By Hailey Dickson, Product Marketing Manager at Khan Academy
A common misconception around AI is that it will just go and do the work/study/learning for us. The inverse is actually true: we only get out of it what we put in.
To effectively learn using AI, we must understand how to communicate with it using text prompts.
“A prompt for an AI is a specific instruction or question given to the AI to generate a desired response or output. It helps provide context and guidance for the AI to understand what is expected of it.” (Khanmigo)
Learning how to create effective prompts is crucial for several reasons. First, well-designed prompts generate more accurate responses. Second, carefully crafted prompts can foster originality in AI-generated content, enabling the exploration of new ideas and perspectives. Lastly, mastering the art of writing good prompts empowers students and teachers to effectively utilize AI technology and achieve meaningful outcomes.
This guide will help you create effective AI prompts for any platform, though the examples we use will come from our new AI tutor called Khanmigo. Our primary objective is to provide world-class AI for education. Therefore, it is essential to learn how to create effective AI prompts that will aid learning. In this blog, we give general guidance and examples of good and bad prompts that will help you create effective AI prompts for your learning journey.
Prompt-Writing Tenets
- Know the purpose of your prompt.
Before you begin generating prompts, it’s crucial to understand their intended use. Will the prompt facilitate learning or test your knowledge? The purpose of the prompt will dictate the type you create. For example, if the goal is to assess your knowledge, you’ll want to generate a quiz prompt or multiple-choice format. If the aim is to support learning, consider using open-ended questions that encourage critical thinking and problem solving.
- Keep the prompt simple and clear.
A well-crafted prompt should be concise and unambiguous, leaving no room for confusion. It should clearly state the problem and provide precise instructions for responding. Ambiguous prompts may lead to incorrect or irrelevant answers.
- Use context, examples and ask for specific output
Including context and examples in your prompt makes it more effective. Context provides additional information about the prompt, making it easier for the user to understand the question.
Examples provide a visual representation of the problem, making it easier for the user to grasp the concept. Using context and examples in your prompt increases the user’s engagement and understanding.
Be specific about what your desired output is; including length, format and tone.
- Offer feedback.
Giving feedback after a prompt is essential because it helps the AI evaluate its understanding of the problem. Feedback can either be positive or negative. Positive feedback encourages the AI to continue down the same path; negative feedback informs the AI of its incorrect response and helps it understand how to solve the problem correctly. Adding feedback to your prompt will improve your learning experience.
Good Prompt Examples
Each of these prompts offers a different challenge and encourages the AI to generate relevant and effective outputs.
- Scenario-based prompt
“Imagine you are a time traveler from the 28th century. Describe the technological advancements that exist in your time.”
- Persuasive prompts
“Write an argument in favor of renewable energy sources compared to fossil fuels, highlighting their long-term sustainability and overall benefits.”
- Creative prompts
“Create a short story about a robot that feels human emotions and has to navigate the complexities of emotional intelligence.”
- Informative prompts
“Explain the process of photosynthesis in simple terms.”
- Discussion prompts
“Discuss the ethical implications of artificial intelligence in decision-making roles.”
Bad Prompt Examples
To create effective prompts, it is also important to understand what makes a prompt bad. A prompt can be ineffective if it is too long, too ambiguous, or too complex for the AI. Additionally, using jargon or technical language can make the prompt difficult to understand.
- Ambiguous prompts
“Talk about it.”
- Overly complex prompt:
“Describe the sociopolitical implications of quantum mechanics in the context of 21st-century geopolitics.”
- Overly broad prompt:
“Discuss life.”
- Prompt with insufficient context:
“Write a story.”
Each of these prompts presents various challenges for an AI. They are either too vague, too complex, too broad, or lack context, which can lead to underwhelming or off-target responses from the AI. By understanding the limitations of these examples, we can craft clear, engaging, and effective prompts that facilitate meaningful AI interactions.
You can learn even more about effective prompting in our new video series on Khan Academy: Practical AI for Instructors and Students
Practice Your Prompt Skills on Khanmigo
Creating effective AI prompts is fundamental to optimizing the learning process when using AI-powered applications like Khanmigo. A well-designed prompt should be simple and clear, and it should have an intended purpose. Providing context, examples, and feedback will improve the user’s learning experience and increase engagement. Avoiding bad prompt design will prevent confusion or disinterest on the user’s end. Follow these general guidelines, make good use of the examples presented in this blog, and you’ll be on your way to creating effective and engaging prompts that will aid your AI-powered learning journey.