On handling online cheating with AI

[Adin]

There are many openly AI enhanced players on KGS now, I haven't noticed anyone removing their rank. So I am not sure that you are right about that.

Have you reported them? I removed myself the rank of someone who was an openly AI enhanced player. If you wish you can send me a PM with their usernames.

[Bill Spight]

I look at it as saying that some robbers are taking away our goods and we should do something about it. And then someone else talks for hours about whether a robber usually wears a hoodie or just a simple shirt, and do they use Android or IPhones etc etc. These things are totally irrelevant, a robber is that guy with YOUR phone in his hand.

How can you tell if a robbery has taken place? What action should you take about a nonexistent crime? If you are going to take any action, you have to learn how to be a detective.

[Knotwilg]

I just checked the KGS TOS at http://gokgs.com/tos.jsp. Although it says "Dishonest play during rated games." is not allowed, it says nothing specifically against using bots, just prohibits "Having other people play rated games for you".

Using an AI as a human player in rated games does qualify as dishonest play. But it is certainly a good suggestion to add it specifically, I will bring that up.

If you're not careful, you're going to have more posts about Bill's posts than about what you actually want to discuss. Put away your anger for a moment will you? I wouldn't be confident in the refereeing of a person so easily aroused.

If you care about something then you are passionate about it. I look at it as saying that some robbers are taking away our goods and we should do something about it. And then someone else talks for hours about whether a robber usually wears a hoodie or just a simple shirt, and do they use Android or IPhones etc etc. These things are totally irrelevant, a robber is that guy with YOUR phone in his hand.

An analogy is a sign of someone becoming intellectually insecure about his own argument yet emotionally attached to it. You'd better stick to the point.

You want to do something against people using bots to boost their ranks online. Several arguments are being offered here, one of which is Bill's tangent, trying to find out if you can at all figure out if someone uses a bot. Whether that exploration is very fruitful to your quest, I don't know, but it's certainly a more valuable contribution than your own rambling against it. You can choose to follow other arguments instead. I think that's perfectly fine.

[jlt]

And of Blackstone's principle that

It is better that ten guilty persons escape than that one innocent suffer.

What do you think the correct ratio is for AI cheating is in Go. I think it is reasonable to differ from that for say the crime of murder and the punishment of lenghty imprisonment or execution. But we shouldn't neglect the reputational damage of declaring someone an AI cheater, indeed I would be happy with a lower standard of evidence against an anonymous online account than one linked to a known human individual.

Robert seems to think that ratio is infinity to 1. The US justice prison system seems to think it's flipped at 1 to 10. My first feeling is maybe a bit lower than Blackstone's, say 5 to 1, is an appropriate target to design for. What do others think is a suitable one for Go cheating?

It depends what kind of punishment you are talking about. A ratio of 5:1 is perfectly fine if you just invalidate the rank of an anonymous account (i.e. the account is no longer allowed to play ranked games), maybe 10:1 if you ban the account and 500:1 if in addition you ban the IP.

On the other hand, if the punishment is to ban an account of an identifiable person and to state publicly that the reason is cheating with AI, then I wouldn't be comfortable with a ratio less than 100000:1.

There are other ways to prevent cheating without ruining a player's reputation. For online tournaments, you can ask a player to be monitored and/or filmed by a camera that shows the player's screen as well as all his body movements.

[ez4u]

I find some of the comments here rather beyond the line of being helpful.

The issue is how should admins police the servers so that we do not see the communities that exist there deteriorate as people become unwilling or too cautious about playing. See Baduk Doctor's 9-dan videos, where he says that these days he only plays known opponents because the 9-dan ranks have become overrun with people playing via AI's.

Personally I think that we are far from detecting human vs bot in any automated fashion. I see quite a lot of variation in move choices among bots, among different versions of the same bots, and depending on how many playouts the bots are given. So what can be done?

First, everyone in the U.S. (i.e. where KGS is located) should stop with the absolute proof discussions. I cannot speak for other countries but in the U.S. the standard is different.

Source: criminal.findlaw.com/criminal-law-basics/the-differences-between-a-criminal-case-and-a-civil-case

"The Standard of Proof

Crimes must generally be proved "beyond a reasonable doubt", whereas civil cases are proved by lower standards of proof such as "the preponderance of the evidence" (which essentially means that it was more likely than not that something occurred in a certain way). The difference in standards exists because civil liability is considered less blameworthy and because the punishments are less severe."

What should the less severe punishments be? The best that I can think of is that if the admins examine a complaint against a player and feel that it is more likely than not that the player used an AI:
1. They should remove the player from any tournament that they are in and nullify their result if there is a tournament involved.
2. KGS should develop an new icon for "suspected bot" or something and apply that icon to the player. That way people should be able to see a ready indication of the judgement of the server with regard to that player. If you don't mind, you can play them. However, people that do mind should have easy access to the judgement of the server.
3. People suspected of using bots should be banned... from automatch rather than from the server. This would lower the likelihood that people who object to using AI's would end up playing against people that do so.

Of course I have not found a better way to confirm whether an AI has been used. My advice is to do what you do throughout the day every day - use your common sense. In other words, if it walks like a duck and quacks like a duck, slap the duck icon on it. However, do not lose any sleep over it. Believe that you are acting for the greater good. In 1970 George Akerlof published "The Market for Lemons" on the economic costs of dishonesty and the negative effects on markets. The paper won him the 2001 Nobel prize. Go servers represent such markets where people gather to exchange entertainment services. The ability of some players to cheat, for whatever reason, does reduce the average experience of the more honest players and makes the servers less attractive. I support the efforts of the admins to police the servers and am particularly appreciative of the unpaid volunteers.

[Adin]

As I mentioned before I think an error rate of 1 in 100 is achievable. And anything worse than 1 in 10 is an obvious failure of the investigator.

I'm not comfortable with "punishments" since there will always be some error. The point is not to punish someone but to protect the community. For example on KGS you just get your rank deactivated and you can ask it back after 6 months when all your ranked games have gone out of the system. You can create a new account and start playing right away (of course, admins may monitor it and remove the rank from it too if shows the same AI behaviour). But there's no banning or public shaming or any such stuff.

In 1970 George Akerlof published "The Market for Lemons" on the economic costs of dishonesty and the negative effects on markets. The paper won him the 2001 Nobel prize. Go servers represent such markets where people gather to exchange entertainment services. The ability of some players to cheat, for whatever reason, does reduce the average experience of the more honest players and makes the servers less attractive.

Well said, it basically takes away some of the enjoyment of Go from us.

[gennan]

If the false positive rate is more than 20% and the sanction is disproportional like an automatic 6 month ban from the server, I would definitely be against it.

If the false positive rate is less than 1% and if the sanctions are proportional like @ez4u proposes, I would definitely be in favour.

But how do we know the false positive rate? I think quite a bit of testing has to be done to measure this rate for whatever bot detection method the admins use (without disclosing the details of those methods).

[RobertJasiek]

You don't want to hear it because you don't want to invest the work: false positives are identified by observing each alleged cheater in person proving his strength in self-decided play for many games.

[Polama]

I'd posit there's two types of cheaters: people who just run the bot and destroy their opponent, who want to play the best move every move. And people who want to get away with cheating, who spot check for blunders or the occasional tricky move, but who are savvy enough to not blindly play clearly bot sequences.

I really don't think you'll catch the latter group with anything short of monitoring. And again, if you have an algorithm that detects bot moves, you can train a bot that doesn't play those moves.

In a lot of ways, though, I think it's the former group that's the bigger problem because it's more visible. If people are complaining that the 9-dan ranks are all bots, that's a perception of a widespread problem. But I think Spinal Tap can point us to a helpful counter-measure: make it go to 11! Bots are better than us at Go. Let the cheaters raise up to a rank above us humans and play amongst themselves.

[Bill Spight]

I'd posit there's two types of cheaters: people who just run the bot and destroy their opponent, who want to play the best move every move. And people who want to get away with cheating, who spot check for blunders or the occasional tricky move, but who are savvy enough to not blindly play clearly bot sequences.

I really don't think you'll catch the latter group with anything short of monitoring. And again, if you have an algorithm that detects bot moves, you can train a bot that doesn't play those moves.

If, however, the bot detectors are based on different preferences between bots and humans that do not have to do with the quality of play, then the extra task of finding human-like plays is likely to reduce the bot's level of play. IIUC, today's top bots are Zero(™) bots is that the initial training on human professional play is a handicap.

In a lot of ways, though, I think it's the former group that's the bigger problem because it's more visible. If people are complaining that the 9-dan ranks are all bots, that's a perception of a widespread problem. But I think Spinal Tap can point us to a helpful counter-measure: make it go to 11! Bots are better than us at Go. Let the cheaters raise up to a rank above us humans and play amongst themselves.

Interesting idea.

[Gomoto]

An analogy is a sign of someone becoming intellectually insecure about his own argument yet emotionally attached to it. You'd better stick to the point.

ad hominem

[Bill Spight]

An analogy is a sign of someone becoming intellectually insecure about his own argument yet emotionally attached to it. You'd better stick to the point.

ad hominem

Yes. But a fallacy?

[daal]

I'd posit there's two types of cheaters: people who just run the bot and destroy their opponent, who want to play the best move every move. And people who want to get away with cheating, who spot check for blunders or the occasional tricky move, but who are savvy enough to not blindly play clearly bot sequences.

I really don't think you'll catch the latter group with anything short of monitoring.

Isn't the whole idea of neural networks that AI can learn to recognize patterns that humans might not be aware of? Feed it a bunch of games where savvy players cheat and let it make the call. For a server like KGS, the combination of automated cheater detection and an automatic temporary ban would certainly serve as a deterrent without ruining anyone's reputation. It doesn't have to always be right, it just has to be the authority.

[Knotwilg]

An analogy is a sign of someone becoming intellectually insecure about his own argument yet emotionally attached to it. You'd better stick to the point.

ad hominem

Yes. But a fallacy?

I apologize for throwing oil on the fire, as we say. Let the discussion continue on topic.

[jann]

who spot check for blunders or the occasional tricky move, but who are savvy enough to not blindly play clearly bot sequences.

I really don't think you'll catch the latter group with anything short of monitoring.

Isn't the whole idea of neural networks that AI can learn to recognize patterns that humans might not be aware of? Feed it a bunch of games where savvy players cheat and let it make the call.

NNs can only find things that are actually there, for example, what could also be found by reasonable human effort. How well would humans do here, even after a long study? With blunder-check method a cheater leaves minimal traces. As mentioned, unusual lack of big blunders may be seen as evidence, but only if it persists in dozens or hundreds of games. (And it is also easily possible to develop a cheating logic that mimics stronger humans' style and errors for example.)

The kind of "automated detection" in some people's mind is prolly something that marks a player after a few suspicious games already (maybe even showing a "botness percentage" besides every player's rank or rating ). But this is simply not realistic for the more clever types of cheating, only for the simplest type.