Creating Clarity: Techniques for Writing Effective AI Prompts

Crafting effective prompts for large language models (LLMs) like ChatGPT is both an art and a science. A well-constructed prompt can transform a vague query into a precise, insightful, and actionable response, making it an indispensable skill for users seeking to maximize the capabilities of AI-driven chatbots. Understanding the nuances of prompt engineering can significantly enhance the quality and relevance of the AI’s output. This guide delves into essential tips and techniques for writing prompts that yield the most accurate and useful results from LLMs like ChatGPT.

Talk to the AI, Don’t Programme It

Large Language Models like ChatGPT are designed to emulate human speech/writing patterns and simulate human-like reasoning. So prompting them with tasks or questions is more like informing them of your needs than programming it. 

Imagine the AI is a human assistant. You wouldn’t just tell a person to “write a piece about effective AI prompts” – or if you did you couldn’t expect good results. The more information you give the AI the better the results will be. And, like a person, if it doesn’t have the information it needs it may interpolate (or ‘guess’) something and get it wrong. 

Provide Loads of Context

With the above point in mind, you’ll get far better results from your prompts if you consider what information a person would need to help you with your request. Asking ChatGPT; “how can I start learning woodworking as a hobby?” will produce a generalised answer. 

But if your prompt was more like; “I am interested in starting woodworking as a hobby but have never done any woodworking before. I have a small workbench and some basic DIY tools. I would like to be able to make my own furniture one day. How can I best get started and develop my skills?” then you’ll get a much more nuanced and helpful answer. 

Always remember that, however intelligent they seem, ChatGPT and other Large Language Models work on probability – which words are most likely to go with each other. Context helps the model do this with better accuracy, and also ensures that it when you ask a question about router bits the answer relates to woodworking tools and not components for networking devices. 

Outline the Output

A critical part of this context is telling the AI what the purpose of the task or question is and what form you want the output to be. Do you want suggestions or a framework for a piece of work you’ll be writing yourself? Ideas for a task or event? Or are you wanting the AI to generate a complete text? 

Constraints can also include the format of the piece – a numbered list, a blog, an essay or a table – and the tone. Do you want an authorial, expert tone or something chatty and informal? ChatGPT will also absorb tone from your prompts and replies, so over a lengthy session of questions, tasks and responses its output will more and more resemble the style of your prompts. Bear this in mind. 

Give the AI a Role to Perform

A very effective way of combining all the above without using up too many words (or your mental energy) is to cast your prompts in the form of giving the AI a role and setting it a task – it will form its responses within that framework If you’re a marketing manager and you’re after ideas for a client’s strategy, tell the AI that and it should provide much more specific and relevant answers. 

If you’re a teacher looking for fresh ideas for engaging students on a subject, tell the AI the grade and size of the class as well as the subject, and prompt it to take on the role of the teacher. It will give quite different outputs if it’s set the same task with different roles – in fact, assigning the AI different roles within the overall task can be a great way of getting a varied insight. 

Get the AI to Check Your Input

The underlying principle of writing good AI prompts is appreciating what the model needs to provide the best possible answers. And, while it may seem oddly recursive, the best way to understand this is to ask the AI itself. 

Write your prompt with all the information and context that you think is relevant. End the prompt by asking the AI to ask you any questions it needs to clarify or expand on the information. That may well reveal areas you’ve missed. Combine the suggestions from the AI with your original prompt and you’ll be sure of getting high-quality results. 

Leave a Comment