Skip to main content

Day 2 · Block 2

E08 - Three AI data-collection modes by example

  • Natural DGP + same instrument: design A/B experimental stimuli with AI (e.g., two highly similar images that differ only on one treatment).
  • Natural DGP + new instrument: research how Prolific could be connected to an AI agent and propose one new data-collection flow.
  • AI DGP: build a tiny synthetic-respondent demo using a few observations inspired by Twin-2K-500 (Toubia et al.).
  • Interview at least one synthetic respondent and assess whether the answers seem plausible.
  • Ask the synthetic respondent survey questions that you are interested in.

Inputs

  • One treatment idea for A/B stimuli
  • Prolific + AI-agent integration notes
  • Tiny synthetic respondent profile inspired by Twin-2K-500

Deliverable

one worked example per mode + plausibility note + cross-mode comparison table.

Target

Submit one worked example for each mode: Natural DGP + same instrument, Natural DGP + new instrument, and AI DGP.

Checklist

  • Make one explicit design decision.
  • Include one verification check.
  • State one limitation or risk.

Common failure modes

  • Examples remain abstract without a concrete artifact.
  • No plausibility check for synthetic respondent interview.

Extension task (optional)

Add one external validation check for the synthetic respondent answers.