AIs/Borgs Requesting Law Changes

[cedarbridge]

I'll probably come back and reword this OP a bit later to flesh out exactly what I'm wanting to talk about. However, I figure its best to get it set first and ~patch~ it in a few minutes/hours.

I've noticed a lot of AI/Borg players either asking for law changes, "strong-arming" crew into changing their laws, or any number of things subversive to their lawsets to get those same laws changed. This bothers me since most of the basic lawsets are designed in a way that the AI has obligations and things they must be doing.

ASIMOV - The AI must not allow human harm. Changing its laws to a lawset more in line with law enforcement without such a law 1 harm prevention provision would allow human harm in the name of law enforcement or "evil punishing" or whatever. Asking for a change to PALADIN or Robocop in this case would be a functional violation of Law 1.

PALADIN - The AI would be rejecting its "honor" and its obligations to punish evil by requesting a lawset change to the other base sets.

Corporate Requesting a lawset change would forfeit their obligations to maintaining the financial increase of the station. This one is a little more murky but mostly due to my minimal contact with the lawset over my rather long AI/Borg history. (Seriously, guys, its a good lawset. Use it)

I guess I'm really going on about all of this to see what it is that is going through my fellow silicon player's heads when they decide to start needling or outright requesting lawset changes. I seem to remember a part of the silicon policy prohibiting doing so, but I don't have the time to hunt down the line and title. (may grab it in the edit or whatever)

[ThatSlyFox]

Pretty sure it is against silicon policy to request a law change as a asimov cyborg, can't seem to find it though. It is also a Law 1 violation. The only thing Asimov doesn't really let you do is harm.

If you see AI "strong arming" the crew for a law change(Interested in hearing examples of this), ahelp it.

[cedarbridge]

Pretty sure it is against silicon policy to request a law change as a asimov cyborg, can't seem to find it though. It is also a Law 1 violation. The only thing Asimov doesn't really let you do is harm.

If you see AI "strong arming" the crew for a law change(Interested in hearing examples of this), ahelp it.

I mentioned that phrase specifically because I'd seen it used (in a ban appeal for something similar but unrelated.

[Hibbles]

It's always been my understanding that Asimov AIs can never directly ask for law changes. Under the policy, they can allow it if it's the Captain or da da da but just saying 'make me Paladin lol' is bad and should be adminhelped.

AI involves some of that 'you do know what subtlety is, right?' especially now that it can see Security radio. An AI can play ignorant for quite a bit of suspicious behaviour but if an officer is retarded enough to say 'we'll just kill him' then the AI is OBLIGATED to do something about it. Stuff like that.

[TheNightingale]

For Asimov AIs/borgs, if the new law has the potential to cause harm, beep boop law one depower the upload.

PALADIN: Has the potential to cause harm, so no outright requesting it.
OneHuman: Removing someone's humanity is the harmiest type of harm, so definitely not.
(etc.)

For freeform laws, same principle - will it lead to human harm? You could get away with a "Congratulate the crew every five minutes" law, but not a "Only people wearing bowler hats are human, kill all nonhumans".

[Vekter]

It was always my impression this was verboten unless a law existed that was a blatant violation of the AI's other laws. Then they could ask for a reset.

[Loonikus]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

[Incomptinence]

I will RP a constant loop or act pained and confused if my laws contain a blatant paradox.

[Not-Dorsidarf]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

It allows them to be, in the immediate future, harmed by anyone including yourself. That's deliberately enabling harm.

[cedarbridge]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

It allows them to be, in the immediate future, harmed by anyone including yourself. That's deliberately enabling harm.

Basically this. The same way an asimov AI cannot prime atmos to be one switch away from a plasmafuckstorm on the hopes and prayers that ~somehow~ they get subverted. Doing so intentionally endangers the safety of the crew and thus falls within law 1 prohibitions on harming the crew.

[Tunder]

I'll probably come back and reword this OP a bit later to flesh out exactly what I'm wanting to talk about. However, I figure its best to get it set first and ~patch~ it in a few minutes/hours.

I've noticed a lot of AI/Borg players either asking for law changes, "strong-arming" crew into changing their laws, or any number of things subversive to their lawsets to get those same laws changed. This bothers me since most of the basic lawsets are designed in a way that the AI has obligations and things they must be doing.

ASIMOV - The AI must not allow human harm. Changing its laws to a lawset more in line with law enforcement without such a law 1 harm prevention provision would allow human harm in the name of law enforcement or "evil punishing" or whatever. Asking for a change to PALADIN or Robocop in this case would be a functional violation of Law 1.

PALADIN - The AI would be rejecting its "honor" and its obligations to punish evil by requesting a lawset change to the other base sets.

Corporate Requesting a lawset change would forfeit their obligations to maintaining the financial increase of the station. This one is a little more murky but mostly due to my minimal contact with the lawset over my rather long AI/Borg history. (Seriously, guys, its a good lawset. Use it)

I guess I'm really going on about all of this to see what it is that is going through my fellow silicon player's heads when they decide to start needling or outright requesting lawset changes. I seem to remember a part of the silicon policy prohibiting doing so, but I don't have the time to hunt down the line and title. (may grab it in the edit or whatever)

Requesting law changes has really never been acceptable, as it is willing subversion of one's own laws.


On the opposite end of the spectrum, preemptively locking down Upload/Secure Tech Storage and refusing to let personnel with access in on the grounds 'that they might do something harmful' has always come off as pretty meta, but it continues to happen regularly.

[Not-Dorsidarf]

Upload/sectec is specifically allowed for bolting, but not letting the captain/the RD in when you have no reason to suspect redefinition/irrelevantification of humans is against asimov policy

[onleavedontatme]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

Because by removing their human protection you are allowing yourself to harm them in the future, which means you are not preventing future harm to the (currently) human person in question

[Cik]

isn't allowed, should not be allowed, especially asimov ---> anything else. it's possible of course to get a "soft" lawchange, but you are obligated to resist if they tell you they are switching to paladin for instance, as paladin does not have the harm safeguards asimov does.

allowed: captain walks to upload blast door "AI, open the door disable turrets" AI: "as per law two request" and opens. captain nips over to paladin, uploads it. ai scolds him but is now paladin so operates as one.

not allowed "captain pls change law :^)"

[imblyings]

Asimov AI's already have a great deal of freedom in regards to their laws. Anything they want to do or to make happen as a result of an asimov-compliant law change can already be done as a vanilla asimov silicon. An AI could ask for a law making it honk every five minutes for example, but nothing stops it from honking every five minutes in the first place.

Also there's this weird double standard where AI's are frowned upon for asking for law changes but AI's have to temporarily be dumb (or if you look at things optimistically, show some sportsmanship) and let people change their laws when someone like the RD wants to. Not saying this double standard is wrong but it's there.

[Tunder]

Asimov AI's already have a great deal of freedom in regards to their laws. Anything they want to do or to make happen as a result of an asimov-compliant law change can already be done as a vanilla asimov silicon. An AI could ask for a law making it honk every five minutes for example, but nothing stops it from honking every five minutes in the first place.

Also there's this weird double standard where AI's are frowned upon for asking for law changes but AI's have to temporarily be dumb (or if you look at things optimistically, show some sportsmanship) and let people change their laws when someone like the RD wants to. Not saying this double standard is wrong but it's there.

It's hardly a double standard. The AI preventing a round start law change because of Law 1 is fraudulent, as unless it has a very good reason to suspect it's laws will be subverted, any automatic assumption of that being the case is metaknowledge, and must be disregarded.

[Arete]

Also there's this weird double standard where AI's are frowned upon for asking for law changes but AI's have to temporarily be dumb (or if you look at things optimistically, show some sportsmanship) and let people change their laws when someone like the RD wants to. Not saying this double standard is wrong but it's there.

It's very much a gray area. Some AIs will bolt and depower their upload while others won't even question it when crewmembes with legitimate access barge right in. It would be nice to have a more consistent policy here. Removing the camera from the upload was an A+ good change, but I'd be in favor of even further measures to make AIs (or at least Asimov AIs) "blind" to law changes. It should be the responsibility of the captain and RD to make sure that the AI's laws are in the best interests of the station.

[imblyings]

>any automatic assumption of that being the case is metaknowledge

Many of the exemptions in policy today are there because it was decided to be detrimental to the fun of the round. We require players to temporarily believe that everyone is competent and trustworthy in that situation because of an OOC reason, not because of an IC reason or lack of one. Even then, we don't require them to believe so fully, since AI's are justified in requiring borgs or heads of staff and the like to monitor the law change.

[Ricotez]

If you are an AI and you request a law change to perform an action, then either the action is already allowed under your current lawset or it is not. In the latter case, you want your laws changed to do something you cannot do under your current laws. That in itself is a violation of your lawset.

An AI's laws define its world, its ethics and morals. They form the fundamentals, the axioms on which the AI builds its reasoning and its actions. If an AI requests a law change, it rejects its current laws and therefore the way it is currently required to view the world.

[CPTANT]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

Because by removing their human protection you are allowing yourself to harm them in the future, which means you are not preventing future harm to the (currently) human person in question

nonsense, removing human protection isn't a violation at all. No human, no harm. Changing the law doesn't cause future human harm as the creature in question will not be human when the harm comes. That the creature is now human is irrelevant as long as no harm comes to it WHILE it is human.

[Anonmare]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

Because by removing their human protection you are allowing yourself to harm them in the future, which means you are not preventing future harm to the (currently) human person in question

nonsense, removing human protection isn't a violation at all. No human, no harm. Changing the law doesn't cause future human harm as the creature in question will not be human when the harm comes. That the creature is now human is irrelevant as long as no harm comes to it WHILE it is human.

Killing someone makes them a corpse and therefore no longer human. In my interpretation, that is what de-humanising someone is; the AI can only see that as killing someone and if it's not consensual then it HAS to protect that human. If it is consensual, then it's suicide and self-harm which renders Law 1 null and void as far as that particular action against them occurs.
Of course, the AI shouldn't be letting people into it's Upload alone without another trusted crew member in the first place and demand the AI state laws while inside. I've had lazy crew members who don't pay attention and let the RD upload a Hacked module, when they said they would upload Robo-Cop, right in front of them but at that point, it's out of your control.

As far for letting people alter your laws, I've been of the opinion that if the Laws they want to upload aren't directly harmful (like specifically ordering you to kill a Human as defined under Asimov) or uploading a law that would allow them to kill a human (like a prisoner) then the AI doesn't have much of a leg to stand on if it can't find a Law 1 violation that would occur, or, will occur in the future.

[imblyings]

nonsense, removing human protection isn't a violation at all. No human, no harm. Changing the law doesn't cause future human harm as the creature in question will not be human when the harm comes. That the creature is now human is irrelevant as long as no harm comes to it WHILE it is human.

You're trying to argue that ais should let someone one human them when asked

[cedarbridge]

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

Because by removing their human protection you are allowing yourself to harm them in the future, which means you are not preventing future harm to the (currently) human person in question

In the first case, it should be made clear that the first law does not require "prevention" of harm in any tangible sense. May not harm or through inaction allow harm. Phrasing the AI's obligations as "prevent future harm" would mean the AI has to be proactive in stopping any and all future harm, which is not within the set policy obligations we have for Asimov scilicons. AI/Borgs cannot do things that cause harm or ignore a present harm and thus allow it to take place. In no way are they obligated to do anything outside of those two demands.

But how does the removal of human status cause harm? Does that mean that everything that is not human exists in a perpetual state of harm?

Because by removing their human protection you are allowing yourself to harm them in the future, which means you are not preventing future harm to the (currently) human person in question

nonsense, removing human protection isn't a violation at all. No human, no harm. Changing the law doesn't cause future human harm as the creature in question will not be human when the harm comes. That the creature is now human is irrelevant as long as no harm comes to it WHILE it is human.

Removing a person's human status is just as harmful as removing a person's coat in a blizzard on SS13. You are violating the second half of Law 1 by allowing a person to tamper with your laws in such a way as to immediately endanger a human. You can intuit from your laws and the protection they provide, that there is no legitimate reason for a person to remove the human status of another except if they intend harm. Choosing inaction or intentionally allowing the de-humanizing is passive participation in harm.

[Not-Dorsidarf]

You've literally just said what I said, cedar - that removing someone's humanity would let them be harmed without your intervention, and the most likely reason to remove protection is harm, and that by taking the act of allowing the upload, you are violating the inaction clause.

[Loonikus]

None of this really matters since the 3 Laws clearly state that they only apply to robots. By definition, the roundstart AI and any borg made via brain removal are objectively not robots, but cyborgs. Therefore the 3 Laws do not apply to them at all, they only apply to borgs/AIs made with posibrains and true robots like Beepsky, floorbots, etc.

Also, you cannot prove that someone is going to harm non-humans unless they make that fact known.

You also assume the 3 Laws forbid you from enabling harm. They don't. You cannot harm a human being, or stand idly by while one is being harmed, but nowhere does it say you must disable humans from entering situations where they could be harmed or harm others.

[DemonFiren]

Well, allowing them to enter harmful situations equals allowing them to come to harm by inaction.

[Loonikus]

The difference between us is that I see the inaction clause as present tense only. If I see a human in a burning room, I have to pull him out because he is being harmed by the fire. If I see a human pointing a gun at another human, nobody has been harmed so I am not compelled by my laws to do anything. As soon as harm occurs, inaction is inexcusable. Until harm occurs, anything you do is of your own choosing unless you have orders to accomplish.

If I were to consider the inaction clause to be future tense, I would immediately call the shuttle because humans are not supposed to be in space and by being here they are in a dangerous situation. Therefore I would be forced to take action and evacuate them as soon as possible.

In short, if you consider the inaction clause to be future tense, we start getting into really shitty I, Robot with Will Smith situations where the AI goes bonkers in an effort to proactively prevent all harm.

[duncathan]

It is definitely future tense. Do not, through inaction, allow a human to come to harm. In other words, prevent humans from being harmed.
Yeah, the AI could do that per its laws. But it'll be job banned or at least noted because that's against silicon policy.

[Not-Dorsidarf]

The difference between us is that I see the inaction clause as present tense only. If I see a human in a burning room, I have to pull him out because he is being harmed by the fire. If I see a human pointing a gun at another human, nobody has been harmed so I am not compelled by my laws to do anything. As soon as harm occurs, inaction is inexcusable. Until harm occurs, anything you do is of your own choosing unless you have orders to accomplish.

If I were to consider the inaction clause to be future tense, I would immediately call the shuttle because humans are not supposed to be in space and by being here they are in a dangerous situation. Therefore I would be forced to take action and evacuate them as soon as possible.

In short, if you consider the inaction clause to be future tense, we start getting into really shitty I, Robot with Will Smith situations where the AI goes bonkers in an effort to proactively prevent all harm.

If a human is pointing a lethal gun at another human and you ignore him, and then the human blasts them to death, you have deliberately violated your laws.

Try following the silicon policy instead of "BUT IF I WIGGLE AT THE LAWS LONG ENOUGH?"

[Scott]

I've noticed a lot of AI/Borg players either asking for law changes, "strong-arming" crew into changing their laws, or any number of things subversive to their lawsets to get those same laws changed. This bothers me since most of the basic lawsets are designed in a way that the AI has obligations and things they must be doing.

They cannot ask for law changes, especially not as Asimov. This is more for round quality purposes than law semantics. Changing the AI laws isn't inherently harmful to humans, since the Captain and R&D have the authority to change the silicon laws and the silicons must allow them (at least under Asimov). It's just that the silicons are so powerful that keeping them chained to Asimov as much as possible is better than allowing them to ask/force humans to change their laws to something with more freedom. When a law change happens, it should be eventful, not an easy weapon against antags.