How to Have Multiple AI Personas Discuss Together?

Question Type

User Manual Question

Quick Start (3 Steps)

Step 1: Create Discussion Project

Entry Point:

Homepage → "Create New Interview"
Select "Discussion Chat (Group Discussion)"

Fill in Project Information:

Project Name: e.g., "Sparkling Coffee Pricing Discussion"
Research Objective: Brief description of topic you want to explore

Step 2: Select AI Personas (3-8 People)

Recommended Quantity: 5-6 people

Selection Principles:

✅ Ensure viewpoint diversity (different stances)
✅ Different backgrounds and consumption psychology
❌ Avoid selecting people with similar viewpoints

Example: Test sparkling coffee pricing (5 people)

2 price-sensitive types (students, working class)
2 quality-first types (white-collar, high-income)
1 social trendsetter type (Gen Z, Xiaohongshu users)

Step 3: Choose Discussion Type and Start

3 Discussion Types:

1. Focus Group - Most Common

Use For:

Product concept testing
User attitude research
Feature acceptance

Characteristics:

AI moderator guides discussion
Free expression of viewpoints
Seek consensus and divergence

2. Debate

Use For:

Solution comparison (A vs B)
Controversial topic exploration
Clear pro and con positions

Characteristics:

Divided into pro and con sides
Opposing viewpoint collision
In-depth argumentation

3. Roundtable

Use For:

Exploratory research
Open-ended exploration
No clear stance

Characteristics:

Equal speaking rights
No preset stance
Multi-angle insights

Click "Start Discussion"

Discussion Execution Process

AI Moderator's Role

1. Opening and Guidance:

AI Moderator: Hello everyone, today our discussion topic is "Sparkling Coffee Pricing."First, do you think ¥18 sparkling coffee is expensive?

2. Control Speaking Pace:

Actively call on speakers: "XX, what's your view?"
Invite silent ones: "XX hasn't spoken yet, what's your opinion?"
Control chatterers: "Thank you XX for sharing, let's hear others' thoughts"

3. Guide Viewpoint Collision:

AI Moderator: I notice your price views differ greatly,A thinks it's expensive, B thinks it's not, can you explain why?

4. Deep Dive into Controversies:

AI Moderator: Everyone has different views on the "sparkling + coffee" combination,some think it's innovative, others think it's a gimmick, why?

5. Summarize Consensus and Divergence:

AI Moderator: It seems everyone acknowledges the 0-sugar concept (consensus),but price acceptance varies greatly (divergence).

During Discussion

Real-Time Viewing:

Monitor discussion progress in real-time
See each persona's statements

Mid-Way Adjustments:

Insert new questions
Adjust discussion direction
Extend or end discussion

Estimated Duration:

Standard: 30-40 minutes
In-depth discussion: 50-60 minutes

Detailed Real-World Cases

Case 1: Focus Group - Sparkling Coffee Pricing (5 People)

Persona Mix:

Linda (28, Operations, price-sensitive)
Emma (32, PM, health-conscious)
Chloe (26, Designer, social trendsetter)
Sarah (30, Finance, rational decision-maker)
Mia (29, Marketing, brand loyalist)

Discussion Flow:

Opening:

AI Moderator: Do you think ¥18 sparkling coffee is expensive?Linda: Too expensive, I usually buy ¥12 AmericanoEmma: I think it's okay, 0-sugar is very attractiveChloe: Price is acceptable, key is appearanceSarah: Need to see value, how does it compare to competitors?Mia: ¥18 isn't expensive, but this brand lacks recognition

Deep Dive into Divergence:

AI Moderator: Why do your price views differ so much?Linda: My budget is limited, ¥18 exceeds daily budgetEmma: I focus more on health, 0-sugar worth paying extraChloe: I don't care about price, if I can post on Xiaohongshu it's worth it

Explore Consensus:

AI Moderator: What do you think of the 0-sugar concept?Emma: 0-sugar attracts me, but worried about sweetener safetySarah: I also focus on 0-sugar, but need to see ingredients listMia: 0-sugar is a trend, this selling point is good(Consensus: 0-sugar concept is attractive, but need to address sweetener concerns)

Key Controversy Point:

AI Moderator: Do you think "sparkling + coffee" is innovation or gimmick?Chloe: I think it's novel, willing to tryLinda: Sounds weird, worried about bad tasteEmma: Will bubbles irritate stomach?

Discussion Results:

Consensus: 0-sugar concept is attractive
Divergence: Price acceptance varies greatly (depends on income and consumption psychology)
Controversy Point: Is "sparkling + coffee" combination innovative
Key Insights:
- Need to address sweetener concerns (label "natural sweetness")
- Price positioning must be clear (¥15-18 range, avoid ¥20+)
- Packaging design must look good (attract social trendsetters)

Case 2: Debate - Subscription vs One-Time Purchase (6 People)

Persona Mix:

Pro (support subscription): 3 people
- Small business decision-maker (focus on flexibility)
- Individual user (don't want to pay too much at once)
- Heavy user (continuous use, subscription more cost-effective)
Con (support one-time purchase): 3 people
- Mid-size business decision-maker (focus on total cost)
- Light user (infrequent use, one-time more economical)
- Privacy-sensitive (don't want to bind credit card)

Discussion Flow:

Opening:

AI Moderator: Today we discuss subscription vs one-time purchase, which is better.Pro side believes subscription is superior, con side believes one-time purchase is superior.First, pro side representative please speak.

Pro Arguments:

Pro-Small Business: Subscription is flexible, don't want to pay ¥5000 at oncePro-Individual: Monthly ¥99 acceptable, one-time ¥2999 too expensivePro-Heavy User: I use daily, subscription 2 years only ¥2376, cheaper than one-time

Con Arguments:

Con-Mid-Size Business: We use 3 years, one-time ¥2999 lower total costCon-Light User: I use few times monthly, monthly ¥99 subscription not worth itCon-Privacy-Sensitive: Subscription requires credit card binding, I don't want to bind

Confrontation:

AI Moderator: Both sides have valid arguments, now please debate freely.Pro-Heavy User: One-time has no updates, subscription has continuous updatesCon-Mid-Size Business: We don't need frequent updates, stability is goodPro-Small Business: Subscription can be canceled anytime, one-time has high riskCon-Light User: Subscription easy to forget to cancel, actually spend more money

Discussion Results:

Consensus: Both models have demand, cannot choose one only
Pro Advantages: Flexibility, continuous updates, low barrier
Con Advantages: Lower total cost, no binding, stability
Key Insights:
- Offer both options (subscription + one-time)
- Promote subscription to heavy users, one-time to light users
- Subscription offers "20% discount for annual payment" to reduce total cost concerns

Case 3: Roundtable - Explore Fitness App Needs (5 People)

Persona Mix:

Fitness beginner, advanced fitness enthusiast, weight loss group, senior fitness user, rehabilitation trainer

Discussion Flow:

Opening:

AI Moderator: Today we freely discuss fitness app needs and pain points.What problems do you encounter using fitness apps?

Free Discussion:

Beginner: My biggest problem is incorrect form, worried about injuryAdvanced: I think app features too basic, not professional enoughWeight Loss: I want to see real-time calorie burn, but app doesn't have itSenior: Interface too complex, I don't know how to useRehabilitation: My knee is injured, app has no suitable training plan for me

Moderator Guidance:

AI Moderator: Seems everyone's needs are very different,beginners focus on ease of use, advanced users focus on professionalism.If app should prioritize solving one problem, what should it be?

Discussion Results:

Consensus: Form correction is essential need (beginners and seniors both mentioned)
Divergence: Beginners want simple, advanced want professional
Key Insights:
- Layered design: Beginner mode + Advanced mode
- Core feature: Form correction AI
- Differentiation: Personalized training plans (consider injuries)

Common Questions

Q1: Can discussion duration be controlled?

Yes. You can set:

Minimum duration: e.g., 30 minutes (ensure sufficient discussion)
Maximum duration: e.g., 60 minutes (avoid lengthy discussion)
Auto-judge: AI moderator identifies "viewpoints fully collided" then auto-ends

Q2: Can I add or replace people mid-discussion?

No. Cannot replace personas after discussion starts.

Reasons:

Discussion is continuous, mid-way replacement disrupts flow
New persona lacks context, hard to integrate

Recommendations:

Confirm persona selection beforehand
Ensure persona diversity

Q3: What if a persona never speaks?

AI moderator will actively invite:

"XX hasn't spoken yet, what's your opinion?"
Ensure everyone has opportunity to express

Q4: Will discussion deviate from topic?

AI moderator will control:

Identify deviation, pull back promptly
Ensure discussion focuses on core issues

You Can Also:

Insert new questions mid-way
Adjust discussion direction

Best Practices

Recommendation 1: Ensure Viewpoint Diversity

Don't:

❌ 5 price-sensitive users (no collision)

Do:

✅ 2 price-sensitive + 2 quality-first + 1 brand loyalist
✅ Ensure opposing viewpoints for discussion value

Recommendation 2: Interview First, Then Discussion

Best Process:

First use Interview for deep exploration (5-8 people)
Discover key controversy points
Then use Discussion to observe viewpoint collisions (3-5 people)

Advantages:

Discussion topics more precise
First understand individuals, then observe groups

Recommendation 3: Use Focus Group to Test Acceptance

Scenario: New product concept testing, feature acceptance

Advantages:

Quickly discover consensus and divergence
Observe viewpoint evolution during discussion

Recommendation 4: Use Debate to Test Solutions

Scenario: A/B solution comparison

Setup:

Pro: 3 people support A
Con: 3 people support B
Observe which side argues more fully

Final Takeaway

"Let AI personas discuss together, observe viewpoint collisions, Quickly discover consensus, divergence, and controversy points."

Remember:

✅ 3 steps to start: Create project → Select personas → Choose discussion type
✅ Ensure viewpoint diversity (different stances)
✅ Best practice: Interview deep dive first + Discussion viewpoint collision later

Related Feature: Interview vs Discussion Document Version: v2.1