Metrics question

What metrics would you use to track the improvement of sense of community within Facebook Groups?

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

Tests defining a metric for a qualitative concept (community feel) that resists simple engagement counts, requiring a genuinely thoughtful proxy.

Clarify what 'sense of community' actually means: mutual support, repeated meaningful interaction among the same members, and feeling recognized, not just group size or post count.
Reject raw post/comment volume as the metric, since a group can have high volume from a few dominant posters without broad community feeling.
Propose a network-based metric: reciprocal interaction rate, the share of members who both comment on others' posts and receive comments on their own within a group, over a set period.
Add a qualitative layer: a periodic in-group survey asking members if they feel a sense of belonging, since some community feeling isn't fully captured by behavioral data alone.
Add a distribution metric: interaction concentration (Gini-coefficient-style), since a healthy community has broadly distributed participation, not just a few power users driving all the activity.

Goes beyond simple activity volume to propose a reciprocity-based metric, which genuinely reflects mutual, community-like interaction versus one-way broadcasting.
Explicitly pairs a behavioral proxy with a survey-based qualitative check, recognizing 'sense of community' is partly subjective and not fully capturable by clicks alone.
Uses an illustrative number, e.g. assume only 15% of members currently both give and receive interaction in a typical group, showing a real gap to close.
Adds a participation-distribution metric, since a community dominated by a few loud voices doesn't feel like community to quieter members.

Defaulting to raw post or comment counts, which don't distinguish real community feeling from one-sided broadcasting.
Relying purely on behavioral data without acknowledging that a subjective concept like 'sense of community' benefits from direct user feedback too.

How would you validate that your behavioral proxy actually correlates with the survey-based trust score?
How would this metric differ for a small, tight-knit group versus a massive public group?