Keeping Roblox Players Safe in Direct Messages: Contact Rules, Detection, Reporting

You are a product manager at Roblox, where players can send each other private messages. How would you build safety features into that messaging, and how would you know they are working? ```hint Start from the harms, then place the defenses List the concrete ways one player can hurt another through a private channel and which players are most exposed. Then ask where each defense belongs: before two players can talk, during a conversation, or after something goes wrong. ``` ### Clarifying Questions - Which channels are in scope: one-to-one messages between friends, group chats, private whispers inside an experience, or voice as well? - Should protections differ by age band (under 13, teens, adults), and how reliable is the age Roblox has for each player today? - Which harms worry the team most: grooming and sexual content aimed at children, harassment between peers, or scams such as fake currency giveaways and account phishing? - How much friction is acceptable: may we restrict who can message whom even if social activity drops? ### What a Strong Answer Covers - A harm taxonomy for private messaging, ranked by severity, with the most exposed players identified - Layered protections across the conversation: who can contact whom, real-time content defenses, in-chat controls and reporting, and enforcement - Age-aware defaults, how age is established, and the role of parents - Trade-offs among safety, privacy, false positives and social engagement - Metrics for harm prevalence and response speed, with engagement and accuracy guardrails, and a rollout plan ### Follow-up Questions - How would you detect grooming that builds slowly over weeks of messages that each look harmless on their own? - Parents want to read their teen's messages, and teens want privacy. Where do you draw the line? - Your strictest default for under-13 messaging would sharply cut how much they message friends. Would you ship it, and what would you watch afterward?