01
Baseline and random assignment
Captures prior belief or baseline preference before assigning respondent condition.
LikertScaleChoiceButtonsrandomization
Cases
Behavioral insights instrument
AI-advice treatment with reveal tracking
BE / nudge consultancy
A controlled intervention pattern for testing whether advice, model confidence, or information acquisition changes respondent choices.
Backed by skills/experiment/source-projects/ai-advice-treatment and the shared experiment runtime. Live model calls remain opt-in and must be logged when used.
https://ai-advice-treatment.exp.maestro.onl/
Complexity snapshot
- apps
- baseline, treatment_exposure, decision_task, debrief
- pages
- 15 pages across consent, baseline profile, prior beliefs, advice/control exposure, reveal, comprehension, decision, risk lottery, explanation, and debrief
- exportFields
- treatment, reveal event, recommendation metadata, baseline/final choice, confidence, timing, mechanism checks
- logic
- control/human/AI arms, click-to-reveal exposure, timer events, payoff display, lottery choice, comprehension checks
Instrument flow
Multi-step respondent logic, not a toy form
02
Advice exposure
Shows precomputed advice or a click-to-reveal recommendation with confidence metadata.
RevealPanelAIRecommendationRevealExperimentTimer
03
Decision and confidence
Captures final choice, confidence, and optional explanation after exposure.
PayoffMatrixChoiceButtonsSliderTextInput
04
Debrief and export
Exports treatment assignment, reveal state, timing, final decision, and explanation.
Instructioncompletion codeCSV export
Export contract
Fields the analyst actually needs
treatment.arm
advice_revealed
advice_choice
advice_confidence
final_choice
confidence_slider
decision_time_ms
Scope boundary
What this case does not claim
Not a multi-player synchronous lab game.
Live LLM advice requires explicit runtime/cost logging.
Build a live link from your own brief.
Upload the project description, old questionnaire, or PDF and iterate in chat.