3. Common Mistakes in Prompt Writing

Lesson 1: Ambiguity in Prompts

1.1 The Problem of Vague Prompts

A common mistake in prompt writing is leaving prompts too vague, which leads to unclear and unsatisfactory responses. Without specific details, the AI struggles to understand the task’s objective, resulting in overly broad or unrelated answers.

Example of a Vague Prompt:

Describe technology.

• Problem: What kind of technology? Computers, smartphones, AI? The scope is too wide.

Improved Version:

Describe the advancements in smartphone technology over the past five years, focusing on improvements in camera quality and battery life.

• Improvement: Narrowed the topic to smartphones and specified areas of focus (camera quality and battery life).

1.2 Flow of Ambiguity and Resolution

graph TD;
    A[Vague Prompt] --> B[Unclear Objective];
    B --> C[Unfocused Response];
    C --> D[Poor User Experience];
    A --> E[Refine Prompt];
    E --> F[Clear Objective];
    F --> G[Targeted Response];

1.3 How to Avoid Ambiguity

• Be Specific: Always define the topic, subtopics, or desired scope.
• State Expectations: Indicate what form the response should take (e.g., explanation, list, story).
• Provide Context: Use relevant background information to clarify the task.

---

Lesson 2: Overly Broad Prompts

2.1 The Problem of Scope

Another common mistake is creating prompts with too broad a scope. Broad prompts make it difficult for the AI to know where to start, often resulting in generic or surface-level responses.

Example of a Broad Prompt:

Explain AI.

• Problem: AI is a vast field. The response could range from machine learning to neural networks, making it difficult to cover all aspects meaningfully.

Improved Version:

Explain how machine learning is used in AI, with a focus on supervised learning techniques and real-world applications.

• Improvement: This version narrows the focus to a specific area (machine learning) and asks for details (supervised learning and real-world examples).

2.2 Effects of Overly Broad Prompts

Overly broad prompts often lead to:

• Unstructured Responses: AI may jump between unrelated topics.
• Lack of Depth: Responses are too shallow or superficial.

2.3 How to Fix Broad Prompts

• Set Clear Boundaries: Define the scope and limit the subject matter.
• Ask for Specific Examples: Narrow down the task to avoid generalizations.

graph TD;
    A[Too Broad Prompt] --> B[Generic Response];
    B --> C[Lack of Depth];
    A --> D[Narrow Focus];
    D --> E[Targeted Response];

---

Lesson 3: Lack of Constraints

3.1 The Importance of Constraints

When prompts lack constraints, the AI might generate responses that are too long, irrelevant, or off-topic. Constraints help guide the model to a more precise and useful response.

Example of a Prompt with No Constraints:

Write a story about a detective.

• Problem: Without length, style, or theme constraints, the response could be unpredictable or unstructured.

Improved Version with Constraints:

Write a 300-word detective story set in a futuristic city. Focus on the detective’s ability to solve crimes using advanced technology.

• Improvement: Adds constraints on length, setting, and theme, guiding the model to produce a more relevant response.

3.2 Types of Useful Constraints

• Word Count: Specify the length to manage the scope of the response.
• Tone or Style: Indicate the desired tone (e.g., formal, casual, creative).
• Subject Focus: Guide the AI to specific aspects of a topic.

3.3 Code Example: Implementing Constraints in a Prompt

Here’s a Python snippet using OpenAI’s API to apply constraints:

import openai

def generate_response(prompt, max_tokens=150):
    response = openai.Completion.create(
        engine="text-davinci-003",
        prompt=prompt,
        max_tokens=max_tokens,
        temperature=0.7  # Creativity control
    )
    return response.choices[0].text.strip()

# Example prompt with constraints
prompt_text = "Write a 200-word article explaining how solar panels work, with a focus on photovoltaic cells."
print(generate_response(prompt_text, max_tokens=200))

---

Lesson 4: Overly Complex Prompts

4.1 The Pitfall of Complexity

When prompts are too complex or contain multiple tasks, it can overwhelm the AI, leading to incomplete or confused responses. Breaking tasks into manageable chunks makes it easier for the model to produce coherent answers.

Example of a Complex Prompt:

Describe the history of artificial intelligence, including key advancements, influential people, and modern applications, in 500 words.

• Problem: The prompt asks for too much information in a limited space, which may result in a disorganized response.

Improved Version:

Describe the key advancements in artificial intelligence over the past decade, in 300 words.

• Improvement: Focuses on advancements in the last 10 years and limits the word count to avoid an overwhelming scope.

4.2 How to Simplify Complex Prompts

• Break Down Tasks: If the task is complex, split it into separate, smaller prompts.
• Focus on One Aspect at a Time: Address one topic or concept in each prompt.

Example of Breaking Down a Complex Task:

Instead of:

Write about AI history, its key figures, and applications.

Break it down into:

1. Prompt 1: "Describe the key advancements in AI history."
2. Prompt 2: "Who are some influential figures in AI development?"
3. Prompt 3: "What are some modern applications of AI?"

graph TD;
    A[Overly Complex Prompt] --> B[Multiple Tasks];
    B --> C[Incomplete Response];
    A --> D[Split into Simpler Prompts];
    D --> E[Focused and Clear Response];

---

Lesson 5: Ignoring the Audience

5.1 Mismatched Prompts and Audience

A common mistake is writing prompts without considering the audience’s knowledge level or expectations. If the prompt is too advanced for a beginner audience or too simple for experts, the response may not be useful.

Example of Ignoring Audience:

Explain quantum computing in detail.

• Problem: This might be too advanced for beginners and doesn’t specify the audience.

Improved Version for Beginners:

Explain the basic concepts of quantum computing in simple terms for someone without a background in physics or computer science.

5.2 Tailoring Prompts for the Audience

• Consider Knowledge Level: Adjust the complexity of the language and content based on the audience’s familiarity with the topic.
• Use Relevant Examples: Relate the content to what the audience cares about or understands.

---

Conclusion

Common mistakes in prompt writing—such as ambiguity, overly broad tasks, lack of constraints, complexity, and ignoring the audience—can result in poor responses. By refining prompts to be specific, clear, and audience-tailored, you can significantly improve the quality of AI-generated content.

---

References

1. OpenAI Documentation on Prompt Engineering: Best Practices for Writing Prompts
2. Effective Communication in AI: John Doe, AI Interaction Guide, 2023.