Banned by something that can't listen to you

The screen goes black for a second and a red message appears with a word you'd never seen before. It's disconnected you. You try again and you can't get in any more.

You look up how to appeal. You find a form that asks for your name, the date and the reason for the sanction. The reason is exactly what you don't know. You fill it in anyway, write three lines explaining that your connection was bad, and hit send.

Nobody replies. Not that week, not ever.

What stays with you isn't the ban

Weeks later you'll still think about it now and then, and if you try to explain it to someone, it'll come out sounding disproportionate. It was a game. There were other servers. You didn't lose anything important.

All true, and even so the unease doesn't go away. Because what went wrong wasn't losing access: it was that the thing that threw you out had no way of hearing you, and that nobody seemed to think it needed one.

Why people comply with things they don't like

The psychologist Tom Tyler devoted decades to a question that seemed to have an obvious answer: why people obey authority. The comfortable answer was fear of punishment. His data said something else.

What he found, published in 1990 and confirmed many times since, is that people comply with a decision above all when the procedure seemed fair to them, fairly independently of whether the outcome benefited them. Someone who loses a court case and feels they were heard accepts the verdict better than someone who wins while feeling mistreated.

He isolated four ingredients. Being able to put forward your side. The decision-maker being seen as neutral and not as someone with a grudge. Being treated with respect along the way. And believing there is good intent behind it.

None of the four says anything about being right.

A detector isn't a judge

With that in hand, look at what an anti-cheat system does.

It observes a pattern, compares it with what it considers normal and acts. It's often right: some people really do cheat, and going after them protects everyone else. But in doing so it fails all four ingredients at a stroke. It doesn't explain what it saw. It doesn't tell a cheat from a connection that's stuttering. It doesn't treat anyone in any particular way, because it doesn't treat. And it can't receive a different version of events, because it has nowhere to put it.

The important thing is that this still holds when the system is right. Being right without a process leaves the innocent defenceless and the guilty convinced they were caught through bad luck. Neither of them learns anything.

And something that's almost never counted: every unexplained ban is also seen by those who stay. They don't learn that there's order. They learn that the place is capable of doing that without being accountable, and from that day on they play a little more defensively.

The principle: almost nobody demands to win, almost everyone demands to be heard

Step out of the server, because this is the part that repeats everywhere.

The patient whose appointment is changed without warning. The passenger whose flight is cancelled with an automated notice. The worker whose shift is changed by a system that doesn't accept replies. In every case the complaint that comes out first isn't "I want the decision to be different": it's "I want someone to tell me why and let me answer".

It's a modest request. And it's the first one suppressed as soon as an organisation grows, because listening doesn't scale and deciding does.

Hence a criterion that works for almost any institution: measure its decency by what it offers the person who has just lost. With the winner, no effort is needed.

What can be done

A concrete, written reason, not a label. A route for replying that reaches a person within a deadline, even if the outcome ends up being that the sanction stands. And visible proportions, so it's clear why this is three days and that is permanent.

If there were a place where no sanction was applied without a written reason and where every appeal reached someone who reads it, I'd like it to be called KobiiCraft.

Now the uncomfortable part. Moderating like this costs time, and time is the only thing that's truly scarce in a community. At a certain scale it's impossible to attend to everyone. Sometimes you have to remove someone quickly to protect others and explain afterwards. And false positives will always exist, because any detector that is never wrong is a detector that detects nothing.

Anyone promising perfect moderation is promising something they don't control. What can be controlled is whether the person who takes the blow has somewhere to reply.

Three lines sent to no one

The worst thing about that form wasn't that it said no. It's that it said nothing, and that forces the person writing to decide on their own what happened.

A place that punishes and doesn't listen hasn't imposed order. It has saved itself a conversation, and it has charged for it in a person who never came back.

Frequently asked questions

Why does a ban you consider unfair hurt so much? Because what breaks is almost never the access, but the treatment. Research on procedural justice shows that people accept decisions that harm them when the procedure seemed fair to them, and reject even correct decisions when they couldn't explain themselves. Losing an account is annoying for a while; a system accusing you without saying of what and without hearing a reply touches something older, which is the feeling of not counting as someone who is owed an explanation.

What is procedural justice? It's the idea, developed by the psychologist Tom Tyler from 1990 onwards, that an authority's legitimacy depends more on how it decides than on what it decides. Four elements sustain it: that the person affected can put forward their side, that whoever decides is seen as neutral rather than arbitrary, that the person is treated with respect, and that there is believed to be good intent behind it. When those four are present, people comply even with decisions that harm them.

Can an automated system moderate fairly? It can detect, but it can't listen, and that limits how far it goes. An automated detector can be right in most cases and still produce an unfair experience, because it doesn't explain what it saw, doesn't tell a bad connection from a cheat and has no way of receiving a different account. The sensible thing is to use it to flag and to leave serious sanctions, or at least the review, in the hands of someone who can read a reply.

What should a decent sanctions process include? Three minimal, cheap things. A concrete, written reason, not a generic label: what was detected, when and where. A real route for replying that reaches a person and has a deadline, even if the answer ends up being that the sanction stands. And visible proportionality, so that it's clear why this is three days and that is permanent. None of that requires technology: it requires time, which is what's truly scarce.