[Carbon policy for Silicons] Criteria for calling borgs/ai bad?

[MooCow12]

viewtopic.php?f=7&t=36217 brought into light a very common situation where a borg is called malf based on false information or information with missing context. Based on my experience the bar for declaring malf ai seems rather low, from simply ignoring dangerous law 2 requests as asimov ai to seeing an asimov borg kill a non human (Yes i was declared as malf before for killing someone who was VISIBLY NONHUMAN)


At the end of the day , silicons are not usually outed based on their gear, rather they are constantly graded based on their behavior and obligations and expectations that every other player has on them, this is a game where people act on limited information and silicons very frequently act on context which makes their actions sometimes hard to predict and it is those inaccurate predictions from the crew that results in malf calls, especially given that when a seemingly antagonistic force is present, investigation of further context is not always available or even safe to do without taking a risk and some criteria would be needed to fall back on for those split second decisions.



So I'll start with some example criteria and factors that seem to be present in terry and sybil to play around with.


1. The silicon(s) are apparently going against/contradicting the laws that you thought they had-




A. The asimov ai wont do a seemingly benign task like open eva storage for a human, (usually from hidden conflictions like command secretly countering their law 2 order on command channel.)

B. The silicons are harming someone you perceive to be a protected individual (human/Asimov)



2. The silicon(s) are breaking META BEHAVIOR-



A. Calling a silicon malf for seemingly randomly killing a nonprotected individual (any nonhuman for asimov) This declaration calls upon the ruling that silicons are normally not allowed to randomly kill without specific exceptions despite laws not protecting
them. (meta gaming????????)

B. Doing an autism project in ai sat also usually doesnt just result in the ai sat being crowded and investigated it also results in the ai being carded in case of malf status. (this one is hard to talk about because it might also call into question what the autism project is) Even nonlethal fortifications and reinforcements to the sat result in large investigations even though asimov silicons have a law 3 reason to do so. I've even had rounds where the sat/ai was investigated just because i spent alittle too much time there.

C. lol tcomms is down

D. Very mild grief like jani borg doing janiborg things



3. Guilty via association and its implications-



Calling a single borg subverted is a lot less disruptive to a round than calling a borg malf since it does not immediately put a target over every other borg and the ai, is there leniency when calling an individual borg malf/hacked/subverted and escalating to calling the ai malf if enough of their borgs are perceived a threat? (Obviously not an option if you think the ai themselves is the one that is bad which would automatically condemn
every other borg on the server even ones that are not sync'd to it due to how trigger happy everyone is)


4. Inability to investigate facts- (playing safe rather than sorry)


This is its own thing because while it doesnt result in you calling malf ai by itself it swings open the door for that to be your only option in countless scenarios.

The vast majority of players only have ONE way to investigate if silicons are bad and that is through hearsay/talking on radio, in many scenarios simply shouting that you think the silicon(s) are malf on radio immediately prompts a discussion based on everyone's perception of their behaviors and the ai / borgs sharing their laws, and context being shared. Screaming ai malf because you perceived them killing a human only for it to be explained right after on radio
that the victim was actually a nonhuman is natural.

Declaring silicon as malf is not just a reaction/response, its an opener for a larger investigation and this step serves as a thin line between declaring a silicon as bad and actually treating the silicons as evil which begs the question as to
if declaring silicons as malf should be sanctioned as long as communication is available to dispute.



TLDR: Where is the bar for declaring silicon(s) bad and is it flexible or static?

[Vekter]

This is going to be pretty much impossible to actually define. Reasonably we can say "It's any situation where you realistically would be led to believe the AI is not following their laws", but that's so fucking broad that I can't actually define it without relying on vibes.

[Higgin]

I'm going to be looking for, if it was a borg misbehaving (non-Asimov,) people investigating first if the borg is connected or not.

Then, if the borg is connected, trying to get and fix the laws.

Then checking the AI's laws directly.

Then, say it's malf.

If there's explicit malf tells before then, with fewer necessary if things are escalating (round state) or the tell is unambiguous (lockdown, overloads, doomsday obviously,) it's more acceptable to jump to malf and act accordingly to every silicon you don't know for sure to not be malf.

That's a rough decision procedure, but like Vekter said - too much to define neatly, ultimately vibes.

[zxaber]

The easiest thing to do would be avoid calling out AI/Borgs Malf, and instead call out what you see them doing. "Borgs attacking a human" opens up people to question why, which in turn can either lead to Malf actually being confirmed, or lead to the AI/borgs explaining what they were doing (IE, non-asimov laws, or that the target wasn't human after all). Calling out Malf should mean that you have plausible proof of it. Hacked APCs, for example, are probably enough to be called out (even though in reality emagged APCs also give the same effect and if you know how to tell the difference you should check).

That said, even if you don't have proof of Malf, you can take action against a borg you suspect is emagged, or a silicon team that you suspect in general has been subverted. Just, y'know, be reasonable about why you suspect such and work towards reviving the silicons when the dust settles.

[Vekter]

The easiest thing to do would be avoid calling out AI/Borgs Malf, and instead call out what you see them doing. "Borgs attacking a human" opens up people to question why, which in turn can either lead to Malf actually being confirmed, or lead to the AI/borgs explaining what they were doing (IE, non-asimov laws, or that the target wasn't human after all). Calling out Malf should mean that you have plausible proof of it. Hacked APCs, for example, are probably enough to be called out (even though in reality emagged APCs also give the same effect and if you know how to tell the difference you should check).

That said, even if you don't have proof of Malf, you can take action against a borg you suspect is emagged, or a silicon team that you suspect in general has been subverted. Just, y'know, be reasonable about why you suspect such and work towards reviving the silicons when the dust settles.

So you're saying that it's often better to avoid using out-of-character terms that describe a very specific set of actions and instead simply say what's happening, as in saying "John Doe is killing someone!" instead of "JOHN DOE BAD?"

Woah, that's crazy.

[Constellado]

I think the only time a person should be punished is if it causes atleast one Borg gets killed after the call. Dying as a Borg is the same as dying as a carbon. You get turned into a ghost and maybe get angry because you died. And, unlike what people say, it's not hard to kill a non malf Borg, especially if you are sec. Sec is never afraid to outright kill a Borg for malf calls.


However, I will say this: Being called malf as an AI is common, and normal. I believe it is part of the AIs job to convince the crew that is not true. Most of the time people see it as a boy cried wolf situation, and end up not taking it seriously anyway.

[Justice12354]

I believe it is part of the AIs job to convince the crew that is not true. Most of the time people see it as a boy cried wolf situation, and end up not taking it seriously anyway.

While the majority of this topic is rather subjective, I do not agree that the AI should make the active effort to prove they're not malfunctioning while anyone in the crew can just freely point fingers at them for the slightest motive. If a crewmember cannot be that freely accused of being a tot, why can an AI? Is it related to the capability of being able to prove their antag status or whether they're subverted? This leads to another concern of mine, which is the reduce in silicon gameplay quality by being basically griefed everytime someone slightly dislikes what you're doing. If this is a balance that we're willing to play around, then sure, but there is still a loose edge here imo

[dragomagol]

My take on this:

You should have a reasonable explanation for calling a borg bad, aka not following laws. Same vibes as someone just yelling "ARREST JOHN JONSON." Like if they're not Asimov and you're getting riled up and starting a lynch mob because the borg isn't listening to you, operator error.

If a borg gets killed or incapacitated for an unreasonable amount of time (like a long brig sentence) and the person who made the callout didn't do any kind of due diligence in investigating the issue (things including but not limited to talking to the AI, talking to the RD, locking down the borg and talking to them), then that is actionable.

[kieth4]

My take on this:

You should have a reasonable explanation for calling a borg bad, aka not following laws. Same vibes as someone just yelling "ARREST JOHN JONSON." Like if they're not Asimov and you're getting riled up and starting a lynch mob because the borg isn't listening to you, operator error.

If a borg gets killed or incapacitated for an unreasonable amount of time (like a long brig sentence) and the person who made the callout didn't do any kind of due diligence in investigating the issue (things including but not limited to talking to the AI, talking to the RD, locking down the borg and talking to them), then that is actionable.