Why
See which of your inputs pushed the advice, and by how much.
Take part
Help us find out what makes AI advice trustworthy
You will read six short investor cases, each with a recommendation from the advisor and one style of explanation. Tell us what you would do with that advice, and the session is done.
- 10 to 15 minutes
- Anonymous, no account
- Six hypothetical cases
Random assignment is what makes the results comparable, so this is the option we recommend. It can also give you a control session with no explanation at all.
Or choose an explanation style yourself
Every style below is one way of answering the same question: why this recommendation, and how much should you rely on it. Your choice is recorded as a choice, so it stays separable from the randomly assigned sessions.
What is explained
Different material about the same recommendation, shown as a static panel.
What would change it
The smallest change to your situation that would flip the advice.
How sure
The advisor's calibrated confidence, with the full probability picture.
All three
Why, what would change it, and how sure, together.
How it is delivered
The same three contents, handed over in a different way. The last one drops the written explanation entirely.
Interactive with all three
All three explanations, plus controls to move the inputs and watch the advice react.
Adaptive
An explanation that adjusts to your financial literacy.
needs a modern GPU browser
Conversational
Chat with an explainer running entirely in your browser.
Interactive only
Move the inputs yourself, with nothing written to explain the advice.
The advisor itself, the neural network or the interpretable scorecard, is assigned at random either way. The session ends with a debrief that tells you which recommendations were deliberately flawed. How the study is designed.