When treating some users secretly changes the others
Asked at Uber
A ride-hailing app A/B tests a feature that makes treatment users request rides more aggressively. Treatment engagement is up, so the team wants to ship. But drivers are a shared, limited resource.
Explain how interference between units violates a core A/B assumption, why it biases the estimate here, and what designs address it.
Your answer
This one is open-ended. Work it through, then check your reasoning against the full solution.