“Factorization of Prompts (5) mini” — Prime Factorization of F — The Three Factors Composing Format: “Style,” “Structure,” and “Length” [Column] Science of Thought (No. 84)

🎯 The Aim of This Column (3 Key Points)

• Experience how specifying “F = Format” changes the shape and tone of an AI's response.

• Learn that to convey “how you want the answer to be presented,” the format can be broken down into three factors: “Style,” “Structure,” and “Length.”

• Understand that format design, as an “exit design” for your inquiry, is the key to tuning your dialogue with AI according to your purpose.

◆ Review of Previous Topics:

In No. 80, we learned that the prompt formula can be expressed as follows:

$${Prompt=P×R×C×F}$$

And here are the four factors that make up this formula:

• $${P}$$ = Purpose: What do you want it to do?

• $${R}$$ = Role: Who do you want it to answer as?

• $${C}$$ = Context: What is the topic?

• $${F}$$ = Format: In what form do you want the answer?

In No. 81, regarding “$${P}$$ = Purpose” in the prompt formula “$${Prompt = P × R × C × F}$$,” we identified three factors as the smallest units: “Understanding,” “Judgment,” and “Action.”

$${P (Purpose) = P_1 (Understanding) × P_2 (Judgment) × P_3 (Action)}$$

In No. 82, we captured “$${R}$$ = Role” using three factors: “Knowledge Level,” “Stance,” and “Social Perspective,” and confirmed that specifying these significantly changes the AI's tone and persona.

$${R (Role) = R_1 (Knowledge Level) × R_2 (Stance) × R_3 (Social Perspective)}$$

In No. 83, we took up “$${C}$$ = Context” and learned how to structure the “design of the topic,” or “what the question is about.”

$${C (Context) = C_1 (Domain) × C_2 (Subject) × C_3 (Situation)}$$

This time, in No. 84, we will learn about the final factor that composes a prompt, “$${F}$$ = Format,” and consider the three prime factors that serve as the “blueprint for output.”

◆ Does AI expression change depending on “the form in which it answers”?

For example, let's consider a prompt like this.

ChatGPT's expected response example:

In this way, ChatGPT will provide a neutral and concise summary of the facts.
However, if you specify the “format” as follows, the structure and tone of the answer will change completely.

ChatGPT's expected response example:

Alternatively, there is also this format.

ChatGPT's expected response example:

Just by specifying “in what format to answer,” ChatGPT's output changes from a mere list of information into a “form that fits the intended purpose.”

That is the power of
$${F = Format}$$.

◆ Prime Factorization of F: Format is made of three elements

“$${F}$$ = Format” in a prompt is the “design blueprint for output” that allows ChatGPT to understand “in what form it should answer.”
By clearly specifying the format, the AI's response becomes not just a list of information, but an expression that fits the reader.

This “format” can also be broken down into the following three factors:

$${F_1}$$ = Expression Style (Style)
┗ Specifications regarding tone, vocabulary, and writing style, such as gentle/academic/casual/polite language, etc.

$${F_2}$$ = Structure
┗ The “shape” of the output, such as bullet points, paragraphs, comparison tables, or story-telling formats.

$${F_3}$$ = Length & Constraints
┗ Specific volume specifications, such as within 200 characters, in 3 lines, in one sentence, or enough to be read in 5 minutes.

By multiplying these three elements, ChatGPT can grasp “what to answer, in what style, at what length, and in what structure.”

If expressed like a mathematical formula:

$${F (Format) = F_1 (Style) × F_2 (Structure) × F_3 (Length & Constraints)}$$

For example, if you specify “casual tone × bullet points × 3 points,” ChatGPT will return a response like a light blog post.
Conversely, if you say “academic tone × paragraph structure × about 800 characters,” it will change to an academic and logical output.

Specifying the format in a prompt is exactly like designing the “vessel” for the output.
Only when the format is designed does information take on a “form that conveys.”

◆ Let’s look at prompt examples

When you clarify the elements that compose “$${F}$$ = Format” (expression style, structure, constraints),
how much does the “appearance” or “way of conveying” of ChatGPT’s output change?
Let’s compare with actual examples.

🔍 Prime factors of F included in the example:

This prompt contains the following format settings (prime factorization of $${F}$$):

• Expression Style ($${F_1}$$) = In a friendly tone

• Structure ($${F_2}$$) = Bullet points with 3 items

• Constraints ($${F_3}$$) = Within 200 characters

In other words, expressed as a formula:

$${F = F_1 (friendly tone) × F_2 (bullet points, 3 items) × F_3 (within 200 characters)}$$

In a prompt constructed this way, it becomes clear to ChatGPT “in what form it should respond,” and it can accurately adjust the style and volume of the output.

🅰️ In the case of an ambiguous prompt

🅰️ Example Answer:

📌 The information is correct, but because there are no instructions regarding sentence structure, tone, or length, some readers might find it 'a bit wordy'.

🅱️ In the case of a structured prompt (explicit F):

🅱️ Example Answer:

📌 Since the expression style, structure, and length are defined from the start, it is output as a 'finished product' suitable for the intended use (e.g., SNS posts, internal sharing, printed materials, etc.).

In this way, simply by structurally designing the $${F}$$ (Format) in a prompt,
ChatGPT's response changes from a 'list of information' to a 'document tailored to the purpose'.

Being conscious of the output format is
the act of conveying to the AI 'how you want it to speak'—in other words, designing an 'interface of expression'.

◆ Adjusting F is 'structural design' of the output

Specifying $${F}$$ = Format in a prompt
is equivalent to the act of designing the 'structure of the output' for the AI.

For example, even with the same content, just changing the format can dramatically change the impression and usability of the answer.

$${F_1 (Key point bullet points) × F_2 (Comparison structure) × F_3 (For readers)}$$
→ A clearly organized checklist-style summary

$${F_1 (Script format) × F_2 (Optimized for video length) × F_3 (For SNS)}$$
→ A short-form script for TikTok or Instagram

$${F_1 (Explanatory text for diagrams) × F_2 (Procedural) × F_3 (For beginners)}$$
→ An easy-to-follow tutorial article for anyone

In short, a prompt is also a 'blueprint' for designing how the answer is 'conveyed'.
The meaning of information changes depending on 'how it is told'.

By consciously designing $${F}$$, the AI's output evolves from 'information' to 'expression'.

◆ Finally, please imagine for a moment

When you ask ChatGPT something,
what are you conveying about what topic, from what perspective, and for what situation you want it to speak?

For example, that topic is—

• which field it belongs to,

• what it targets,

• and to whom and in what situation it should be conveyed.

When the “background” of the question is clarified in this way,
it becomes more than just a request for information;
it becomes an expression of your own 'way of choosing thoughts'.

By clarifying the context, the question deepens, and the answer becomes richer.

This column was written in co-creation with generative AI, ChatGPT. It is an attempt to strengthen the design of thought, with the proposition of 'what to make the topic'.

🔜 Next Preview: 'Prompts are an “Interface for Thought”'

In the next installment (No. 84), we will provide a summary of the 'Prompt Factorization Studies' series so far.

We have previously broken down the four elements of 'Purpose ($${P}$$)', 'Role ($${R}$$)', 'Context ($${C}$$)', and 'Format ($${F}$$)', and clarified the prime factors inherent in each.

Finally, next time, we will present the overall picture that integrates these and deepen our structural understanding of 'Prompt = Blueprint for Inquiry'.

• Why does dialogue with generative AI sometimes 'not click'?

• How can you accurately convey your thoughts to AI?

And in the final chapter, we will view 'prompt design' as the budding of a new academic discipline and redefine it as an 'interface for thought' that supports your own creation of questions.

💡 What kind of question would you like to 'design'?

Please look forward to it.

Profile | Kazuomi Matsunaga

🧭Thinking Navigator

Biography:Graduated from Kyushu University, Faculty of Engineering (Materials Engineering) ⇒ Engineer at a metal materials manufacturer in the Chukyo region (Production Engineering) ⇒ Advertising sales for a local free paper in Nagasaki Prefecture ⇒ Educational institution staff (← currently here). Living in Nagasaki Prefecture, in my 50s. Both of my children are now university students who have left the nest, and I live with my wife.
I am a writer who enjoys playing with thoughts daily at the intersection of “questions” and “curiosity.” Recently, I have been writing about my own unique perspectives on themes such as education, storytelling, psychology, AI, and design, centered around the serialized column series “The Science of Thought.” I am exploring “free design of knowledge” that transcends existing frameworks. My motto is “Imagination opens up the world.”
While working in my main profession in university and education-related fields, I value carefully observing the moments when “people start to like something” and the scenes where “questions sprout.”
🧭 I would be happy if I could deliver articles that help a “new perspective” sprout within you after reading them.
📮 Impressions and messages of support are always welcome!