Many in the educational community have a great deal of fear and concern regarding how ChatGPT and AI in general are used in the classroom by students. I believe that one of the ways that this fear and concern can be alleviated is to start to learn and use ChatGPT for educational tasks and build comfort with this new and growing technology.

I have recently begun the process of converting many of my older rubrics into state standard aligned rubrics in order to build better learning data for my classroom. This has been a slow process as I manually write a descriptor for each level of the rubric per item. For the purposes of this article I will be focusing on only of set of standards for the discussion at hand. I have been working for the past year on building descriptors for each of the state benchmarks for my program that I teach. I then take these benchmark descriptors and use them to build rubrics for various types of assessment in my classes. This frankly can be a tedious task to build all these different descriptors, I knew that there had to be a better way to build these items for my assessment. My thoughts turned to better ways to accomplish this task using the various AI technologies that are available today. Though this thought process I realized that most of my descriptors were fairly formulaic in how they were created. This lead me to a work on a methodology that would consistently produce quality rubric criteria descriptors.

Writing Better Prompts for AI

Through creating numerous prompts for AI systems to generate a specific information in a given format. I have used this to create rubric descriptors in an eight point format that my school uses. I have found that writing a useful prompt contains three portions that give you the most useful and repeatable results. These three are:

• Criteria:
• Specific Information:
• Output Formatting:

As we begin to write prompts for AI, I think it is also important to keep in mind that you typically won’t get the best results on your first try and that you will likely need to refine your prompt two or three times before getting solid answer or document created. As you write more prompts for AI systems (text generators, image generators, etc.) it with the practice and understanding what results come from what prompts that we get better at writing prompts. As you are writing the three different portions (criteria, specific information and output formatting), it is important to know keep in mind they don’t have to put in a liner order and that ordering these pieces differently in the prompt to generate different results. When we talk about criteria, these are items that must be included in the resulting output. Specific informaiton is exact facts, figures or other exact information that you are looking to have included. This portion can be the hardest to get correct in the prompt since getting the information you want out of the prompt can be a challenge while you are learning. Finally we have Output Formatting which is a written description or descriptor of how you want the outputted information to look for example in a table or a list or a certain number paragraphs.