Why are cheaters not necessarily banned?

A go educator recently posted a video in which they ran into a very obvious AI cheater on OGS and got smoked. Unfortunate, but it happens.

Out of curiosity I had a look at the opponent’s account and noticed that almost all of their games have been annulled. Many of the annulments were originally wins by resignation but now have the message “This game has been annulled by a moderator”. It looked pretty clear that the player was using AI in those games as well, and I couldn’t see any other reason for the annulment.

So this made me wonder why the user has not been entirely banned from the site - I assumed that repeated moderator interventions based on AI use would be enough. Are the specifics of the policies spelled out anywhere?

(I’m not going to mention any details about the incident that led me to this question - I flagged the game in question for AI use just in case nobody else has. I’m more interested in the policy, not that specific player / game.)

One possible argument is that if you ban them, unless that banning is very clever, they will just come back again and do the same behaviour and you have to identify them all over again, whereas if you let them stay at least you know who they are to keep annulling their games. That’s the idea behind shadow chat bans.

2 Likes

In any case ban happens too at times

2 Likes

In the particular case that I believe that you are referring to, it seems that it was simply a clerical oversight, and the moderator handling the case did intend to ban that user, but simply forgot to do so.

3 Likes

Oh, well, that’s a reasonable explanation if that’s what happened. I’m still sort of curious - how many incidents of this sort does it take before a user is banned?

The matter is not as easy as you think for multiple reasons .

There is a team of benevolent moderating by receiving the reports. The first difficulty is to understand a report to be able to check what’s going on. Then we have a procedure with mutual agreement about what we should do on each report. Not all is easy but we fixed guidelines and we still debate unclear cases. Most is kind of automated and you can’t ask everything to a benevolent taking on his free time, like checking the history of a player or studying a game in depth.

So yes we have moderation and decisions are taken. But in the actual state of the art we can’t answer more precisely your question.

2 Likes

Oh for sure, I have no doubt that it is incredibly difficult to make a ruling on whether or not someone is cheating, particularly given limited time and evidence. My question is about what happens once it is determined that someone cheated - does OGS have a zero tolerance policy? Or do they get a warning and then a ban? Or something else?

In principle they get a warning, but a new account that has cheated in all their games may get a ban directly.

2 Likes

I can elaborate on this.

First I will say that this is a matter of judgement, case by case.

However, you can expect that a person who has AI detected will be immediately banned if:

  • They’ve been detected before and warned OR if
  • It appears that they have been using AI the whole time - they came here and started cheating.

The opposite is true if a person recently started using AI after playing honestly. In that case they get a warning.

In all cases, once we are sufficiently sure that AI has been used in a game, then any games following that, and likely suspects prior, are annulled.

There’s a current discussion amongst the people working hard on this as to whether banning itself is effective enough.

Someone who is really determined can get around various levels of defence we have against “just create another account and return”.

This kind of topic is under continual review - obviously behind closed doors (surely it is obvious why).

We are continually actively improving detection and control measures.

The single most challenging thing we face is making convictions that we can defend when the person comes to the forum or goes to redit saying “hey they banned me for AI, clearly this isn’t AI”.

All the armchair experts leap in with obvious observations, failing to appreciate that we have far more data than is on the table.

When this happens (someone banned who disagrees) we’re the bad guys.

When we haven’t banned someone who ostensibly “obviously is using” we’re the bad guys.

Please be reassured that we are working hard on it.

Please know that if your AI report is knocked back, it is still on the person’s record, part of the accumulating evidence. It’s not knocked back because “we’re too dumb to know” …

8 Likes

Thanks for the full explanation - my curiosity is satisfied!

Seems like a reasonable policy, for what it’s worth. No policy is going to be uncontroversial or easy to implement, but some policy should exist, IMO. The fact that OGS tries to deal with the problem at all puts it ahead of other sites, e.g. Fox.

2 Likes

I wish we could share more awareness of our efforts.

It’s a darn thankless task :squinting_face_with_tongue:

But those involved hate AI cheating as much as anyone, it makes it worth it…

1 Like

This is much better than IGS pandanet

3 Likes