Quick guide to prompt mastery

Learn how to correctly construct your prompts for the best performance.

Alright, let's dive into the art of crafting the perfect prompt for our AI Chats System. Whether you're a seasoned NDIS Support Coordinator or just starting out, mastering the art of prompt creation can significantly enhance your experience and efficiency. Here are some tried-and-true strategies to get the best results from our AI.

Write Clear Instructions

First things first, clarity is key. The AI isn't a mind reader (yet!), so the more specific you are, the better the results. If you need concise answers, ask for them. If you want detailed, expert-level responses, specify that too. The less the AI has to guess, the more likely you’ll get what you need.

Tactics:

  • Include details in your query to get more relevant answers.

  • Ask the model to adopt a persona.

  • Use delimiters to clearly indicate distinct parts of the input.

  • Specify the steps required to complete a task.

  • Provide examples.

  • Specify the desired length of the output.

Example:

Instead of saying, "Help me with a case note," try, "Help me write a case note for a participant who has just started a new therapy program. Include details about the therapy, the participant's initial reactions, and any next steps."

Provide Reference Text

If you have specific information that the AI should use, provide it! This helps the AI avoid fabricating details and ensures the accuracy of the response.

Tactics:

  • Instruct the model to answer using a reference text.

  • Instruct the model to answer with citations from a reference text.

Example:

Provide the text: "The participant started a new occupational therapy program focusing on improving fine motor skills." Then ask, "Using the provided text, write a case note summarising the first session."

Split Complex Tasks into Simpler Subtasks

Break down complex tasks into simpler, manageable steps. This reduces error rates and makes the AI's job easier.

Tactics:

  • Use intent classification to identify the most relevant instructions for a user query.

  • Summarise long documents piecewise and construct a full summary recursively.

Example:

Instead of asking, "Summarise the entire NDIS plan," try, "First, summarise the participant's goals. Then, summarise the funding allocations. Finally, summarise the support services provided."

Give the Model Time to "Think"

Sometimes, the AI needs a moment to reason through a problem. Asking for a "chain of thought" can help it arrive at more accurate conclusions.

Tactics:

  • Instruct the model to work out its own solution before rushing to a conclusion.

  • Use inner monologue or a sequence of queries to hide the model's reasoning process.

  • Ask the model if it missed anything on previous passes.

Example:

Ask, "Explain the steps you took to determine the best support plan for this participant, and then provide the final recommendation."

Test Changes Systematically

Improving your prompt's performance is easier if you can measure it. Make modifications and test them against a set of examples to ensure you're getting better overall results.

Tactic:

  • Evaluate model outputs with reference to gold-standard answers.

Example:

Create a test suite of common tasks you perform and see how different prompts affect the AI's performance. Adjust accordingly to find what works best.

And there you have it! Mastering these strategies will make your interactions with the AI Chats System smoother and more effective. Experiment, have fun, and remember, clear and detailed prompts are your best friends. Happy prompting!

Did this answer your question?
😞
😐
😁