Table of Contents

1. Dashboard

The Dashboard gives you a snapshot of the platform's current state: total personas, audiences, simulations, and segments. It shows your most recent audiences with persona counts and population estimates, plus recent simulations with their progress. Use it as a quick status check before diving into a specific workflow.

2. Audiences

An audience is a named group of personas that share a set of criteria — demographics, medical conditions, and behavioral traits. Think of it as a saved cohort definition.

The Audiences page lists all your audiences with name, description, persona count, population estimate, and state. From here you can:

• Insights — View cohort-level analytics for the audience
• Personas — Browse the individual personas in the audience
• Delete — Remove the audience (personas themselves are kept)

3. Persona Explorer

Browse all synthetic personas in the system. Each persona has a demographic segment (age bracket, sex, region, income quintile, education, employment) assigned from Census PUMS data, plus specific traits sampled from NHIS and MEPS survey distributions.

Use the filters at the top to narrow by any demographic dimension. The table shows each persona's name, age, sex, location (state), income, education, employment status, and a health badge. Actions:

• Chat — Start a 1:1 conversation with this persona
• Delete — Remove the persona and all associated data (conversations, memories, trait records)

4. Segment Explorer

Segments are the underlying demographic building blocks — every unique combination of age bracket, sex, region, income quintile, education, and employment status. Each has a Census-derived population estimate.

Filter and sort segments to understand population distribution. The "Personas" column shows how many synthetic personas exist in each segment. This is useful for identifying underrepresented segments that may need more persona generation.

5. Build Persona

Create an individual persona with specific attributes. Two modes are available:

Build by Form — Fill in demographics (name, age, sex, region, income, education, employment), medical conditions, health status (overall, mental, BMI, smoking, alcohol), healthcare behaviors (insurance, visits, telehealth), attitudes (trust scales 1-5, cost sensitivity), spending (annual totals by category), and lifestyle (time use, social patterns). Only name, sex, and age bracket are required — everything else is optional and will be sampled from survey distributions if left blank.

Build by Query — Describe the persona in natural language (e.g., "Maria, 47, female, from Cleveland, makes $72k, bachelor's degree, employed, has diabetes and hypertension"). The AI extracts structured fields and shows them for confirmation before creating.

6. Build Audience

Define a cohort by combining filters. Two modes are available:

Build by Form — Select age brackets, sex, regions, income quintiles, education levels, employment statuses, medical conditions, and trait filters (cost sensitivity, trust in medicine). Set a target persona count to auto-generate new personas if existing matches fall short, or leave it blank to use all matching personas.

Build by Query — Describe the audience in natural language (e.g., "Midwest Diabetics 45-64 — people aged 45-64 in the midwest with diabetes"). The AI parses your criteria and shows the extracted filters for confirmation.

Example presets are provided as starting points — click any example card to pre-fill the form.

7. Audience Insights

Select an audience to view cohort-level analytics. Two tabs are available:

Insights — Aggregated statistics across six categories: health status, healthcare behaviors, spending, attitudes, lifestyle, and social. Each field shows the mean/median, distribution, and survey citation. Filter by category to focus on what matters for your use case.

Personas — Browse the individual personas within the selected audience. Same table format as Persona Explorer, scoped to the audience.

8. Run Simulation

Simulations send a prompt to every persona in one or more audiences. Each persona responds in-character based on their full profile — demographics, conditions, behaviors, attitudes, and spending patterns.

To run a simulation: give it a title, select one or more audiences, write the prompt (the question you want to ask), and optionally add trial/study context (background about the protocol, medication, or intervention).

Example presets cover common scenarios: weekly injection trials, placebo-controlled studies, decentralized trials, frequent blood draw studies, and pediatric caregiver decisions.

9. Simulation Results

Select a simulation to view its results. Two tabs:

Individual Responses — Each persona's response is shown with their demographic badge (age, sex, income, education). Read individual perspectives to understand the range of reactions.

Aggregate Analysis — An AI-generated summary that identifies themes, patterns, and sentiment across all responses. Includes key themes with frequency and representative quotes, plus demographic-specific breakdowns.

10. Chat

Have one-on-one conversations with any persona. The chat interface has three panels:

• Left — Conversation list. Click any previous conversation to resume it.
• Center — The chat area. Type messages and the persona responds in-character.
• Right — Persona detail panel showing full demographics, conditions, traits, and memories.

Click + New Chat to pick a persona. Use filters to find the right person. You can optionally set trial context before starting — this primes the persona with background about a specific study or protocol.

Conversations are saved automatically. The persona builds memories from the conversation that persist across sessions, creating a more realistic ongoing relationship.

Data Sources

All persona traits are grounded in real survey data. Here are the primary sources:

• Census PUMS (ACS) — Demographics, population estimates, income, education, employment. Microdata records define the demographic segments.
• CDC NHIS — Health conditions, health status, access to care, smoking, BMI
• AHRQ MEPS — Healthcare utilization, spending, insurance, out-of-pocket costs, cost sensitivity
• Census Places — Geographic reference data for U.S. cities and towns, used for state and location assignment

Example Audiences

Click an example to pre-fill the form below

Audience Definition

Demographics

Conditions

Trait Filters

Behavioral traits that affect trial retention. Derived from AHRQ MEPS survey data.

Audience Size

Identity & Demographics

Conditions

Health Status

Healthcare Behaviors

Attitudes

Spending

Lifestyle & Social