Product design question

Design a Facebook feature that measures user happiness.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Metrics design skill applied to a fuzzy human construct, checking whether you can operationalize 'happiness' into measurable, ethical proxies.

Clarify what happiness means here: momentary positive sentiment while using the app, or broader wellbeing from using it, since Meta has been criticized for ignoring the latter.
Reject self-report-only measures as the sole signal, since surveys have low response rates and recall bias, but keep them as a periodic validation check.
Propose behavioral proxies: ratio of positive reactions to negative ones on content a user posts or comments on, time spent on content associated with self-reported high mood in past surveys, and reduced doom-scrolling patterns like long passive sessions with low interaction.
Combine proxies into a wellbeing index rather than a single metric, since any one signal alone can be gamed or misleading.
Address the ethical risk directly: this data must not be used to manipulate engagement, only to inform product changes that genuinely improve wellbeing.
Define success as improvement in the wellbeing index correlated with periodic survey validation, not raw engagement.

Builds a composite index instead of one fragile proxy metric, showing awareness that happiness cannot be reduced to a single number.
Explicitly separates measurement from misuse, naming the ethical risk of using happiness data to increase addictive engagement instead.
Anchors behavioral proxies to periodic survey validation, catching cases where behavior and self-reported mood diverge.

Proposing time spent as the sole happiness metric, which actually measures engagement, not wellbeing, and can be directly contradictory.
Ignoring the ethical risk that this feature could be misused to optimize for addictive rather than healthy engagement.

How would you prevent this metric from being used to justify more addictive design?
How would you validate that your proxies actually correlate with real happiness?