BiasMarket Lab
When AI-generated product framing distorts belief, does it also change choice, willingness to pay, and market outcomes?
A research platform studying how AI-generated product summaries -- positive-valence framing, selective omission of weaknesses -- shift consumer beliefs away from ground truth, and what that implies for choice, stated willingness to pay, and (via simulation) market share and consumer surplus.
Behavioral results
Simulated pilotSynthetic-agent data only, used to validate the analysis pipeline. Never evidence about human behavior.
905 included trials from a rule-based synthetic-agent population, generated to validate the analysis pipeline before any human is recruited. See /results.
Market simulation
No human dataWhat-if market simulation with user-set or assumed parameters. Never a measurement of any real market, platform, or seller.
Every number under /market is labeled “INTERACTIVE SIMULATION” and is adjustable by the reader -- it is a what-if tool, not a measurement of any real platform.
No human participants have been recruited.
No IRB or ethics submission exists for this project. Every number on this site carries a status badge read directly from the artifact that produced it -- never hand-typed, never implied to be about real people until that badge says so.
Explore the site
Results
Primary outcomes, forest plots, H6 moderation, Bonferroni-corrected significance.
Beliefs
Belief-error distribution around the omitted battery-life attribute.
Choice
Choice-share treatment effects, the H1 test.
Market
Interactive multi-seller market simulation -- never an empirical finding.
Interventions
The planned fact-table-disclosure study (H5) -- not yet run.
Stimuli
Every AI-generated stimulus, its provenance, and the manipulation-check audit.
Methods
The 3-arm design, preregistration, and confirmatory analysis plan.
Data
What's real, what's placeholder, and current storage status.
Paper
Manuscript abstract, contributions, and the (empty) results section.