Prompt Engineering – Categories vs Strategies
- Prompt Strategies: Form how the questions are asked to the model & use different techniques to guide AI responses. Examples of prompt strategies include few-shot or direct prompting.
- Prompt Categories: These are what is asked, and relate more to the content & scope of the question. Examples of categories include open-ended questions or informational requests.
You might imagine them in levels:
Level 1 (Top of Mind): General, top-of-mind questions as prompts to the model.
Level 2 (prompt categories): Testing different categories of questions such as “Describe the benefits of Product X” (instructional) vs “How does Product X compare to Product Y?” (comparative)
Level 3 (prompt strategies): Testing a direct question with no examples (zero-shot) vs the results from giving a question & a few examples (few-shot)
Prompt Categories:
Open-Ended Prompts: Designed for detailed, expansive responses without specific correct answers.
Example: “Describe the impact of your latest marketing campaign.”
Closed-Ended Prompts: Require specific, brief responses, often yes/no.
Example: “Did the recent ad increase customer engagement?”
Reflective Prompts: Encourage introspection on thoughts, feelings, or experiences.
Example: “Reflect on the effectiveness of your last social media strategy.”
Informational Prompts: Seek direct facts or information.
Example: “What are the key features of Product X?”
Creative Prompts: Spark creativity and imaginative responses.
Example: “Imagine a world where your product solves the biggest industry challenge. Describe it.”
Instructional Prompts: Provide clear, precise directions or tasks.
Example: “Outline the steps for executing a successful email marketing campaign.”
Comparative Prompts: Invite comparison or evaluation between items or ideas.
Example: “Compare the ROI between traditional advertising and digital advertising.
Exploratory Prompts: Stimulate curiosity and discovery without a set endpoint.
Example: “Explore potential markets for launching our new product line.“
Scenario-Based Prompts: Create hypothetical situations for problem-solving or decision-making.
Example: “As the marketing manager facing budget cuts, how would you reallocate resources to maintain campaign effectiveness?”
Opinion-Based Prompts: Ask for personal views or evaluations on topics.
Example: “What is your perspective on the use of influencer marketing in building brand awareness?”