Being specific

Core Concept

• Being specific means providing guidelines or steps to direct the model toward a particular output
• Without specificity, models can go in infinite directions (varying length, adding/removing elements, changing structure)

Two Types of Guidelines

Type A: Output Attributes

• List qualities or characteristics the output should have
• Control aspects like length, structure, or specific attributes
• Example: "Keep story under 1000 words, include rising action, add at least one supporting character"

Type B: Process Steps

• Provide specific steps the model should follow
• Forces the model to think through particular considerations
• Increases output quality by guiding the reasoning process
• Example: "First brainstorm three talents, pick the most interesting, outline a revealing scene, consider supporting characters"

When to Use Each Approach

Always use Type A (attributes):

• Recommended for virtually every prompt
• Provides baseline guidance on desired output qualities

Use Type B (steps) when:

• Working on complex problems
• You want to force consideration of wider perspectives
• Additional topics or data points might otherwise be overlooked
• Example: Analyzing why sales dropped requires considering multiple viewpoints and data sources

Practical Application

• Both techniques can and should be mixed together in professional prompts
• Adding specificity significantly improves output quality (demonstrated by score improvement from 3. 92 to 7. 86)
• Specificity helps ensure you get the output you actually want

Takeaways

• Being specific in prompts means providing guidelines or steps to direct the model toward particular outputs, preventing it from going in infinite directions
• Use Type A (output attributes) to control length, structure, and qualities—recommended for virtually every prompt
• Use Type B (process steps) for complex problems to force consideration of multiple perspectives and prevent overlooking important data points
• Combining both techniques in a single prompt significantly improves output quality
• Specificity ensures you get the output you actually want rather than leaving it to the model's interpretation