Facebook prompts to prune your friend list: launch call and candidate selection

Facebook is considering a feature that would proactively suggest that a user unfriend certain people on their friend list. Should Facebook launch it? And if it goes ahead, how should Facebook decide which friends to suggest? ### Clarifying Questions - What problem is the feature meant to solve: a cluttered feed, users holding back from posting because their audience is too broad, unwanted contact, or friend lists that no longer reflect real relationships? - Must every suggestion lead to unfriending, or could it also offer lighter options such as unfollowing, snoozing or moving the person to a restricted audience? - Where would suggestions appear: in a review flow the user opens, or as unprompted prompts in the feed? And would the other person ever learn about it? ### Part 1 — Make the launch call Decide whether Facebook should launch unfriend suggestions as described, launch them in a changed form, or not launch them at all. Weigh what users and Facebook would gain against what could go wrong. ```hint Two people per suggestion Each suggestion involves the person who sees it and the friend who might be removed. Think through what each of them gains or loses, and what fewer connections would mean for the network over time. ``` #### What This Part Should Cover - The user problem the feature would address and how you would confirm that it is real - The effect on the friend who is removed, on sharing and on the value of the network - Emotional, privacy and reputational risks, and lighter alternatives to unfriending - A clear recommendation and the evidence that would change it ### Part 2 — Choose which friends to suggest Assuming the feature goes ahead, explain how Facebook should pick and rank the friends it suggests, who must never be suggested, and how you would tell whether the suggestions are good. ```hint Start with the user's own behavior What has the user already done, or stopped doing, with a particular friend? Those signals reveal more about the relationship than anything about the friend's own activity. ``` #### What This Part Should Cover - Candidate signals and how they are combined into a ranking - Exclusions for sensitive relationships and sensitive data - How suggestions are presented, how many and how often, and how a user undoes one - Offline and online evaluation, including how often users regret an action ### What a Strong Answer Covers - A goal centered on users' control over their audience and feed, not on engagement alone - A selection design consistent with the launch call, so the form recommended in Part 1 shapes how candidates are chosen and shown - A success metric with guardrails such as reversals, complaints, posting activity and time spent - A rollout and experiment plan that measures effects on both the user and the friends who are removed ### Follow-up Questions - In testing, users act on many suggestions but then post less often. What do you do? - How would you keep the feature from suggesting people in ways that expose sensitive events, such as a recent breakup or a friend's death? - Would you build the same feature for accounts a user follows on Instagram? What would change?