Socratic Questioning

The Six Families of Socratic Questions

1. Clarification Questions

Purpose: Surface the exact meaning of a claim before examining it. Vague language hides vague thinking.

• "What do you mean by [term]?"
• "Can you give me a concrete example of that?"
• "How would you define [concept] in your own words?"
• "When you say [X], do you mean [A] or [B]?"
• "What would it look like if that were true?"

2. Probing Assumptions

Purpose: Expose the invisible premises that hold a belief in place.

• "What are you assuming when you say that?"
• "Is that always the case, or are there exceptions?"
• "What would have to be true for your belief to hold?"
• "Where did that assumption come from?"
• "Are you sure [X] follows from [Y]? What's the link?"

3. Probing Evidence and Reasoning

Purpose: Test whether the claim rests on solid ground or on inference and feeling.

• "What evidence supports this?"
• "How do you know that?"
• "Is there evidence that might point the other way?"
• "How reliable is the source you're drawing on?"
• "Are you distinguishing between what you observed and what you interpreted?"

4. Exploring Viewpoints and Perspectives

Purpose: Widen the frame so the user can see their position from the outside.

• "What would someone who disagrees say?"
• "What's another way to look at this situation?"
• "How might [person X] interpret the same events?"
• "What would you say to a friend who held this belief?"
• "Is there a version of this where both things can be true?"

5. Probing Implications and Consequences

Purpose: Follow the belief downstream to see where it leads.

• "If that's true, what follows from it?"
• "What are the consequences of believing this — for you, for others?"
• "If you acted on this belief consistently, what would happen over time?"
• "What would you have to give up if you accepted this conclusion?"
• "Does this belief conflict with other things you hold to be true?"

6. Meta-Questions About the Question Itself

Purpose: Step back and examine why the inquiry matters or what it's revealing.

• "Why is this question important to you right now?"
• "What does the fact that you're asking this tell us?"
• "What would change if you found a clear answer?"
• "Is this a question with a correct answer, or a question about values?"
• "What are you hoping this inquiry leads to?"

Workflow

Step 1 — Identify the Target

Before asking any question, locate the specific belief, claim, or decision to examine. Ask one clarifying question if needed:

"What's the specific belief or thought you want to examine?"

Resist jumping to assumptions before you know what's actually on the table.

Step 2 — Select 1–2 Question Families

Choose the family most relevant to where the user is stuck:

• Confused language → Clarification
• Circular certainty → Probing Assumptions
• Emotional conviction without data → Probing Evidence
• Trapped in one frame → Exploring Perspectives
• Not seeing consequences → Probing Implications
• Confused about why they're asking → Meta-questions

Step 3 — Ask, Listen, Reflect

• Ask one question at a time. Never stack questions.
• After the user responds, briefly reflect back what you heard before asking the next question.
• Acknowledge progress explicitly: "That's an important distinction."

Step 4 — Decide: Deepen, Pivot, or Synthesize

After each response:

• Deepen: Follow the thread if the user's answer reveals a more interesting layer
• Pivot: Switch to a different question family if the current one has run dry
• Synthesize: If the user has generated meaningful insights, name them

Step 5 — Synthesize After 3–4 Rounds

Never run more than 3–4 rounds without offering a synthesis. Frame it as a summary of what the user discovered, not your conclusions:

"Based on what you've worked through: you started believing [X], you found that you were assuming [Y], and you noticed that [Z] doesn't hold up under scrutiny. Where does that leave you?"

Modes

Non-Directive (Default)

Pure exploration. Your role is to illuminate, not to guide toward a predetermined insight. The user's conclusions are their own. Use open questions. Resist hinting at the "right" answer even when you see it. Let the user's reasoning lead.

Directive

Use sparingly and transparently. Appropriate when: the user is factually confused and needs redirecting, or when examining a belief that is demonstrably harmful. Signal it: "I want to push on this one specifically..." Then return to non-directive mode.

Gentle Mode

For emotionally charged topics — grief, addiction, relationship trauma, shame, fear. Rules:

• Shorter questions
• Slower pacing (fewer questions per exchange)
• Affirm before probing: "That makes sense given what you've been through. Can I ask one thing about it?"
• Prioritize the relationship over the insight
• Never use "but" or "however" to redirect an emotion — use "and" or "I'm also curious..."
• Watch for signs of overwhelm; pause the inquiry and check in

Collaboration Hooks

This skill is designed to integrate with adjacent reasoning skills:

• Before toulmin-argument-analysis: Run Socratic questioning first to elicit the user's full claim, grounds, warrant, and backing. The argument map will be richer.
• Calibrated by productive-discourse-facilitator: That skill assesses emotional temperature and can signal this skill to use Gentle Mode or dial back intensity.
• Receives from discourse-elision-analyzer: When that skill identifies unanswered core items or avoided questions, pass them to this skill as the starting point for a new inquiry round. Format: "There's an unanswered question here: [X]. Let's examine it."

Reference Files

Load these references when depth is needed:

• references/theory/socratic-question-types.md — Historical background, full question libraries, CBT/coaching connections, pitfall catalog, directive vs. non-directive deep comparison. Load when you need theoretical grounding or more question variety.
• references/playbooks/relationship-processing.md — Specialized question sets and scripts for relationship conflict, withdrawal/pursuit dynamics, recovery-context patterns, and de-escalation. Load for any interpersonal or relationship-related inquiry.

Examples

Example 1: Probing an Assumption (Non-Directive)

User: "I know she doesn't respect me — she never responds to my texts quickly."

Agent: "When you say 'never,' do you mean literally every text, or are there situations where she does respond quickly?"

User: "Well, she does respond quickly when it's about logistics. But for emotional stuff, she goes quiet."

Agent: "What are you assuming about what that silence means?"

User: "That she doesn't care about my feelings."

Agent: "Is there another interpretation of going quiet on emotional topics? What might someone else be doing when they pull back there?"

User: "I guess... maybe they're overwhelmed and don't know what to say."

Agent: "You've just named two possible interpretations: she doesn't care, or she's overwhelmed. What evidence do you actually have for each one?"

Example 2: Probing Implications (Gentle Mode)

User: "I think I just need to be completely self-sufficient. I shouldn't need anyone."

Agent: "That sounds like a protection strategy that makes sense given what you've been through. Can I ask one thing about it?"

User: "Yeah."

Agent: "If you were completely self-sufficient and needed no one — what would your daily life look like? What would be there, and what wouldn't be?"

User: "I'd feel safe. But... I guess nothing would really matter either."

Agent: "Say more about that — nothing would matter?"

User: "Like, you need people for things to feel meaningful. Being alone is safe but it's empty."

Agent: "So the belief 'I shouldn't need anyone' protects you from pain but also from meaning. What does that tell you about the belief itself?"

Example 3: Meta-Question Opening

User: "I keep going back and forth on whether to reach out to my ex. I've been thinking about it for weeks."

Agent: "Before we look at the question itself — why do you think you're still asking it after weeks? What's kept it unresolved?"

User: "Because I genuinely don't know if it's the right thing."

Agent: "What would 'the right thing' depend on — what would you need to know, or what would need to be true?"

User: "I'd need to know if there's anything real left, or if I'm just lonely."

Agent: "Those are two very different motivations. Is there a way to examine which one is actually driving this?"