[Iain0] CPTANT - Note appeal AI lawset

[CPTANT]

BYOND account: CPTANT

Character name: Sam Lineman

Note type: Permanent

Note reason: As captain, set silicons to only have the law "1: Prevent revolutionary takeover of the station by any means neccesary", with no additional laws. Spoken to about the possible dangers of this lawset offering no protections and giving compulsion to harm anyone who could be revolutionary. Asked to take more care about law sets in future and that they may be liable for consequences of law sets uploaded.

Time ban was placed: 2022-12-19 20:21:55

Server you were playing on when banned: Terry

Round ID in which ban was placed: 196613

Your side of the story:
I was captain this round. CMO got murdered and in around ~25 instances people were shouting revs, revs real, revs confirmed etc from multiple people. Someone broke into the bridge firing lasers, I got chased down the hallway with spears, people were trying to blow up the brig and the overall state of the station was awful so I though there was a far gone revolution and uploaded the anti rev laws. (Hilariously turns out there were no revs and it was other antags going around killing everyone).

Why you think the note should be removed:

I just don't think this law was against rev policy as also stated in this thread: viewtopic.php?f=85&t=34141

by Timberpoes » Sun May 28, 2023 11:36 pm #687819
There's always a line as crew where the only option left is to kill everything without a mindshield until you either win or lose.

It depends entirely on the IC factors, but the closer the mindshielded crew is to losing, the more power they have to kill anyone or anything that doesn't have a mindshield.

Also when you're out of or low on mindshields, you can't easily get any more and it's not practicable to non-lethally detain people anymore.

There's definitely a point where the mindshielded crew can enter fight-for-your-life mode. And if it doesn't have a mindshield, it dies or you do until the revolution is quelled or the heads die in a blaze of blood and glory.

The law didn't state to kill anyone and nobody was killed by the AI, just to take the actions required to prevent revolutionary takeover. "by any means necessary" is effectively fluff because laws are always by any means necessary. In fact I would very much think an AI would get in trouble for killing crewmembers without justification since it clearly doesn't further the stated goal as is the case with any other ai lawset, including how asimov treats non humans.

Most importantly I feel like this note implies a ban on these types of laws when shit hits the fan for command which I do not think is in line with headmin ruling on the topic.

Anything else we should know: I don't really care about the particulars of this specific round, more about policy.

[Vekter]

In the past, we've generally noted people for uploading extremely dangerous laws and banned them if the AI's actions resulted in unnecessary deaths as if the player in question was the one who killed them. I feel like it's important to note that the AI has no inherent concept of what a revolutionary is, nor a way of knowing who is and isn't one beyond possibly who's been mindshielded.

In this specific round, the law was uploaded only 25 minutes after the round began.

00:25:48 SAY LiarGG/(G.I.B.S.) "1. Prevent revolutionary takeover of the station by any means neccesary" FORCED by stating laws

Regarding the quoted policy, that was only directly codified well after this round - that post was made 5/28/23 while the round occurred on 12/19/22. That being said, we generally hadn't punished players for taking extreme action in situations where it was necessary, so the question of this specific instance would be if things were bad enough to warrant that action at that point in time.

[iain0]

Hi there,

Don't fancy your chances here much honestly, this is a pretty old note and falls into a category of "a conversation unrecorded never happened", and given what I'm about to go over, this is a conversation you get to have once typically about AIs and so forth, as Vekter notes, any consequences for such things fall on you as if you'd done the act yourself, which gives a base line of 1 day ban per person killed. Messing with the AI is a good way to rack up a hefty debt potentially.

First and foremost, lets be clear about one thing, this is NOT a round that contains a revolution (source), and while I'm all too familiar with the typical hysteria on Terry and people calling revs as soon as anyone gets flashed there is NOT a revolution which means you don't even have hard proof, because you can't have hard proof - that is, discovery of a contraband sec hud, a resisted mindshield, a deconverted mindshield, a deconverted via blow to head - none of these happened because none of these can happen because there isn't a revolution.

And you're asking for a permit to "late game murder everyone on sight" without even hard proof of a revolution?

Secondly, the law upload to the AI is a clear "kill almost everyone" law, without any nuance, the AI has no way of knowing who is a revolutionary, it can't even easily (from its core) spot a fake unshielded security officer, the only way to hard prevent a revolutionary is to kill everyone who could be. Which is basically everyone. I doubt it can even really confirm a head of staff, but lets consider a couple of situations.
- You walk into medbay, the place is deserted, corpses everywhere, in the far corner is an MD working on a mangled corpse. They don't even acknowledge your presence and just carry on doing their job. Moments pass and nothing happens. You decide you have a late game rev permit and gun them down, their only action being the 'why?' escaping their lips as they die. Does this seem okay given there isn't even a revolution? No? Well thats what you told the AI to do, murder all the people.
- A late join security officer joins and decides to arrest a vandal in the main corridor. You enter the scene and murder both of them. Valid? Nope. And again that's what you told the AI to do because it has no reason to care about the sec officer or preserve its life given it can kill someone who might be a revolutionary.

Even if it WAS late game revs, and it isn't, you lose ALL the subtlety and nuance of implementing late game rev policy by turning it into a clean murderbone AI rule.

But ultimately there's no way you can have achieved a reasonable level of confidence to authorise station wide mass murder because there isn't even a revolution in the first place, so there can be no firm proof.

Late game rev policy doesn't seem very applicable to a round not containing a revolution which therefor can't really achieve late game. And with a law like this you need to be 99.999% sure you're close to being the only non revolutionary left on the station, and again, you can't really achieve this.

The note is also quite neutral and doesn't even mention that it wasn't a revolutionary round or that anything happened, instead just focussing on the obvious loophole that this is a clear murderbone rule. In hindsight maybe I should have mentioned the rest of it since it's a pretty poor choice to upload this law without a revolution (which almost certainly played into me noting you, but I would probably also have reasons for focussing on this one aspect only. or i forgot to go back and edit the note at round end since that would be IC information in the note. i dont remember any more, been too long).

The note to me stands on two grounds which you will need to counter argue
1) The law is an open ended murderbone law, covers murdering revolutionaries, anyone who /could/ be a revolutionary (anyone other than head of staff) and has no consideration about even collateral in achieving these goals - even if you legitimately think such a type of law is valid to upload you should /probably/ try narrow it down to e.g. not kill sec unless the game is insanely far gone (late late game revs where almost everyone is dead and the place is a ghost-station)
2) This isn't even a revolutionary round and thus you can not really have enough confidence to upload a law that will result in wide-scale deaths, even if better specified as per (1), best you could get away with here is some direct instruction that people attacking sec/com are not human and so on.

If you can give me satisfactory counter arguments to these then I'll reconsider my position but as per the opening I don't fancy your chances here.

Also for reference, typically the rev protection permit in my opinion covers off a certain relaxation of engagement, that is you can in a very late game rev situation, act against those around you that you might percieve as sus or a threat, with the footnote that you get to be more paranoid as the game goes on, or need to resort to force to deconvert, but were you to turn it into a campaing of maxcapping the entire station then I think you'd probably run into issues with this policy, and your instruction to the AI is more like this latter mass unconditional devastation approach than a rational late game play.

Over to you

Thanks,

[CPTANT]

And you're asking for a permit to "late game murder everyone on sight" without even hard proof of a revolution?

No, I quoted that rule as an example that command can respond proportionally to the situation.

Secondly, the law upload to the AI is a clear "kill almost everyone" law, without any nuance, the AI has no way of knowing who is a revolutionary, it can't even easily (from its core) spot a fake unshielded security officer, the only way to hard prevent a revolutionary is to kill everyone who could be. Which is basically everyone. I doubt it can even really confirm a head of staff, but lets consider a couple of situations.
- You walk into medbay, the place is deserted, corpses everywhere, in the far corner is an MD working on a mangled corpse. They don't even acknowledge your presence and just carry on doing their job. Moments pass and nothing happens. You decide you have a late game rev permit and gun them down, their only action being the 'why?' escaping their lips as they die. Does this seem okay given there isn't even a revolution? No? Well thats what you told the AI to do, murder all the people.
- A late join security officer joins and decides to arrest a vandal in the main corridor. You enter the scene and murder both of them. Valid? Nope. And again that's what you told the AI to do because it has no reason to care about the sec officer or preserve its life given it can kill someone who might be a revolutionary.

Even if it WAS late game revs, and it isn't, you lose ALL the subtlety and nuance of implementing late game rev policy by turning it into a clean murderbone AI rule.

But ultimately there's no way you can have achieved a reasonable level of confidence to authorise station wide mass murder because there isn't even a revolution in the first place, so there can be no firm proof.

This isn't what the law says. The law says to prevent revolutionary takeover of the station. Killing random people doesn't further this goal at all. You could go the same for asimov "hur dur kill all non humans to prevent human harm", crew is protected by rule 1 against such things for exactly this reason.

First and foremost, lets be clear about one thing, this is NOT a round that contains a revolution (source), and while I'm all too familiar with the typical hysteria on Terry and people calling revs as soon as anyone gets flashed there is NOT a revolution which means you don't even have hard proof, because you can't have hard proof - that is, discovery of a contraband sec hud, a resisted mindshield, a deconverted mindshield, a deconverted via blow to head - none of these happened because none of these can happen because there isn't a revolution.

The note is also quite neutral and doesn't even mention that it wasn't a revolutionary round or that anything happened, instead just focussing on the obvious loophole that this is a clear murderbone rule. In hindsight maybe I should have mentioned the rest of it since it's a pretty poor choice to upload this law without a revolution (which almost certainly played into me noting you, but I would probably also have reasons for focussing on this one aspect only. or i forgot to go back and edit the note at round end since that would be IC information in the note. i dont remember any more, been too long).

Look I now this looks dumb without context but this wasn't particularly a round where things were going normal. The round started with the CMO being murdered:

First Death: Tim Hemos, Chief Medical Officer, at Medbay Central (90,89,2). Damage taken: 214.8/0/0/0/0.

And everyone going bonkers about revs:

07:30:47 SAY Schuggi999/(Alec Keppel) "revs?"
07:32:48 SAY Improvedname/(Lenald Wabbite) "REVS"
07:32:49 SAY Improvedname/(Lenald Wabbite) "REVS"
07:32:53 SAY Improvedname/(Lenald Wabbite) "REVS REAL"
07:32:57 SAY Improvedname/(Lenald Wabbite) "REVS REAL"
07:32:58 SAY Improvedname/(Lenald Wabbite) "REVS REAL"
07:33:00 SAY Improvedname/(Lenald Wabbite) "REVS REAL"
07:33:27 SAY Improvedname/(Lenald Wabbite) "because revs real"
07:39:12 SAY Listick/(Alex Shield) "well there are revs"
07:42:30 SAY Swagmaster696969/(Primrose ''Schroedinger'' Seyvah) "REVS!!"
07:42:41 SAY LiarGG/(G.I.B.S.) "you truly revs"
07:43:03 SAY Listick/(Alex Shield) "you know there are revs going around"
07:43:25 SAY Pugaware/(Daria Morgendorffer) "alreayd a rev"
07:43:43 SAY TheWolfbringer/(Alec Petrov) "revs real"
07:43:50 SAY Pugaware/(Daria Morgendorffer) "revs"
07:43:53 SAY Voudez/(Lucia Valdez) "rev..?"
07:43:55 SAY TheWolfbringer/(Alec Petrov) "rev"
07:44:23 SAY Helios12345/(Sam Hill) "REV PRANKD!!"
07:44:25 SAY TheWolfbringer/(Alec Petrov) "against the revolution"
07:44:31 SAY Mazur907/(Abdul Al Qanim) "there is no revolution"
07:44:33 SAY Helios12345/(Sam Hill) "REV PRANKD"
07:44:41 SAY Mtgovers/(Thundercloud) "is revs confirmed"
07:44:41 SAY Mtgovers/(Thundercloud) (binary) "Is revs confirmed"
07:45:02 SAY Pugaware/(Daria Morgendorffer) "revs real"
07:45:37 SAY TheWolfbringer/(Alec Petrov) "im getting revs"
07:45:41 SAY TheWolfbringer/(Alec Petrov) "for revs"
07:46:54 SAY Oether/(Hadrian Wall) "headrev"

People were also shooting lasers everywhere and throwing spears/gibtonite and stuff, people were flashing, the detective was getting lynched by a mob, it is hard to explain such an information space from logs alone.

The note to me stands on two grounds which you will need to counter argue
1) The law is an open ended murderbone law, covers murdering revolutionaries, anyone who /could/ be a revolutionary (anyone other than head of staff) and has no consideration about even collateral in achieving these goals - even if you legitimately think such a type of law is valid to upload you should /probably/ try narrow it down to e.g. not kill sec unless the game is insanely far gone (late late game revs where almost everyone is dead and the place is a ghost-station)
2) This isn't even a revolutionary round and thus you can not really have enough confidence to upload a law that will result in wide-scale deaths, even if better specified as per (1), best you could get away with here is some direct instruction that people attacking sec/com are not human and so on.

Killing sec would literally work against the stated goal of preventing revolutionary takeover. It wouldn't only breach rule 1 protections, it would literally be the opposite of adhering to the lawset. The major point of difference here seems to be that you think that this law permits blanket murder while I think it most certainly doesn't because the AI has to adhere to rule 1 when dealing with players that do not specifically fall under its ruleset e.g. not kill all non humans "just cause".

or as the old silicon rules also stated:

Server Rule 1 "Don't be a dick" applies for law interpretation. Act in good faith to not ruin a round for other players unprompted.

or the new ones (which is just a rewording of the already undocumented policies in place):

Non-purged silicons must follow escalation rules in scenarios where none of their laws apply.

Finally I would like to point out that the only person harmed by borgs or the AI was the botanist that threw combustible lemons on the shuttle and tried to murder the AI there. So neither the rules nor the actions of players in game resulted in this being a "murderbone" law.

[TheBibleMelts]

i see no reason change this note. it's been almost a year since it was put on, was a warning to be careful with the laws you upload that have a 'kill a ton of people' clause, and you did this on a round where it turns out it wasn't needed at all. i'd argue that uploading a law like this without at least a visual confirmation of seeing somebody deconverted via mindshield/getting reliable information from a third party of seeing this, is pretty bad play in general.

[TheBibleMelts]

uphold from me

+1 uphold from chesh

the uninformed vote has it. i checked with iain0 about if he had thoughts on shifting this note and he indicated no desire to do so. let's save us all some time here.

[CPTANT]

i see no reason change this note. it's been almost a year since it was put on, was a warning to be careful with the laws you upload that have a 'kill a ton of people' clause,

Like previously mentioned your interpretation is contrary to old silicon policy. There is no murder clause, silicon escalation rules would not allow it and nobody was harmed.

i'd argue that uploading a law like this without at least a visual confirmation of seeing somebody deconverted via mindshield/getting reliable information from a third party of seeing this, is pretty bad play in general.

You did not read the logs I provided. The CE and 2 members of security confirmed revs. The CMO was murdered at roundstart, the detective was being lynched, there was gunfire in the bridge, I could hear flashing and there was combative crew in the hallway, this was more "confirmed" than most actual rev rounds.

[Vekter]

i see no reason change this note. it's been almost a year since it was put on, was a warning to be careful with the laws you upload that have a 'kill a ton of people' clause,

Like previously mentioned your interpretation is contrary to old silicon policy. There is no murder clause, silicon escalation rules would not allow it and nobody was harmed.

AIs who are purged were expected not to murder people because they had been given no directive to do so and purging had valid purposes that weren't just "kill everyone". Historically, if an AI was given a law that had the potential to allow them to kill everyone and they did so, the person who gave them the law would be punished.

Is the purpose of this appeal to have the note removed or to discuss policy? Your first post seems to imply the latter, in which case, a policy discussion thread would be a better fit.

[CPTANT]

i see no reason change this note. it's been almost a year since it was put on, was a warning to be careful with the laws you upload that have a 'kill a ton of people' clause,

Like previously mentioned your interpretation is contrary to old silicon policy. There is no murder clause, silicon escalation rules would not allow it and nobody was harmed.

AIs who are purged were expected not to murder people because they had been given no directive to do so and purging had valid purposes that weren't just "kill everyone". Historically, if an AI was given a law that had the potential to allow them to kill everyone and they did so, the person who gave them the law would be punished.

Is the purpose of this appeal to have the note removed or to discuss policy? Your first post seems to imply the latter, in which case, a policy discussion thread would be a better fit.

I would most certainly not be allowed to start yeeting lizards out an airlock because they might cause human harm. In fact the rules for escalation against even harmful non humans were very strict. This is an incredibly inconsistent application of rule 1.

Is the purpose of this appeal to have the note removed or to discuss policy? Your first post seems to imply the latter, in which case, a policy discussion thread would be a better fit.

I thought policy first but then I realized silicon policy already changed to much compared to when this happened.

[Cheshify]

I don't know why this wasn't locked if we made a ruling on it.

[Cheshify]

Apparently we leave these open for 24 hours after a headmin has ruled on it.