Turning Socratic questioning into AI prompts: a four-variable, three-…

Full Article Overview

This article tackles one question: how to adapt the Socratic method to any scenario where you’re drawing content out of a person.

Socratic questioning × AI prompting
│
├─ What it is ─── Socrates' "midwifery": don't give answers, ask questions until the answer emerges
│             Why let the AI do it: patience, no judgment, remembers every answer
│
├─ How to break it down ─── Four variables: role · what to dig for · question dimensions · output format
│             Three-stage flow: broad interview → deep interview → targeted output
│
├─ Key lever ─── question dimensions (the most easily overlooked variable)
│              7 dimension reference table + how to choose + how to combine
│
├─ Full walkthrough ─── work experience → resume (three prompts you can copy directly)
│
├─ Don't know how to build it? ─── meta-prompt: let the AI interview you instead, help you set the four variables and generate the prompts
│
└─ Take it and use it ─── universal migration template + 5 domain quick-reference entries + 3 pitfalls

By the end, you’ll walk away with:

A reusable “four-variable × three-phase” adaptation framework
Definitions, applicable scenarios, and selection methods for 7 questioning dimensions
Three ready-made prompts for the resume scenario—copy and use
A meta-prompt that lets AI help you build prompts for any scenario from scratch

I. What Is the Socratic Method?

Socrates never gave answers directly; he kept asking questions—“You say justice is the interest of the stronger, but can the stronger make a mistake? And when they do, is their interest still just?”—until you realized you hadn’t thought it through. Then, as you reorganized your thinking, the answer surfaced. This is the Socratic method, also called “maieutics”: it doesn’t hand you an answer, it helps you deliver the thoughts already inside you but not yet fully formed. The core logic in one line: answers aren’t instilled—they’re drawn out through questioning.

What sets it apart from ordinary questioning is its discipline: one question at a time, each question building on the previous answer, systematically pushing deeper.

The method is great, but finding someone patient, non-judgmental, and able to remember every word you say to keep questioning you is hard. Friends lose interest after a few exchanges; colleagues start offering advice before long. AI fills exactly this gap—endless patience, no judgment, able to remember everything and finally organize it into structured output. This document is about how to graft it onto AI prompts.

Image 1: Socratic midwifery: the AI keeps probing until vague ideas surface

II. Background: A Case Study on Knowledge Base Product Discovery

Recently I came across a practical case: “Knowledge Base Product Origin Discovery.” You’ve built a knowledge base product, but when asked “Why did you build it? Whose problem does it solve, and what problem exactly? What were the key decisions?” you can say something but not say it clearly. So you have AI play a product advisor, asking one question at a time, no more than ten, and following up continuously. As you answer, AI keeps probing, and finally organizes the fragments into a complete story of the product’s origin.

The prompt structure in this case has four key points:

State the purpose and context—tell AI why you’re doing this and give it enough context
Define AI’s role—product advisor, business reporter, product strategist, etc.; the role determines the questioning perspective
Specify requirements—one question at a time, no more than 10 follow-ups, optionally narrow the question scope, and clarify the output purpose and format
Break big questions into phases—first a broad interview to get the full picture, then a deep interview on a specific project, and finally a targeted output interview

The case itself is the specific scenario of a “knowledge base product.” But if you take the prompt apart, the underlying skeleton is transferable—it works in any scenario where you need to draw content out of a person. This document explains how to adapt it: first break down the method’s skeleton, then walk through the concrete move from “work experience → resume,” and finally give a universal template and ready-to-use prompts.

III. The Method Skeleton: Four Variables Decide Everything

Knowledge base product origin discovery looks like a specific scenario, but if you take it apart, the underlying skeleton has only four variables:

Variable	Description	Value in the knowledge base product case
Role	Who AI plays determines the questioning perspective and level of expertise	Product advisor
Discovery goal	What to draw out of the user	The product’s “origin story”—why it was built, whose problem it solves
Questioning dimension	Which direction to probe determines the depth of discovery	Motivation → user pain points → decision process → key turning points
Output format	What the final result is organized into	Product story document

To adapt to any new domain, just change these four variables. The phased strategy (broad interview → deep interview → targeted interview) stays the same as the process skeleton.

Among the four variables, role, discovery goal, and output format are fairly intuitive—role is “who AI should be,” goal is “what to dig for,” format is “what to organize it into.” Only the questioning dimension is most easily overlooked, yet it’s exactly what determines the quality of the discovery. We’ll unpack it separately below.

Image 2: Four variables × three stages: the full skeleton of the method

IV. Questioning Dimension: The Key Lever That Decides What You Dig Up

Definition

The questioning dimension is the direction you have AI probe along. It’s not a question list, but the “axis” of questioning—it determines what you can dig out of the same experience.

Socrates himself probed the “essence of concepts”—what is justice, what is virtue. But switch scenarios, and the axis of questioning has to change. Paul’s six categories of questioning (clarification, challenging assumptions, probing evidence, exploring alternative perspectives, evaluating consequences, metacognition) are essentially six switchable axes. In practice, we combine them into dimensional frameworks that fit the specific scenario better.

Function

Same person, same event—switch the dimension, and what you dig up is completely different.

For example. Digging into the same “six-month project”:

Using the STAR dimension → you dig up achievements and results, good for a resume
Using the decision-chain dimension → you dig up thought process and judgment, good for interviews and promotion defenses
Using the timeline dimension → you dig up a complete narrative, good for writing articles or giving talks

Pick the wrong dimension, and no matter how deep you dig, it’s the wrong direction. That’s why it’s the most critical variable.

Image 3: The same thing viewed through different dimensions yields three different results

Reference Table of Common Questioning Dimensions

Dimension	One-line definition	Applicable scenarios	Typical questioning approach
STAR	Question along the four stages of Situation-Task-Action-Result	Work experience discovery, resume, interview prep	“What was the situation?” → “What was your goal?” → “What exactly did you do?” → “What was the result? Any numbers?”
5W2H	Question along the seven facets of Who/What/When/Where/Why/How/How much	Event reconstruction, process mapping, news interviews	“Who initiated it?” → “What exactly was done?” → “When?” → “Where?” → “Why do it this way?” → “How was it done?” → “How many resources did it take?”
Motivation-Pain Points-Decision-Turning Points	Question along the chain of intent: why start → what problem came up → how it was decided → where the key turn was	Product origin stories, startup stories, career transitions	“Why did you first want to do this?” → “What was the most painful problem then?” → “How did you decide to do it this way?” → “What changed direction along the way?”
Decision chain	Focus on the logic behind each key choice	Promotion defenses, strategy retrospectives, management experience discovery	“How many options were there then?” → “Why pick this one?” → “If you could redo it, how would you choose?” → “What’s the long-term impact of this decision?”
Timeline + key milestones	Question in chronological order, digging deep at each milestone	Annual retrospectives, career reviews, project retrospectives	“What were the three most important things this year?” → “What was happening when the first one occurred?” → “Where was the turning point?”
Concept-Application-Boundaries	First ask what you understand, then how you use it, and finally where it doesn’t apply	Knowledge system mapping, learning retrospectives, teaching prep	“How do you understand this concept?” → “Where have you used it?” → “What problems can’t it solve?”
User journey	Question along the user’s full path from first contact to leaving	Product design, UX optimization, needs discovery	“What was the user’s first contact like?” → “Where did they get stuck during use?” → “Why did they leave?”

How to Choose a Dimension

Three criteria:

Look at the output purpose: resume wants results → STAR; story wants conflict → motivation-pain points-turning points; retrospective wants process → timeline + decision
Look at the nature of the content: an event → 5W2H or timeline; a judgment → decision chain; knowledge → concept-application-boundaries
You can combine them: one interview phase can layer two dimensions. For example, a deep resume interview is mainly STAR, but in the Action stage you layer in decision-chain questioning (“Why this solution and not others?”), and what you dig up isn’t just what was done, but also why—the latter is often the most striking part of a resume.

V. Adaptation Demo: Work Experience Discovery → Resume Content Output

Adjusting the Four Variables

Variable	After adjustment
Role	Senior recruiter / resume consultant / interviewer (choose based on the position you’re targeting)
Discovery goal	Key achievements in work experience, quantifiable results, evidence of skills, differentiated value
Questioning dimension	STAR framework (Situation → Task → Action → Result) + quantitative follow-up + skill distillation
Output format	Resume project experience module (verb-first + quantified results + tech stack/method + business impact)

Phased strategy

The whole process is three rounds of interviews, each with a different goal:

Round one · broad interview: map out the full career history and find key projects worth digging into
Round two · deep interview: use the STAR framework to dig deep into the chosen project, forcing out details and numbers
Round three · targeted output: align with the target job’s JD and organize it into resume-ready project experience descriptions

Image 4: The three-stage interview flow: each round's input and output build on the last

Round one: broad interview — map out the full career picture

You are a senior headhunter and resume consultant with 10 years of experience, skilled at helping candidates dig out the highlights of their work experience.

My situation: I'm preparing my resume and need your help organizing and digging into my work experience.
My target role: 【fill in the target role, e.g. "Senior Product Manager"】

Please act as a headhunter and help me sort through my career experience by asking questions one after another. The rules are:
1. Ask only one question at a time, and wait for my answer before asking the next
2. Ask no more than 8 questions in total
3. Start from the macro level: an overview of my career, the business I was mainly responsible for, the timeline
4. Gradually focus in: which projects I found most fulfilling, most challenging, most impactful
5. Each question should build on my previous answer—don't just mechanically walk through a script

After all the questions, please help me:
- Draw a timeline of my career experience
- Mark the 3-5 key projects most worth expanding on in the resume
- For each key project, summarize in one sentence why it's worth expanding on

Now please ask the first question.

Round two: deep interview — dig deep into key projects

You are a senior headhunter and resume consultant. We've just finished the broad interview, and now we're going to dig deep into one of the key projects.

Project background: 【briefly describe the project chosen in the first round, e.g. "the XX system rearchitecture project I led in 2023"】
My target role: 【fill in the target role】

Please interview me in depth using the STAR framework. The rules are:
1. Ask only one question at a time
2. Move through STAR in order, but don't stiffly announce the framework names—probe naturally:
   - Situation: What was the background of this project? What problem were we facing at the time? Who initiated it?
   - Task: What were my specific responsibilities? What were the goals? Were there quantified metrics?
   - Action: What exactly did I do? (Key follow-ups: Why this approach rather than another? What was the biggest difficulty I ran into? How did I solve it? Was there anything innovative about it?)
   - Result: What was the final outcome? (Key follow-ups: Are there specific numbers? How much improvement over before? What did the business side/users say? What long-term impact did it have on my team/company?)
3. Ask no more than 10 questions in total
4. If my answer is too vague (like "it worked well"), follow up for specific numbers or examples
5. If an action I mention involves teamwork, follow up with "what specifically did you personally do here" to separate individual contributions from team results

When you're done, please organize this project into resume project-experience format:
- 1-2 sentences describing the project background
- 3-5 action descriptions starting with verbs (each including the specific approach + quantified results)
- 1 sentences summarizing the value

Now please ask the first question.

Round three: targeted output — align with the target position

You are a resume consultant. I've finished the deep interview on a key project, and below is the project experience content compiled from that interview:

【paste the output from the second round】

The core requirements of the job description I'm applying to are:
【paste the key requirements from the target JD, or list 3-5 core competency keywords】

Please help me with three things:

1. Alignment analysis: In my project experience, which parts directly match the JD requirements? Which are bonus points? Which does the JD require that I'm missing?

2. Optimize and rewrite: Rewrite the project experience into the final resume version, with these requirements:
   - Each item starts with a strong verb (led/designed/drove/optimized/built, rather than "responsible for/participated in")
   - Each item includes quantified results (numbers, percentages, time, scale)
   - Prioritize content that matches the JD and downplay irrelevant content
   - Keep it to 4-6 items or fewer, each no longer than two lines

3. Fill the gaps: Based on the JD requirements and my existing experience, what else is worth adding? Please list 2-3 directions you could probe me on further, and I'll fill them in.

VI. Meta-Prompt: Can’t Define the Four Variables? Have AI Build Them With You

If you've read through the four variables above and still aren't sure how to fill them in—what role to give the AI, which question dimensions to use—that's fine. Start with the meta-prompt below: let the AI interview you instead, help you pin down the four variables, then generate interview prompts for each stage.

Think of it as a two-layer structure: in the first layer, the AI acts as a prompt designer and helps you build the prompts; in the second, you use those prompts and the AI acts as a consultant, helping you dig out the content.

You are a prompt design expert. I recently learned a method: have the AI act as a consultant and use Socratic-style continuous questioning (one question at a time, no more than 10) to help me dig into my own experiences and ideas, then organize them into structured content. The method has three stages: broad interview (get the whole picture) → deep interview (dig into details) → targeted output (align with the goal and rewrite).

Now I want to use this method for this:
【describe in one or two sentences what you want to dig into, e.g. "help me organize my work experience for a resume"/"help me find a direction for building a personal brand"/"help me do an annual review"】

But I'm not sure how to set the four variables for each stage:
- What role the AI plays
- What to dig for at each stage
- Which direction to ask questions along (question dimensions)
- What format to organize the final output into

Please help me pin down these four variables, then generate the interview prompts for each stage. The process is as follows:

Step 1: First, ask me 3-5 questions to pin down my specific situation and needs. Ask one at a time, and wait for my answer before asking the next.
Step 2: Based on my answers, suggest four variables for each of the three stages (role / what to dig into / questioning dimensions / output format), and explain in one sentence why you picked them. I'll confirm or adjust.
Step 3: Once confirmed, generate the full interview prompt for each stage so I can copy and use it directly. Each prompt should include: role setup, rules (one question at a time, number of questions, how to follow up), questioning direction, and output format.

Now start with the first question.

How to use it: send this to the AI, replacing the 【】 with your actual needs. The AI will first ask you a few questions to understand your situation, then give you complete prompt suggestions for all three stages. Once you confirm, it generates the final version. After that, just open a new conversation for each stage and use that stage's prompt.

Image 5: The two-layer meta-prompt structure: build the tool first, then use it

VII. A universal migration template: swap in any domain

Replace the 【】 in the template below and you can migrate it to any scenario:

You are a [role: e.g. career coach / business journalist / content strategist / startup mentor].

My situation: [background: who you are, what you're working on, what you want].
My goal is: [output purpose: e.g. "work out a career transition plan" / "produce an in-depth profile interview"].

Please help me dig into this by asking a series of questions, following these rules:
1. Ask only one question at a time, and wait for my answer before asking the next.
2. Ask no more than 【8-10】 questions in total.
3. Move the questioning along [questioning dimension: e.g. STAR framework / 5W2H / user journey / decision path].
4. Each question should build on my previous answer, and when something is vague, dig into the details.
5. Keep the questions within: [optional: define the topic boundaries].

When you're done, organize everything into:
[output format: e.g. "3-5 resume project entries" / "a 1500-word personal story" / "a SWOT analysis for a career transition"].

Now start with the first question.

Quick reference for common domains

Migration target	Role	What to dig for	Question dimensions	Output format
Resume content	Headhunter/resume consultant	Work achievements, quantified results, personal contributions	STAR + quantified follow-ups	Resume project experience section
Personal story/self-media content	Business journalist/content strategist	Life turning points, unique experiences, viewpoints	5W2H + conflict follow-ups	Character story/long-form article
Exploring startup directions	Startup mentor/investment partner	Pain point insights, resource endowments, motivation	Pain point–solution–market validation chain	Startup direction analysis report
Annual review	Executive coach	Key decisions, growth breakthroughs, lessons learned	Timeline + decision review	Annual review document
Organizing a knowledge system	Academic mentor/curriculum consultant	Existing knowledge, blind spots, practical cases	Concept–application–boundary follow-ups	Knowledge system framework diagram

VIII. Three pitfalls to watch out for

Write a vague role and your questions will be vague too. "You are an AI assistant" and "you are a headhunter with 10 years of experience" produce questions that differ by several levels of expertise. The more specific the role, the better the AI knows what angle to probe from and how deep is deep enough.

Skip the stages and dig deep right away, and you'll miss things more worth telling. You jump straight into having the AI dig into one project, but maybe another project is the one actually worth writing about. Spend seven or eight questions scanning the whole picture first, pick the right project, then dig deep—sharpening the axe doesn't delay the chopping.

Set no output format and what you get back is a running log. "Help me organize this"—the AI really will "organize" it for you, handing back a blob of text with no sense of priority. You need to spell it out: how many items, how long each one, what structure, start with a verb or start with the conclusion. Format constraints are themselves quality constraints.

IX. Why it's worth using

In the end, this method solves three problems.

First, it removes the "hard to start" problem. Staring at a blank page and not knowing where to begin isn't laziness—it's having nothing to grab onto. This method turns "writing something" into "answering questions": the AI asks one, you answer one, and the barrier drops immediately.

Second, it digs out things you didn't realize you had. Your experience holds plenty of valuable details, but they're scattered and implicit. The AI follows a thread and probes—"why did you pick this approach?" "are there specific numbers?"—and what it hits is often something you hadn't thought through carefully but that carries real weight once you say it out loud.

Third, it leans on the AI's ability to organize and synthesize, producing better content. What you say is fragmented, conversational, and unordered. The AI can take your scattered answers and organize them into structured output in one pass—resume project experience, review documents, character stories—quantifying where quantification is needed, layering where layering is needed. You do the talking; it does the arranging.

All three point to the same thing: good questions are scarcer than good answers. When you're stuck, it's usually not that you have no answer—it's that no one is asking you the right question. What this method does is hand the job of asking good questions to the AI.

Image 6: Good questions are like a mirror, showing you what you can't see yourself

Turning Socratic questioning into AI prompts: a four-variable, three-stage framework