Except you can’t. It’s what I keep explaining, and people here either fail to grasp why or refuse to believe it. I’ll assume it’s the former and try to explain in terms that I hope are sufficiently clear.

Think about all the land claim and building abuses that are reportable. Not what you believe should be reportable – i.e. not your own pet peeves – but those that are reportable. Also, bear in mind that “reportable” is not the same thing as “infraction”. What I’m asking you to do is think about things that would be worth reporting, so that the admins would investigate if it’s an infraction.

The reason why I’m asking you to think about these things is because that’s the first step when trying to come up with code that you’re asking devs to write. So try to come up with some stuff.

I’m betting that many of you are now thinking about the size of claim and/or build. Think broader. What about walling off someone’s base? What about blocking access to something important? What about a building that spells out something that is forbidden by the rules? What about foundation spam? What about running a line of foundations to the border of someone’s claim – without walling them off – to prevent them from expanding? What about bunching up resource-intensive placeables in a way that negatively affects performance?

If you’ve thought of all of those on your own, congratulations! Things get harder from here. Now you need to think about how to detect these things. Don’t worry, you don’t have to come up with an algorithm, you merely have to spend some time thinking about it, in order to grasp the complexity. If your answer is “that’s not my job”, then all I can say is “that’s fair, but then you don’t get to be frustrated when your proposal is ignored or rejected out of hand”. Otherwise, try to spend some effort on thinking through the variety of things you’re trying to detect and the edge cases and ways to try to game the system.

At this point, you’ve hopefully realized this is not as magically easy as you might have thought. In fact, you might even doubt whether it’s doable. Yeah, good question. But let’s assume that if you pile enough code and tools on top of this problem, it’s doable.

Here comes the “fun” part.

First of all, your detection system will either have false positives or false negatives. A false positive means that the system flagged something for inspection and it’s not an infraction. A false negative means that someone committed an infraction and your system didn’t flag it for inspection.

False negative would defeat the goal you set, which is to eliminate the need for human reporting. So you would have to try to design a system that is “suspicious” enough to have no false negatives, but that system would generate a lot more false positives.

“That’s all right, false positives are okay, because things get flagged and then admins investigate them. Problem solved, right?” Wrong. If you think things are bad now because Zendesk gets clogged with spurious reports from players that admins have to waste time investigating, think about how bad it would be if there was an automated system flagging scores, if not hundreds, of potential infractions on every damn server.

Even if you had a way of dealing with that problem, there’s another one. Let’s say the system flags a potential infraction, then the human admin looks at the server and determines it was a false positive. What now? How do we ensure that the system won’t flag the same thing again, so that the admin doesn’t have to inspect it again the next day?

Even if the admin had a way of putting some kind of a mark directly inside of the game database, saying “this is okay”, what happens when the build changes? Maybe the owner adds just one wall or replaces it with a door or places a torch or removes it. Should it now be considered different enough to be flagged again? If not, you risk false negatives.

At this point, I’m hoping I’ve explained it well enough to make everyone understand that the problem, as stated originally, is not solvable with technology.

Now let’s deal with some typical objections to this whole explanation :wink:

“Perfect is the enemy of good.”

The most typical objection is that “you don’t need to completely eliminate human reporting” or “you don’t need the system to solve every problem”. The idea is that if you build this magical system, it will take care of some problems, so that both the players and the admins can have a much easier time dealing with the rest.

Fair enough, but the devil’s in the details and in the subsequent cost-benefit analysis :wink:

How much of the problem do you want to solve with tech? The idea is to make the cost of the day-to-day admin chores low enough that a handful of people they have right now can do a good job, right? So even if your system doesn’t have to catch everything, how do you deal with all the problems I’ve described above? Is developing such a system feasible? Is the cost worth it? Will the use of the system add day-to-day admin costs that aren’t currently present?

“X shouldn’t be an infraction.”

This is also a typical objection and it comes in equal proportion from hordes of PVP players who think that their favorite abuse of the building system should be allowed and from a handful of free-speech-absolutists who get triggered by the idea that a privately owned platform isn’t obliged to provide its services to trolls.

At any rate, this objection is completely off-topic in a discussion like this. People who disagree with the rules themselves should go start a discussion about how they want the rules to change. This is a discussion about how to enforce the existing rules.

“AI can do it.”

No, it can’t. Prove me wrong. I’ve worked 4 years with AI and ML, but that was before the LLMs became the hot new breakthrough, so there’s your chance: explain how an LLM can solve this. Who knows, maybe you have a great idea that would be fascinating to read about.

But do bear in mind all that I said about false positives and resolving them :wink:


If you’ve actually read through all of that, thank you for coming to my TED Talk :stuck_out_tongue:

Of course, I’ve been here long enough to know that it’s highly unlikely that even those who read that wall of text will actually change their minds, or even offer a decent counter-argument. But hey, gotta try, right? :wink:

But even if you didn’t read what I wrote, there’s one little thing I would like you to think about. Have you ever tried to reach a customer service and had to deal with an automated system that asked you to say what you’re calling about and then that system tried to answer what you said? Have you ever tried using one of those sites that has this chat feature where you can type in your problem, only to realize you’re not even chatting with a human? Yeah, those are the result of some company thinking they can cut down on customer support costs using technology.

That’s what you’re asking for.

4 Likes