Try a Mini Experiment
What you’ll see
An ambiguous box contains 100 balls. Each ball can be one of four colors, but you do not know the actual color composition of the box. You will see five draws with replacement from this box.
What you’ll do
You will make ten binary choices. In each question, one or more colors are designated as winning colors, and you receive a hypothetical $5 prize if the ball drawn from a box is one of those colors. You choose which box you want the ball drawn from: a known box, where you know exactly how many balls are winning colors, or the ambiguous box, where you only observe the five previous draws.
How each question is chosen
The questions are selected adaptively using Dynamically Optimized Sequential Experimentation (DOSE). Based on your previous answers, DOSE selects the next question that is expected to be most informative about your ambiguity attitudes.
What happens after ten questions
Different ambiguity models will be fitted to your answers and used to predict choices you were never asked about. You will answer only questions involving two or three winning colors, while the models will predict your choices for single-color events.
The $5 prize is hypothetical: nothing is paid, nothing is stored, and nothing you do here leaves your browser.
Example Draws
The ambiguous box contains 100 balls, each colored either GREEN, BLUE, YELLOW, or ORANGE. The composition of the box is unknown to you.
Below, you will see five example draws from the box, one at a time. Each ball is returned to the box before the next is drawn. The summary of your example draws will appear on every following question for your reference.
What the models say about you
Choose one of the bets below—none of which you were asked about—to see what each model predicts you would choose. See which bets could be asked, and which were kept out.
See the fitted curve
The curve maps the subjective probabilityYour belief about how likely the event is. of each event to its corresponding matching probabilityThe matching probability is the fraction of winning-color balls in the known box that would make you indifferent between betting that a winning color is drawn from the known box and from the ambiguous box.. The dashed line represents ambiguity neutrality, and deviations from it reflect the model's fitted ambiguity attitude. Click a model name to highlight its curve.
Want to know more?
Everything above is a within-condition prediction: the models were fitted to choices from one box and used to predict other choices involving that same box.
Cross-condition prediction is different. A model that performs best within a condition is not guaranteed to perform best across conditions, because cross-condition prediction also depends on how stable the model's estimated ambiguity-attitude parameters remain when the situation changes.
This is exactly what my job market paper studies! Read more here.