For this guideline we suggest that you:

• stay on point,
• stay true to your topic and your ultimate goal,

so the tool generates relevant and purposeful output for you and your situation.

Also try to keep it real and authentic for the tool. It is trained on data sets after all, so you will need to provide it with language that it can understand. 

Make sure you stay true to your overall OUTPUT GOAL. To do this you can do the following:

• provide some context, background.
• provide an intent.

You may also want to provide a short example or examples, to provide the tool with the right direction you want it to take. Something like

• "similar to ....."
• "you can model the information on..."  

Give it a mission – this should include the: 

• task — what you want it to perform – write, brainstorm, create, explain, summarise, list etc.
• format / type of output you want the tool to generate: email, paragraph, blog post, tweet, FaceBook post, bulleted list, table, flow chart, synopsis, outline, letter, image etc.

This relates to the MISSION. For this guideline, you should include:

• very precise and specific particulars and details.
• very precise and specific instructions of what you want the output to include.

You may also want to include:

• relevant sections or parts.
• Boolean logic: you can draft a prompt in the same way you would construct a search strategy. For more guidance on the use of Booleans, please have a look at this book.

This can encompass the following:  

• the kind of attitude and spirit you want the tool to adopt: For text output you can use clear language, concise and to the point, succinct, academic, friendly. For image output this would be a little different: the tone could be ultra realistic image; highly realistic image of.... evoking a sense of cinematic grandeur etc.
• your target audience – who you want the output to cater to.

When you have completed the PROMPT elements, don’t forget to evaluate your prompt: this may include refining and fine tuning:

• checking for spelling / grammar errors,
• checking whether you have given it a persona, stayed on point and true to your goal and provided enough context, particulars and details,

to allow the tool to fulfil its mission and produce the content you are after!  

After you have formulated your prompt you need to deploy it in the AI tool you want to use. You will essentially be having a conversation or chat with the tool! It is a chatbot after all and there will exist a convergence of human factors and robot / machine - like factors. The end result of such convergence, is the creation of an output (either written, visual or audio) that is right for you! You may need to extend the chat with the bot, if you don’t succeed the first time. Here are some tips if the bot doesn’t produce what you are after:  

• ask the bot to extrapolate or expand on points you feel require further attention or explanation.
• experiment with the phrasing you have used.
• try follow – up prompts.  
• ask follow – up  questions relating to the output.  

Also, the chat bots are very polite! They may: 

• ask you to "provide a more descriptive prompt".
• let you know that a prompt was too "vague" to generate content.
• tell you that they "don’t have enough context to understand what you’re referring to".
• ask you to try a "longer, more descriptive prompt".
• ask you to "please provide more details or clarify your request". 

On the upside, they are always happy to assist you!

You can use these PROMPT-ED principles and guidelines and apply them to the creation of your own outputs, to convey your ideas. Here is an example of a prompt that outlines the PROMPT guidelines of PROMPT-ED and the respective output. 

The information provided in this section of the Guide should be enough to allow you to create prompts to generate output for your purposes and needs. For more advanced information on prompting / prompt engineering, you can have a look at the following information:

• OpenAI document on prompting engineering
• Vanderbilt Universities article on prompt patterns
• This article from The Conversation
• This article from the Australian Law Librarian

Fig 1. Image generated using Microsoft Edge's Copilot, powered by DALL·E from the prompt "create an ultra realistic image depicting an artificial intelligence assistant multitasking. make the assistant look like an android. add some futuristic elements to the scene".

 

You can also explore the following LinkedIn Learning Courses that focus on prompting:

• Introduction to Prompt Engineering for Generative AI – By Ronnie Sheer 
• Prompt Engineering with ChatGPT – By Ronnie Sheer 
• Prompt Engineering: How to Talk to the AIs – By Xavier Amatriain 
• Advanced Prompt Engineering Techniques – By Morten Rand-Hendriksen

Fig 2. Image generated using Microsoft Edge's Copilot, powered by DALL·E from the prompt "you are a well renowned artist. you have been commissioned to create a piece of artwork, for the Artificial Intelligence Institute. they have asked you to create some artwork that will help to convey a chatbot that really enjoys answering questions online. you can model the chatbot on IBM'S WATSON. the image should look highly realistic and evoke a sense of cinematic grandeur, but look like an ancient artefact from the future".

You can also consult prompt websites:

• Prompthero
• Promptvibes