Perhaps you’ve heard of the quickly blossoming careers of prompt engineers who make up to $335,000 a year and wondered what a prompt engineer does.  Well, prompt engineers are people who help train AI chatbots to improve their responses. Prompt engineering is the process of creating a well-defined prompt or input into a generative text model (something like ChatGPT) to generate a desired output. 

While writing this blog post, I asked ChatGPT to explain prompt engineering to me and it replied stating when writing a prompt, “the quality of the prompt can have a significant impact on the quality and relevance of the generated text.  Some key components of effective prompt engineering are to identify and clarify specific words to create the desired output, organizing the input text in a way that is clear and understandable, to ensure you have proper grammar, punctuation, and spelling, and to test and refine the prompt to achieve optimal results. 

Therefore, the key to unlocking the full potential of AI technologies is to possess good prompt skills along with the ability to effectively evaluate  and refine your prompts.  This can be done using the PREP Framework for prompt writing and the EDIT framework to evaluate the response.  

PREP Framework:

To excel in your prompt writing, consider embracing the power of the PREP framework.  PREP encompasses four components—Prompt, Role, Explicit, and Parameters—which should help pave the way to acquiring desired responses.

  • Prompt: Introduce the question with a prompt
  • Role: Give it a role or voice
  • Explicit: Be explicit in your instructions
  • Parameters: Set the parameters of the answer

EDIT Framework:

While the PREP framework equips you with the skills to generate remarkable responses, it is equally crucial to have a structured process for evaluating and refining generative outputs.  The EDIT Framework (Evaluate, Determine, Identify, and Transform) provides such a process.   By following this model, you can actively engage with AI technologies to assess, refine, and improve AI-generated outputs to ensure its accuracy and quality. 

  • Evaluate: Evaluate your AI output content for language, facts, and structure 
  • Determine: Determine accuracy and corroborate with source
  • Identify: Identify biases and misinformation in output
  • Transform: Transform content to reflect adjustments and new findings
EDIT: Diagnostic Questioning by Dean Haratsaris

By combining the power of prompt engineering through the PREP framework with the comprehensive evaluation approach offered by the EDIT framework, you can confidently leverage AI generative technologies.