Design and critique teen-parent impact experiment

Meta is preparing a feature that would let parents register and connect to their teenager’s account. Leadership is concerned about possible harms to teens, such as declines in well-being or lower-quality engagement. Design a study to estimate the causal effect of parental registration on teen outcomes.

Answer the following precisely:

1. Define the success metrics and guardrails: specify primary and secondary outcomes, their expected direction, and the assumptions behind the minimum detectable effects. Give concrete examples such as time spent on the platform, harmful-content impressions, report or mute rates, session-length volatility, and churn.
2. Experimental design: choose the unit of randomization—teen, parent, household, or cluster based on the social graph—and justify that choice under possible interference or SUTVA violations, for example messages between linked family members or peer spillovers. Propose a practical assignment mechanism, such as an encouragement design or stepped-wedge rollout, and define the exposure.
3. Power and duration: outline the inputs for sample size, expected compliance or take-up, and how you would handle staggered adoption and late joiners.
4. Measurement and attribution: specify whether you would use ITT or TOT, how you would handle partial compliance, mislinked accounts, and attrition. Propose CUPED or covariate adjustment to improve precision.
5. Threats to validity and mitigations: identify at least five concrete risks—for example, selection bias among families that opt in, network interference, policy-induced behavior changes, seasonality or back-to-school effects, and measurement error in well-being proxies. For each risk, provide a mitigation, such as household-level randomization, graph clustering, pre-exposure matching, difference-in-differences with teen fixed effects, synthetic controls, instrumental variables, or exclusion windows.
6. If an RCT is infeasible, propose a credible quasi-experiment: specify the design—for example, regression discontinuity at age thresholds, an instrument based on exogenous invite timing, or difference-in-differences on the same teens before and after with matched controls—along with the identification assumptions, diagnostics, and robustness checks.
7. Ethics and safety: define eligibility filters, safety holdouts, monitoring, and stop conditions for adverse outcomes. Explain how you would communicate results and make a launch decision under uncertainty.

Overview: This question tests a data scientist’s ability to apply causal inference and experimental design to measure how parental registration and account linking affect teen outcomes, covering metric definition, power analysis, compliance, measurement, and ethical safeguards.