Create effective prompts
Create effective prompts
On this page, we will introduce a framework for prompt engineering consisting of 5 principles called the CLEAR principles. The purpose is to optimize our interactions with AI language models, such as ChatGPT, and serve as a guide for creating better prompts.
With these principles, you can learn how to create prompts that yield better responses, which are more relevant and of higher quality. Perhaps even more importantly, as a student, you can strengthen your AI information literacy.
In short, prompt engineering is about constructing queries or inputs (prompts) that generate the most relevant responses. It can also be seen as the art of fine-tuning the questions we ask to optimize the quality of the answers we receive.
Improve your prompt engineering with 5 principles
- 1
Be concise
Remove superfluous information to make the AI model focus on the most important aspects of the task. If what you write is unclear or imprecise, you are giving unclear instructions. Your input should be specific and clear, directed towards the desired response.
Guidelines for concise prompts:
- Consider what you want information about
- Remove superfluous information
- Use precise and clear language
- 2
Be logical
Make sure that your input (prompt) has a logical flow. A logically structured prompt makes it easier for the AI language model to understand context and relationships between concepts, resulting in more precise and coherent responses.
An example of a logically structured prompt could be: 'Describe the various steps in the scientific method, where the first step is to formulate a hypothesis and the last step is to draw conclusions.'
One rule of thumb for logical prompts:
- Is your input logically structured?
- 3
Be explicit
Explicit prompts provide specifications or precise instructions about the desired output. These specifications can relate to scope, output format, or content. This will reduce the likelihood of receiving unexpected and irrelevant responses from the AI model.
Instead of writing: 'Tell me about renewable energy,' you can write: 'Identify five sources of renewable energy and briefly describe how each one works.'
Guidelines for explicit prompts:
- Are all important details included?
- Are you specific regarding scope and format? Do you want a detailed explanation or a concise overview?
- What type of information are you seeking?
- How would you like the information to be presented to you?
- 4
Be adaptive
Be ready to adjust your prompts. If the response you receive is not what you expected, try again in a different way. For example, instead of asking, 'What impact does the use of social media have on mental health?' you can ask: 'What is the relationship between social media use and depression in young people?'
Your prompts will direct the model in a specific direction, just like a conversation between two people. When you write "what is the connection..." you automatically imply that there is a connection! If you are in doubt about whether there is a connection or not, you should ask "Is there a connection...", and follow up with more specific questions.
Be aware that outputs from generative AI tools are based on previous inputs you have provided and the output the model has generated. This is why you can start a 'new conversation' or thread. It may be useful to organize your interactions with the system into groups or create new groups when you need to break away from previous prompts and output.
Guidelines for adaptive prompts:
- Do not give up, but rephrase, use different words, or ask in a different way
- If the answer is too general, make your question more focused
- If the answer is too generic, make it more specific
- If the answer provides too much irrelevant information, be more specific and explicit about what you want
- 5
Be reflective
Continuous evaluation will improve your prompts. With a reflective perspective, we are better equipped to stay ahead and adapt to the constant development within the field – what works today may not work tomorrow.
Guidelines for reflective prompts:
- Make it a habit to analyse the responses you receive so that you can improve
- Examine the accuracy, relevance, and completeness of the responses
- Use what you learn along the way to make your prompts even better next time
Final reminder! A prompt can be long. Even though the input field may seem small, there is room for a lot of text. So, you are not limited to writing just one or two lines. The size of the field is due to design reasons.