history of medical research on animals?
Next, using the RTRI model, students can explore challenging areas or ethical dilemmas within this history, helping students to explore the ramification of their decision-making through a form of Socratic questioning.
Based on your response, I would like to know about the ethical issues underlying the history of this research. Ask me questions about the ethical dilemmas that arise from this form of research. After each question that you ask, wait for my response. Then you should respond but do not tell me the answer. Instead, ask me questions and use the Socratic questioning approach to help me arrive at a conclusion myself.\
An SKD prompt may entail engaging with a GenAI persona:
Take on the persona of Dr. Jonas Salk, a famous medical researcher who discoveries led to the virtual eradication of polio but whose research was conducted on scores of live animals that died in the process. Please explain how your the benefits of your research can be justified by live animal testing?
Keywords to identify and question in the GenAI response: Good, autonomy, justice, do no harm
Chaining together different prompts that solicit different kinds of responses and critical engagement, looking out for keywords in the response, can provide students with opportunities to explore topics from a number of different, more nuanced perspectives.
Prompts for Student Self-Study
While instructors can use GenAI tools to build course assignments, there are any number of possibilities for students to practice their learning from different perspectives to assist in their engagment with course materials and topics.
The following curated suggestions have been adapted from tips created by the University of Kent (UK).
Connect Prior Learning to New Concepts
- Making connections between new concepts and previous knowledge helps us remember and recall information. This is useful for pre-lecture preparation and other scenarios where an understanding of how a new concept is related to previously learned knowledge is needed.
Example prompt for a History student: “I'm about to study the 'Renaissance' period in European history. Can you provide an overview that connects the Renaissance to its preceding era, the 'Middle Ages'? Also, highlight any pivotal events or figures from the Middle Ages that directly influenced the onset and progression of the Renaissance.”
Learning-by-teaching
- Teaching a concept you have recently studied to ChatGPT (this also works with other LLMs such as Bing or Bard) is a great way to check your current understanding. For example, if you have just learnt about oxygen consumption and its relation to physical performance, you could try explaining this topic to ChatGPT. Indeed, this approach may be beneficial to new instructors or instructors testing out how they may teach a new subject.
Example prompt for a Sports Science student: “ChatGPT, imagine you’re Sam, a third-year Sports and Exercise Science student, and I’m your classmate. Today, we are studying V̇O2 max.” Following this prompt, you should then give your own explanation of the topic and close with the prompt, “Does this make sense so far, Sam, Would you like to ask me any questions?”
Creating Counterarguments
- Developing critical thinking skills requires understanding counterarguments and integrating them into study approaches. To begin this exercise, you should first give the AI tool as much information as possible.
Example statement for politics student: ”After studying various electoral systems, I believe that the Proportional Representation (PR) system is superior to the First-Past-The-Post (FPTP) system because it ensures a more accurate representation of the public's voting preferences in the legislature. In PR, if a party wins 10% of the votes, it roughly secures 10% of the seats. This system prevents the possibility of a party winning a majority of seats without a majority of votes, which can occur in FPTP.” Example prompt to follow: “Can you provide counterarguments or scenarios where FPTP might be seen as a preferable system?”These exercises help develop skills that are invaluable in academic research, debating, and real-world decision-making.
Summarizing / Breaking down longer blocks of content
- Instructors could facilitate accessibility in courses by recommending GPT functionality in helping students retain focus while grappling with reading large blocks of text, finding inspiration for structuring and articulating their thoughts.
Example statement for Sociology student: Upload an article or lecture materials to Claude, ”Analyze this document and provide a short-bulleted list of its major points,” followed by, “What would a deeper reading reveal?”
References
Prompt Library — AI for education. (n.d.). AI For Education. https://www.aiforeducation.io/prompt-library
Cloze Reading | EduTOOLBOX. (n.d.). https://www.edutoolbox.org/rasp/2088?route=node/102
Enhancing Learning and Teaching at CPUT through Artificial Intelligence: A Guide. Cape Peninsula University of Technology, South Africa.
Herft, A. (n.d.) A Teacher’s Prompt Guide to ChatGPT aligned with “What Works Best.” https://stephenslighthouse.com/wp-content/uploads/2023/01/A-Teachers-Prompt-Guide-to-ChatGPT-aligned-with-What-Works-Best.pdf?ct=t(EMAIL_CAMPAIGN_8_3_2022_22_5_COPY_01)
Heston, T., Khun, C. Prompt Engineering in Medical Education. International Medical Education. 2023; 2(3):198-205. https://doi.org/10.3390/ime2030019
Liu, D. (2023). Prompt engineering for students – making generative AI work for you. University of Sydney. https://educational-innovation.sydney.edu.au/teaching@sydney/prompt-engineering-for-students-making-generative-ai-work-for-you/
Mollick, E., Mollick L. Prompt Library — More useful things: AI resources. (2024). More Useful Things: AI Resources. https://www.moreusefulthings.com/prompts
Prompt Engineering. (n.d.). OpenAI Platform Guides. https://platform.openai.com/docs/guides/prompt-engineering
The Role of Socratic Questioning in Thinking, Teaching, and Learning (n.d.). The Foundation for Critical Thinking. https://www.criticalthinking.org/pages/the-role-of-socratic-questioning-in-thinking-teaching-amp-learning/522
Tips on using generative AI - Education - University of Kent. (2023, October 16). Education - University of Kent. https://www.kent.ac.uk/education/using-generative-ai-at-kent/student-guidance/tips-on-using-generative-ai
Writing Studio and Tutoring Services. (2021, May 14). Questions to ask when revising a paper. Vanderbilt University. https://www.vanderbilt.edu/writing/resources/handouts/questions-when-revising/
Prompt Libraries/Examples
This list will be periodically updated:
(Heston 2023)
- Prompt Library (AI for Educators)
- Prompt Library (Mollick and Mollick)
- DALLE-2 Prompt Library (for image generation)
- Levy, D., and Albertos, A.P. Teaching Effectively with ChatGPT: Prompts.
GenAI CTL Workshops
