Kid-Friendly by Design: Reporting, Feedback, and a Watchful Eye

An online game that kids play needs more than a "report" button in the fine print. How we implement protective mechanisms as a consistent design principle — as of today, with an honest look at what's still missing.

Reporting where it happens. In the game there's the /report chat command: it automatically quotes the last chat lines along with it — whoever reports doesn't have to retype or prove anything. Every report carries the world ID so we know the context. When entering a world, a short hint points out the function, and /help mentions it too.

Feedback is not a complaint. In the portal we deliberately separated "Feedback & Ideas" from "Report a player" — with its own card and friendly language. Someone who wants to share an idea shouldn't fill out a form that feels like pressing charges. And for parents there's a dedicated notice section both on the landing page and with the rules.

What's in progress: moderation in layers. For names and chat, a layered plan is ready: first normalization of obfuscations (turning "B4dW0rd" into "badword"), then fuzzy rule checking, then — only for borderline cases — an LLM check, and at the end always a human admin queue. Chat is to be masked locally right away and spot-checked asynchronously. What matters is the order: the cheap, deterministic layers catch the bulk, the expensive AI only sees the doubtful cases, and no automatic verdict is final.

The principle behind it all: protective functions must be as simple as the thing they're supposed to protect against. A kid who experiences something weird types /report — you can't demand more than that.