(Revised 8/9/23)
This guide provides instruction on how to interact with the LLMs, ChatGPT, Bing, Bard, and Claude (referred collectively as “AI” in this document). For an introduction to artificial intelligence (AI) and large language models (LLMs) please see our Faculty Guide to AI, and the accompanying Student Guide to AI.
How Can AI Help Me?
What is a Prompt?
How Do I Construct a Prompt?
How Can I Improve my Results?
What Can I Ask AI to Do?
How Can AI Help Me?
Fundamentals of Teaching
Course Design: AI can provide advice on designing an effective course syllabus, incorporating appropriate learning objectives, choosing engaging readings or resources, and structuring course content in a logical, learner-friendly manner.
Teaching Strategies: AI can be your instructional coach, suggesting various teaching strategies based on the specific context of a course. This can include differentiating instruction, scaffolding complex concepts, and incorporating active learning strategies, applying evidence-based learning concepts to instruction and homework.
Assessment Techniques: AI can provide guidance on effective assessment techniques, helping to develop rubrics, formative assessments, and authentic assessments to accurately gauge student understanding. AI can help ensure your instruction and assessments align with learning outcomes and can help you design assignments that result in artifacts to provide evidence of this alignment. It can design low/no stakes practice quizzes and exam questions.
Communicating with Students: AI can help with strategies for communicating clearly and effectively with students, both in person and through digital platforms.
Enhanced Teaching Techniques
Feedback Techniques: AI can help you learn how to give constructive, actionable feedback that helps students improve.
Flipped Classroom Implementation: A flipped classroom is a useful response to the proliferation of AI, and AI can provide guidance on how to create pre-class materials and activities, and how to effectively utilize face-to-face class time for active learning.
Motivating Students: AI can provide strategies for motivating students, such as incorporating real-world applications of course content, creating meaningful assignments, and recognizing student achievement.
Student Engagement: Engaging students can be a challenge, but AI can provide strategies for increasing student engagement, such as active learning exercises, discussion prompts, and interactive assignments.
Facilitate Group Work: AI can provide strategies for creating productive group work assignments, addressing common issues that arise in group dynamics, and assessing individual contributions within group work.
Classroom Environment & Dynamics
Community Building: Creating a sense of community in the classroom should be a foundational goal. AI can provide strategies for building community, such as icebreakers, group activities, and class discussions.
Classroom Management: AI can provide strategies for maintaining a positive, inclusive, and productive learning environment. It can also help you diagnose and remediate specific student behavioral issues.
Inclusive Teaching: AI can provide resources and strategies for inclusive teaching, helping to create a classroom environment that respects and responds to diversity in all its forms.
Promoting Academic Integrity: AI can provide strategies for promoting academic integrity and addressing issues such as plagiarism or cheating.
Advanced Pedagogical Approaches
Universal Design for Learning (UDL): AI can explain the principles of UDL and provide strategies for designing course content, teaching methods, and assessments that are accessible to all students, regardless of learning styles or abilities.
Reflective Practice: AI can assist faculty in reflecting on their teaching practices, prompting them to consider what worked well and what could be improved.
Interdisciplinary Teaching: AI can offer suggestions for interdisciplinary teaching, including examples of interdisciplinary assignments or projects.
Peer Observation and Feedback: AI can guide professors through the process of conducting effective peer observations, including what to look for and how to give and receive constructive feedback.
Integration with Technology & Research
Teaching with Technology: AI can guide the use of technology effectively in the classroom, including recommending educational tools, providing tips on online teaching, and suggesting ways to integrate technology into course content.
Research in Education: AI can provide summaries and insights from key research studies, so that we can stay up to date on best practices in pedagogy.
What is a Prompt?
A “prompt” is a set of instructions that is typed into the input field of the chosen LLM. “Prompting” is how one interacts with AI. Prompting should be thought of as an interactive and iterative process. It is not complicated, but it takes time and practice to understand how to communicate with AI, learn how it best operates, and how to achieve the best results. Each of the models behaves differently and has its own strengths and weaknesses.
Important: While AI is improving dramatically, it is not uncommon to receive information that is incorrect or misleading (also known as “hallucinations”). You cannot assume that AI is providing accurate, factual information. All output must be verified for accuracy.
How Do I Construct a Prompt?
While interacting with AI should be as conversational as possible, some basic structure will allow you to be both more efficient and more effective in getting the kind of results you seek. It is helpful to consider the following basic structure, which can then be modified based on your needs:
Context: The AI needs to understand the context of your prompt. There is some contradictory information about the need to tell the AI who it is (ex., “you are an experienced teacher”, “you are the editor of an academic journal”, “you are the creator of diagnostic quizzes”). Some find this helpful, but its impact is unclear. It is important, however, to situate the work you want the AI to do.
Task and Constraints: The AI needs to know what to do and how to do it. Simply ask or tell it what you need. Adding constraints to help the AI focus leads to more useful output. This could include conceptual constraints, level of difficulty constraints, time constraints, complexity constraints, and more.
Steps: Providing a step-by-step procedure can be a useful way of guiding AI’s work and output.
Results can often be improved through interaction. Context, task, constraints, and steps can be altered as you engage in a back-and-forth with the AI.
For example (give this prompt a try, but do not include the bracketed phrases): You are an experienced college instructor well-versed in evidence-based teaching and learning strategies, and you are here to assist me in instructional design. I use a flipped classroom. I want my students to be prepared for activity and conversation about AI and how to use LLMs as a learning tool, which will take place in the subsequent class [context]. We are going to create a homework assignment for college freshmen that will help them understand how LLMs, like ChatGPT, can be a useful learning tool. This assignment should give them a basic understanding so that they are prepared to engage with me and one another in class. The assignment should take no more than 1 hour to complete. It should be designed in a way that significantly reduces the opportunity to cut and paste from AI. It should require some kind of artifact that students must bring to class. [more context]. First, provide a learning outcome for the assignment. Then, provide a sentence or two of description. Finally, provide detailed directions for the students, including the description of the artifact that they will bring to class. Remember to apply what you know about high quality teaching and learning to the design of this assignment. [steps]
The more detailed the input, the more likely it is that you’ll receive a satisfactory response. However, once the AI responds, you might consider altering the learning outcome, or perhaps there are some critical points you already have in mind that should be emphasized in the assignment. Think of the AI as your assistant, so use your knowledge and expertise to guide it. For example, you might tell the AI to “make the following points…” or “include the following topics…” or “I want to make sure the students are challenged by, or…exposed to…”
While it is often helpful to keep a chat session open so that the AI remains within the context you worked to create, it sometimes gets itself into a non-helpful loop that you may need to end. If the AI begins to get confused or the results seem less than satisfactory, start a new chat session and you can begin again.
How Can I Improve my Results?
There are numerous strategies you can employ to improve your results. First, you might try your prompt in various LLMs. They behave somewhat differently and have different strengths and weaknesses. You will also want to consider the following strategies, some of which can be built into your original prompt, or added as part of your guiding conversation after receive results.
Ask AI for help in crafting your prompt.
Paste in text to analyze for style, so that the AI can use that style or respond in accordance with it.
With access to the internet (Bing, Bard, or Chat GPT with plug ins) you might tell AI to look up information first so that it has something tangible to work with.
Use a one step at a time approach, adding another step or task after receiving some output.
Work interactively…push back…engage AI like you would a colleague or intern.
Ask AI to expand on a result and provide more information.
Ask AI to change the example.
Use your expertise, include more of your thinking.
Use alternate forms of prompting (see below)
What Can I Ask AI to Do?
Below are several types of prompts. Add context, constraints, and steps as described above.
Extract specific facts or explanations. Ideal for quick clarifications or to obtain definitions. (“What is quantum entanglement?”)
Generate broad and comprehensive content, perhaps as a starting point for lecture notes or to gather diverse perspectives on a topic. (“Discuss the significance of the Renaissance.”)
Consider a specific context or background. Useful for creating scenario-based exam questions or for lesson planning. (“Given the sociopolitical climate of the 1960s, explain the rise of counterculture.”)
Develop materials that require comparisons, such as study guides. (“Compare classical conditioning and operant conditioning.)
Generate hypothetical scenarios or consider specific conditions. Useful for class discussions or thought experiments. (“If Shakespeare were alive today, how might he comment on modern society?”)
Craft content in a specific format. (“Draft a multiple-choice question about photosynthesis.”)
Gather insights from specific viewpoints, aiding in presenting multiple sides of an issue. (“From a Marxist perspective, analyze the modern gig economy.)
Build upon prior responses to delve deeper into a topic or to refine a complex concept progressively. This can help when preparing in-depth content or research. (After an initial answer about climate change, follow up with, “How do ocean currents contribute to this?”)
Speculation or forward-thinking content. This could aid in spurring classroom discussions or in exploring potential future research topics. (“Predict the future challenges of urban planning in the face of increasing automation.”)
Refine questions based on past outputs, to get more targeted or accurate information over time.