Why we don't ban people

The short version: a ban removes a person and teaches nobody anything — not the person, not the room. We use permanent mutes, graduated responses, and a real appeal path instead. It is more work, it is occasionally wrong in the direction of leniency, and we think that trade is correct for a community this size.

What a ban actually accomplishes

Two things, and only two. It stops the behaviour immediately, and it removes the person from the space. Both are real. Neither is the same as making the room safer, and servers routinely confuse the three.

The problem with ban-first moderation is what it does to the people who stay. A ban is invisible. Nobody sees the evidence, nobody sees the reasoning, and nobody learns where the line is. What they learn instead is that the line is wherever a moderator decides it is on a given day — which is the opposite of a rule.

What replaces it

Four mechanisms, in the order they take effect.

A permanent mute, not a removal

A muted member cannot post. They can still read, still see what the community is, and still be reached by staff. That distinction matters: removal makes someone a stranger, a mute keeps them in the room where a conversation is possible.

Graduated responses

A warning, then a mute, then a longer one, with each step recorded so the next moderator can see the history rather than guessing. Severity tracks behaviour over time, not the mood of whoever is on duty.

Heat that decays

Our text moderation uses a tiered heat system that decays over time. Someone who had a bad week in March should not still be paying for it in September. Decay is what makes rehabilitation a real thing rather than a slogan — without it, every system becomes permanent by accretion.

Intent checked before consequence

An honest mistake and a deliberate provocation should not produce the same outcome. This is the expensive part, because it requires a human reading context. It is also the part that a bot cannot do, which is why moderation here is human-led rather than automated verdicts.

What this costs us

It is slower. Every case that would have been one click becomes a conversation, a record, and sometimes an appeal. It is emotionally expensive for staff, who have to hold judgement rather than apply a rule.

And it is sometimes wrong. Someone who should have been removed gets one more chance and uses it badly. That is a real harm and we should name it rather than pretend the system is free.

The counter-argument: ban-first systems also get it wrong — in the direction of removing people who did not deserve it — and nobody ever reviews those, because a ban that generates no complaint leaves no trace.

When removal is still the answer

This is not pacifism. Some behaviour is not a mistake and is not rehabilitated: anything sexual involving minors, hate speech, doxxing, or a pattern that shows the person is not interested in being part of a community at all. Those are not cases for a graduated response.

The distinction we use is between someone who did a bad thing and someone who is here to do bad things. The first gets a process. The second gets removed, and the eight rule departments spell out where that line sits.

Why this is written down

Because a policy that only exists in moderators' heads is not a policy — it is discretion, and discretion tends to favour whoever is loudest or longest-tenured. Writing it down means a member can predict the outcome of their own case, and can point at the page when the outcome does not match.

The full version, department by department, is on Safety & Guidelines. If you want the general argument rather than ours, what good moderation looks like covers it without the specifics. And if you have been on the receiving end of a decision, the appeal process is the practical page.