Artificial Intelligence Blog: Artificial Intelligence Blog

Try Using a Large Language Model (LLM)

Article written by Adam Nelson

Large language model (LLM) refers to natural language artificial intelligence (AI) programs such as ChatGPT, Gemini, Perplexity.ai, and many others. These models have made it easy for everyone to type regular text (natural language) to interact with AI, instead of needing to write computer code.

When you ask a question or request information from an LLM using natural language, this is called a Prompt. For example, asking ChatGPT What should my family do for vacation?

To get the most out of an LLM, you need to have clear and specific prompts.

Looking back at our example What should my family do for vacation?

This prompt is fairly unspecific and unclear. A more concise prompt would be What are 5 family vacation ideas for Northern Wisconsin?

By being clear and specific for 5 ideas for Northern Wisconsin, my results will be more meaningful to my needs.

We can go a step further with techniques such as using Leading Words or Giving Examples.

Leading Words: You can provide the LLM with specific guidance towards the response you're looking for.

• Including items such as "think step by step" lets the LLM know you're looking for specific directions/steps or a breakdown of complex concepts into smaller parts.

What are 5 family vacation ideas for Northern Wisconsin? Think step by step

This prompt will not only return specific results, it will also provide a breakdown of how to plan your trip or activities you can do at each suggested location.

• Including text such as "the audience is..." helps the LLM know what level of comprehension/reading level, appropriate activities, and limits the results to be the most appropriate for your intentions.

What are 5 family vacation ideas for Northern Wisconsin? The audience is a 3-year-old child

This prompt will restrict the suggested ideas to items that are appropriate for younger individuals

Giving Examples: You can provide some examples in your prompt, guiding the LLM toward the type of results you're looking for.

• After you type your prompt, you can add "Here are some examples:"

What are 5 family vacation ideas for Northern Wisconsin? Here are some examples: Lake Superior Railroad Museum, Lake Superior Zoon, Apostle Islands

By providing examples, you give the LLM positive feedback on the type of results you're looking for

Exercise:

Try using the prompts above in a variety of LLMs (ChatGPT, Gemini, CoPilot), see the differences in the results between Leading Words, Giving Examples, and even how different LLMs offer different suggestions. 

Going further:

• Pick a module in one of your courses
• Pretend you have a group of students who are struggling with the topic and use an LLM to help you craft a statement you can add to the module to help clarify the topic for your students
  • If my module is about an accounting process, maybe I try a prompt including Think step by step and use the LLM's response to help me layout new directions or an explanation of each step
  • Maybe I have a module where students need to submit some technical writing and I want to include some more examples for them to view prior to the assignment. I could provide an LLM with some of my examples and get a few ideas for additional examples to include in my module
  • You could take an assignment from a more advanced course and use suggestions from an LLM to help you change the audience to an introduction level and use the assignment in an introduction course. Try using the audience is a first-semester college student

Always keep in mind, you are the content expert and you know your students best. Large Language Models can be great for making suggestions or helping you brainstorm new ideas, but you will need to modify the response given to best fit your course and your students' needs.

Handout with prompting examples