MVCA X
Monet · by Schema
MVCA X Monet
01 Source

What do you want to run?

Describe your own study and let AI build it — or upload data from a finished run to re-open its analysis.

✦ Describe your own
↺ Load saved study
Upload data
1 Describe your study
In plain English, give the conditions, the stimulus each group sees, and the dependent measure — or paste a whole document, or import a file (a Qualtrics PDF, a Word doc, a screenshot…). AI reads it into a runnable study you can edit.
Supports up to 8 conditions across up to 3 crossed factors — single-factor 1×3 … 1×8, or factorials like 2×2, 2×3, 2×4, 2×2×2. For a factorial, the AI detects the factors and you pick which one is the IV vs a moderator in the Theory step.
Not sure how to phrase it? Start from an example — click to load, then edit.
Pricing psychology Scarcity framing

We test whether a scarce framing ("only 50 left") raises purchase intent versus an abundant framing ("always in stock"). Both groups see the same wristwatch ($380). Dependent measure: purchase intent on a 1–7 scale. We expect scarce > abundant, and the effect to be stronger for status-motivated shoppers.

Ad messaging Emotional vs rational

We compare an emotional ad ("Belong to something bigger") against a rational ad ("12-hour battery, 2-year warranty") for the same pair of headphones. Dependent measure: ad persuasiveness, 1–7. We expect the emotional ad to win for hedonic shoppers but the rational ad to win for utilitarian shoppers.

Product design Rounded vs angular

We test whether a rounded bottle silhouette feels warmer and more likeable than an angular one. Both show the same soda brand and price. Dependent measures: warmth (1–7) and willingness to buy (1–7). We expect rounded > angular on warmth, mediated by perceived friendliness.

2 Key psychological variables optional
Name the constructs you expect to matter; AI maps them to the model's substrates. Leave blank for none.
Who answers? By default, a 100-person US-representative panel — ready to run, no setup. Build a custom persona panel (census presets or your own filters) to scope the audience.
no selection
02 How it runs

Every persona thinks it through.

MVCA runs in naturalistic mode. Each persona's answer is produced by the full 12-process cognitive cascade — and, in parallel, a bare LLM answers the same prompt directly. The gap between them is your finding.

Watch a stimulus move through the cascade — hover or tap any stage to inspect it.
①
The cascadeA stimulus flows through 12 cognitive processes — attention → appraisal → synthesis → judgment — each reading the persona's literature-grounded substrates.
②
The baselineThe same persona & stimulus also go straight to the LLM, with no architecture — the "does the model already know?" control.
③
The contrastA real effect = the cascade moves the measure the way the literature predicts and the bare LLM does not.
03 Cognitive style

Bias the cascade.

Each slider shapes how this person thinks — fast or careful, cool or feeling, inward or outward. The portrait on the right shows the mind shift in real time.

Quick presets
Deliberation
system-1 ↔ system-2
0.0
Affect responsiveness
cool ↔ hot
0.0
Social attunement
private ↔ public
0.0
Counter-signal sensitivity
literal ↔ ironic
0.0
Norm conformity
autonomous ↔ normative
0.0

How this mind will process

A balanced, typical way of taking the world in.
04 Model

Pick your engine.

Different providers, different cost / latency / behaviour profiles. The cascade is provider-agnostic.

05 Sample

How many personas?

Each persona × condition is one cell of the cascade. The estimate updates live.

personas per condition
Cells
—
LLM calls
—
Wall time
—
parallel = 100 · model from settings
06 Confirm

Ready to run.

Review and launch. The cascade streams progress live; you can stop any time.

— Running

Cascade in progress.

the cognitive cascade is running across your sample

Starting…
—
0% — / — cells
— Results

Run complete.

—

DV by condition
DV distribution
cascade vs llm_control vs published benchmark
Process activation
12 cascade processes in execution order, each row showing the top-5 substrates by mean engagement weight across all calls to that process during this run
loading process breakdown…
Top discovered moderators
the 3 substrates whose persona-level value most strongly tracks the DV (Big Five traits excluded — those get their own dedicated section below)
Big Five segmentation
how each persona-level Big Five trait correlates with the DV in this run — fixed, canonical-order panel that runs regardless of which traits the study predicted
Download run .zip

Build a persona panel

Pick a one-click census preset, or define a custom panel by demographic & consumption filters. The pool becomes selectable as your run's persona source.

pools: —
Your pools
No panels built yet — pick a preset or define one below.
1 · Start from a preset — fills the filters below (optional)
loading presets…
2 · Filters — edit anything, then build. These always drive the panel.
starting…