Clarification on code flagged for safety review

I do find it ironic how ROBLOX is now enforcing the rules regarding cursing for something as private and invisible as comments in scripts, while one of their somewhat old internal physics tests have a script with an expletive in a comment that was presumably left by some ROBLOX engineer down the line who re-read the script and was shocked at the spelling in the very old script.

23 Likes

You’ve described the exact reason Roblox is doing this though, the entire issue is the fact that some user’s are seeing these vulgar comments in code, and clearly that code is not very private if you have access to it.

I do agree that completely private code should not be moderated, but any code that is publicly available, such as what you just provided, should be moderated.

1 Like

I think this proves almost everyone’s point about how stupid this is.

Sonetimes using profanity is a good way to laugh about some stupid code you wrote or a good way to motivate yourself, and when you aren’t allowed to even use profanity for your own use, programming just becomes mundane and boring.

And the fact that roblox engineers do it themself, just shows how stupid it would be to enforce this, especially on private scripts.

Do you have the place link to confirm this?

I agree that Open source code that anyone can access should be moderated, to protect users. However it’s still very much pointless moderating private code that only a few people will ever see.

3 Likes

Just because Roblox engineers did it in the past doesn’t prove that this is “stupid”. Anyone can realize that they were wrong even if it becomes hypocritical, in this case Roblox realized allowing profane language in code is wrong. Using profane language in any code is just unprofessional anyways, if you’re using profane language in your code whilst working with other developers, then Roblox is correctly moderating this.

Once again, the only code that should not be moderated is fully private code which nobody has access to, meaning that the code is not on a place with team create enabled or that the game isn’t uncopylocked.

Also the code that @Borsy provided is open source.

I have begun warning others against development on the platform on more publicly accessible media already. Unless they change, I will not stop.

3 Likes

Looks like it. That, or it’s being moderated by some AI that’s being rewarded for true positives and not punished for false positives.

3 Likes

They already have.

DevRel has responded over weekends before multiple times. Even then, it is now Monday at noon in my timezone, and the beginning of the work day over at Roblox’s. We should be getting responses because the amount of complaints on this thread should be boosting the “respond to concerns” priority up to math.huge.

In order: Yes, and no.
They moderate comments and other content that end users will never see.

That policy was changed to zero tolerance recently too.

And if Roblox is going to change everything to zero tolerance on us, developers must change to zero tolerance on Roblox.

That is precisely their attitude. This is a repeat pattern that’s been going on since I’ve been watching the DevForums (a thing that’s been going on for ~3-4 years now, part of which while signed out)

This doesn’t surprise me at all, and if a developer gets desperate I could definitely see them using the last of any cash they have to drive down and attack Roblox directly in person.

That’s news to me, but it’s good to know Roblox is being hypocritical.

Doubtful. There’s an implementation used for quickly approximating roots of a number that people have linked a couple times here that has a couple of swears in it. This code is used in a professional capacity, as far as I know. Bottom line, the contents of a script don’t matter so long as the output is professional and appropriate. The client, unless they poke around in your script/game/software (which, might I add, violates the EULA of most software) has no way of knowing what exactly you’ve put in your code.

All in all, looks like it’s time to give Roblox a reality check and organize ourselves into a single entity that they have inadvertently enslaved themselves to.

10 Likes

I also find it ridiculous how people are being outrighted banned because of comments in a script that never intended to be seen by other people (including moderators). I am sorry but moderation abuse has happened and the amount of false positives on roblox that would cause a moderator to access my code is definitely quite high. For example: My username is flagged by the filter in game. So does that mean all my code with my username is going to read? Official_MrBob101 is certainly not an inappropriate name (yet the filter thinks otherwise)! I have absolutely no problem with moderators reading open sourced code because that is publicly seen and can cause issues in terms of bypassing rules and filters. But code meant for my eyes only? Or even two people in a TC? I find that absurd and downright draconian. I agreed with you guys removing closed source modules as that was a security risk. But my private code being read by strangers? Absolutely not!

6 Likes

If this applies to code that only I and my team will be able to see, I find this change completely unnecessary. I don’t want sensitive data like API keys to be seen by people who I do not know. I think this also opens the door to a lot of false bans like other users have pointed out.

3 Likes

I only semi-support this update, there should just be an automatic filter that keeps developers from putting bad things in their scripts. Nothing else.

The reason they’re doing this is understandable though, there’s many people who will hide infection scripts and profanity in some free models.

If someone was punished for having inappropriate content in their scripts, that’s fine. But is it really necessary to have an actual person check every script on Roblox?

That just takes too much time and money plus everyone suffers from it, just go with an automatic filter so everyone feels more safer.

WHAT THE FILTER DOES:

If the automatic filter finds a curse word or detects an infection script, the script either gets sent to review by an experienced scripter who works closely with Roblox. Or it just tags any inappropriate content.

If it finds nothing, nothing happens to the code or the player.

2 Likes

The problem is, scripting is such a complex thing that something a filter may consider “bad” is not bad at all; look at how many false-positives Roblox’s standard chat has.

Also, free models are already reportable for script content.

No matter who you ask, this is just Roblox acting WAY too preemptively. They’re concerned about a particular issue, and are pushing out dumb changes/actions. No different than half of the governments of the world do pretty regularly, and what TP hoarders are doing now.

1 Like

I dislike this new system on code reviewing, as any developer’s code is seen by people that they don’t know, often without the developer knowing it. This often leads to false and/or confusing moderation actions on developers. This system also puts Roblox’s moderation team into inefficiency, as they could be using their time more wisely and go after serious violations of the terms of service, such as inappropriate games. I would not like a third party peeking at every line of code that I’ve written in a game, possibly hoping to catch something that they consider to be “bad” and using that as an excuse to put my game under review or even take down my game entirely. There’s already been some cases of questionable warnings/bans such as this and I’m suspecting if some of the moderators have a personal bias against certain games/users but it seems like that moderator was scrutinizing every bit of code in his game just to find something with a supposedly inappropiate meaning and take the game down.

An automatic filter would definitely work, as it is completely unbiased AND doesn’t put any extra work onto the Roblox moderators. The automatic filter would work like the Roblox chat filter, only less aggressive and checks comments only:

// ####
print "Hello world"

Note that the print script is not filtered, just so that if the filter is overreactive, code in people’s games won’t be messed up.

3 Likes

True, the only thing the filter should do then is notify an experienced scripter to overlook the moderated code if it finds something bad…instead of automatically handing out punishments.

Gonna update my orginal post in a bit.

Roblox’s moderation processes are confidential, and I’m not privy to how they work exactly. This is all conjecture based on how customer support works in the industry, so yes, Roblox staff could be typing all of those answers out manually. But more likely (almost guaranteed) they are using macros for this.

That would not perform very well in practice, unfortunately as it isn’t what the filter is trained for. However I think that your idea is definitely the right attitude towards it.

I mean, if they suddenly have the ability to magically see everyone’s scripts…making a filter that overviews scripts shouldn’t really be difficult either tbh…

A huge change deserves a huge update honestly.

1 Like

Not quite. The filter is based on a natural language model - meaning that it is designed to understand primarily user chats. There would be a whole host of problems with filtering comments, and the result wouldn’t always end well.

Plus, unfortunately, the idea is not just about filtering comments. Filtering strings and other things, which really can’t be done automatically, is also required.

1 Like

This is a great comment. You are absolutely right. If there is a rogue administrator, we all know what happens…

4 Likes