Univeristy of Pennsylvania
Close

Prompting Strategies for Instructors

09Jan’25

Generative AI

Learning to prompt effectively can save time and produce better results. Prompts are the directions or requests that you, the user, provide to generative AI. The content of the prompts you enter directly influences the responses you receive. 

Considerations for Effective Prompting

  • Thoughtful prompting is worth the effort. Learning to prompt effectively can take practice but ultimately saves time by more quickly and consistently generating AI output that meets your expectations. 
  • Remember to iterate. If the output you receive isn’t exactly what you want, you can ask AI to revise, expand, adjust, or combine ideas instead of starting over. 
  • Technology continually evolves. Prompting frameworks are suggestions, not scripts. Use this information as a starting point for further experimentation as AI tools continue to develop.

General Prompting Recommendations

In the early days of generative AI, it was necessary to compose incredibly detailed prompts using a rigid structure. As AI’s capabilities evolve, prompting strategies have become simpler and more flexible for most routine tasks, but there are still some helpful guidelines to keep in mind. 

  • Be specific. As AI advances, it remains vital to explain exactly what output you want. The more precisely you state your request, the more likely the AI tool will produce the output you expect. 
  • Provide examples. If you want the output in a specific format or style, consider providing AI with an example, either directly within the prompt or as uploaded context. Then, ask it to create something similar according to your specifications. 
  • Iterate for revisions. When AI’s response is on the right track, but not quite what you want, ask it to revise rather than starting over. Be specific about the improvements you expect and avoid vague statements like “make this better”. 
  • Know when to start over. If your prompt is about a new topic that doesn’t reference previous exchanges, start a new chat instead of continuing the current conversation. This is more energy-efficient and can prevent the AI from conflating multiple topics. 

These recommendations are sufficient for most simple, single-use prompts. For highly complex tasks such as creating simulations or executing multi-step processes, we recommend exploring prompt libraries from Wharton Generative AI Labs or AI in Education, or requesting a CETLI consultation to help you get started. 

Prompting Framework: RICO

For those who prefer a more structured approach, you can apply the recommendations above by prompting AI with a Role, Instructions, Context, and Output (RICO). 

Start by giving the AI a specific persona or scenario to take on. This helps characterize its response within an appropriate topic, tone, or style. The more specific information you provide, the better. For example:

  • “You are an expert professor of organic chemistry…”
  • “You are developing a course on clinical interaction practices…”
  • “You teach a first-year undergraduate political science class…”

Tell the AI the core task you want it to accomplish. Be as specific as you can, but know that this step and the following step may blend together. For example:

  • “Design an in-class activity…”
  • “Generate five case studies…”
  • “Write an executive summary…”

Provide as much context to your instructions as possible. Context can include details like intended audience, background information, or examples of the desired output. You may describe these within the prompt or add uploaded files, such as your lecture notes, slides, or course materials that are not under copyright and do not contain personal information. For example: 

  • “The in-class activity should be for 30 students and take approximately 20 minutes to complete. The learning objectives are…”
  • “Each case study should focus on a pediatric patient ages 5-12 and include current vitals, medical history, laboratory results…”
  • “Your summary should highlight the main ideas in the attached PowerPoint presentation…”

Specify the format of the output. Depending on the tool you are using, you may request both text and multimedia elements. For example:

  • “Provide step-by-step directions for the activity, including the estimated amount of time for each step.”
  • “Include a 100-word clinical note and a bulleted list of test results.”
  • “Generate a timeline of important events and a line graph depicting the population change over time.”

Applying the RICO Framework

When you put together all four components of the RICO framework, you may end up with a prompt such as these:

  • “You are a psychology professor teaching a graduate course on human cognitive development. Write five course learning objectives, each using a verb from Bloom’s Taxonomy. Format the objectives as a numbered list.”
  • “You teach a first-year undergraduate writing seminar. Brainstorm 10 strategies to help your students practice synthesizing information from multiple sources. Create a bulleted list of strategy ideas that includes both in-class and at-home activities.”
  • “You want to send a reminder message to all students enrolled in your online software engineering course. Draft a friendly email that summarizes 2-3 key takeaways from the attached lecture notes and reminds students about the midterm exam due this Friday.”

This framework is only a starting point. Remember that you can always iterate, or ask AI to revise, to improve upon its initial output. You should also plan to review and edit the output yourself before utilizing it in teaching or communication.

CETLI Can Help

CETLI staff are available to discuss ideas or concerns related to generative AI and your teaching, and we can work with your program or department to facilitate conversations about this technology. Contact CETLI to learn more.

Generative AI

More Resources

Center for Excellence in Teaching, Learning and Innovation