Numbers & Estimates
Does the First Number Pull Your Estimate?
Try two versions of the same pricing question and see why a starting number can pull later estimates toward it.
Experiments Lab
See the manipulation before you memorise the label.
Each demo changes one part of a decision and lets you compare the alternative version after you answer. Your response stays in the browser. We do not treat these demos as scientific measurements or collect them as study data.
How it works
One question. One changed variable. One evidence trail.
A random condition appears first. Make a quick judgment, then reveal the other version and inspect what changed. The useful lesson is the design of the comparison, not whether one click proves that you are biased.
Framing
Does Survival vs Mortality Wording Change the Feeling?
The numbers stay identical while the wording changes from survival to mortality. Notice whether your evaluation moves with the frame.
Choice Architecture
Can a Third Option Change Which Plan You Prefer?
Choose between two plans, then compare the same choice after a clearly weaker third option appears.
Projects & Money
Would You Continue If the Past Investment Disappeared?
The future economics stay the same. Only one version reminds you how much has already been spent.
Learning & Review
Does the Outcome Change How Good the Decision Looks?
Rate the same decision process after either a lucky or unlucky result.
Forecasting
What Happens When You See the Reference Class?
Estimate a project from the inside view, then compare it with a version that includes how long similar work actually took.
Research foundation
These demos are interfaces, not new findings.
Every experiment points to an evidence-reviewed concept. The machine-readable experiment specifications can also be reused later for controlled human studies or AI-model benchmarks.
Observatory
Move from controlled examples to public information.
The Cognitive Bias Observatory measures transparent headline-level framing and attention signals while preserving source provenance and sampling limits.