Predict what will happen.
Understand why.
Simulate any population with social networks and temporal dynamics. Then interview individual agents about their reasoning. The only platform that combines quantitative simulation with qualitative interrogation.
NYC Congestion Pricing · 500 agents · t=5/10
Position Distribution
Exposure: 78% · Avg conviction: 0.45 · Shares: 34
Network propagation active · 3 clusters identified
Why did you change your mind about the tax?
I was open to it at first — less traffic in Midtown sounded great. But when my neighbor Carlos showed me the cost breakdown — $15 a day, that's $300+/month — I reconsidered. Then Priya at work said she'd just avoid Manhattan entirely. It felt like punishing outer-borough commuters instead of fixing the subway.
What would change your mind?
How It Works
Three steps. Complete understanding.
No other platform combines quantitative simulation with qualitative interrogation and continuous calibration.
Simulate
Build a population. Run a scenario.
Describe your population in natural language. doxia.ai discovers attributes, grounds them in real statistics, samples agents, connects them in a social network, and simulates how they respond to your scenario over multiple timesteps.
"500 New York commuters, ages 25-65"
12 attributes discovered · Statistical grounding: strong
Support: 34% | Oppose: 48% | Neutral: 18%
Interrogate
Interview any agent. Understand why.
Click on any agent in your results. Ask them why they made their choice. They respond grounded in their persona, their simulation history, the network influences they received, and their conviction trajectory. No other platform offers this.
Why did you switch to oppose?
"My neighbor Carlos showed me it'd cost $15/day. Then Priya said she'd avoid Manhattan. It felt like punishing outer-borough drivers."
Grounded in: 7 timesteps of simulation history, 3 peer influences, 2 position changes
Calibrate
Compare prediction vs. reality. Improve.
After implementation, upload real outcome data. See where the simulation was accurate and where it missed. The system recalibrates its models automatically. Every study makes the next one more accurate.
Predicted
48% oppose
Actual
52% oppose
Error
4pp
Model recalibrated. Next prediction: +12% accuracy.
The Killer Feature
Post-simulation interviews
See that 12% of your users would churn? Click on one. Ask why. They answer grounded in their persona, their simulation history, and the social influences they received. Nobody else offers this.
Maria Santos
Agent #042 · Marketing Manager
Position Trajectory
Key Influences
Say-Do Gap
Public: Firmly oppose
Private: Some environmental sympathy
Why did you change your mind about congestion pricing?
Honestly, I was open to it at first. Less traffic in Midtown sounded great. But when Carlos next door showed me the actual cost breakdown — $15 a day, that's over $300 a month just for driving to work — I started to reconsider. Then Priya told me she'd just avoid Manhattan entirely, which would hurt the restaurants I love. It started feeling like the city was punishing outer-borough commuters instead of improving the subway.
Would a public transit credit change your position?
That's actually the kind of thing that would make me reconsider. If they offset the fee with a MetroCard credit, I'd feel like the city is investing in solutions, not just taxing problems. I know David from my online group supports it for exactly that reason — he takes the L train, so it doesn't hit him. But for drivers from Brooklyn like me, we need real alternatives first.
Capabilities
Everything you need to understand your audience
Five layers of technology working together. From statistically grounded populations to deep agent interviews and continuous learning.
Statistical Population Grounding
Every attribute distribution sourced from real-world data — census, surveys, published research — with citations. Upload your own CRM or survey data for proprietary grounding.
Social Network Simulation
Small-world topology with weighted edge types (partner, coworker, neighbor). Conviction-gated sharing with flip resistance. Opinions propagate like they do in reality.
Two-Pass LLM Reasoning
Pass 1: agent reasons in character. Pass 2: structured outcome extraction. Eliminates central tendency bias that plagues single-pass approaches.
Say-Do Gap Modeling
Every agent has public and private positions. What they say vs. what they'd actually do. The gap where all the value lies in boycotts, politics, and pricing.
Post-Simulation Interviews
Chat with any agent after the simulation. They respond grounded in their full simulation history, network influences, and conviction trajectory.
RAG-Grounded Populations
Upload CRM data, past surveys, or interview transcripts. The system calibrates population distributions from your proprietary data instead of public statistics.
Temporal Evolution
Multi-timestep simulation with sliding-window memory. Watch attitudes shift over time as information propagates and agents update their beliefs.
Continuous Calibration Loop
Compare predictions against real outcomes. The system tracks accuracy by segment and recalibrates automatically. Every study makes the next one more accurate.
Validation
12 studies. 9 pass. Public artifacts.
The only reproducible public benchmark in the synthetic research market. Every spec, seed, and config is publicly available.
Netflix Password Sharing
Bud Light Boycott
Spotify Price Hike
NYC Congestion Pricing
+ 8 additional studies (6 pass, 2 miss). Full results and methodology at doxia.ai/blog
Use Cases
Every decision. Pre-tested.
Simulate first. Interview the results. Decide with confidence.
Pricing Strategy
Product ManagerSimulate 1,000 customers reacting to a 20% price increase. See 12% would churn. Interview the churners to understand what retention offer would keep them.
Public Policy
Policy AnalystModel 5,000 constituents responding to a congestion tax. Track how opposition propagates through social networks. Interview swing voters to design better messaging.
Product Launch
Growth LeadSimulate early adopter networks for a new feature. Identify which user segments become evangelists and which become detractors. Interview each to optimize onboarding.
Campaign Messaging
Political StrategistTest 3 different campaign messages on simulated voter populations. See which creates the most persuadable movement. Interview undecideds to understand what resonates.
Crisis Response
Communications DirectorSimulate how a product recall propagates through customer networks. Identify high-influence critics. Interview them to design response messaging before the crisis hits.
Market Research
Strategy ConsultantReplace a $50K focus group with a 2,000-agent simulation. Get distributional outcomes, network dynamics, and deep interviews — in hours instead of months.
Pricing
10-50x cheaper than traditional research
A $50K focus group or a $199/month simulation with deeper insights. You choose.
Starter
For individual researchers and evaluators
- ✓5 studies/month
- ✓100 agents max per study
- ✓3 timesteps max
- ✓5 agent interviews/study
- ✓Community support
Pro
For product teams and small consultancies
- ✓20 studies/month
- ✓1,000 agents per study
- ✓Unlimited timesteps
- ✓Unlimited interviews
- ✓RAG data upload
- ✓API access
- ✓Email support
Team
For mid-market teams and agencies (up to 5 seats)
- ✓50 studies/month
- ✓5,000 agents per study
- ✓Unlimited everything
- ✓Batch interviews
- ✓Compare studies
- ✓Team collaboration
- ✓Priority support
Enterprise
For large consultancies and policy organizations
- ✓Unlimited studies
- ✓50,000+ agents
- ✓SSO / SAML
- ✓SOC 2 compliance
- ✓Dedicated infrastructure
- ✓Custom model training
- ✓White-label available
- ✓Dedicated CSM
All plans include population simulation, social network modeling, and temporal evolution. Higher tiers add interviews, RAG, calibration, and team features.
Full Transparency
Open methodology. Auditable results.
Every simulation is fully inspectable. Population specs, attribute distributions with sources, agent reasoning traces, network propagation paths, and conviction trajectories. No black boxes. Every result can be audited and reproduced.
Learn About Our MethodologyStop guessing. Start simulating.
doxia.ai is launching in 2026. Get early access to the platform that predicts how populations respond to any scenario — then lets you interview them about why.
Simulate populations. Interview the results. Calibrate with reality.