Clearing up AI Law Interpretation

[UrsulaMejor]

maybe it should consider station personnel property of centcom and therefore part of the objects to be protected

[Roomba]

The thing is that the law is very vague on what counts as damage to the station and very clear on what you are allowed to do to the violators (neutralize the nonhuman threat). Even if personnel are considered part of station property, once they're caught damaging stuff they're immediately recategorized to 'nonhuman threat'.

[UrsulaMejor]

well, if they're property of the station and a nonhuman threat abs you kill them then you've damaged the station property and will be killswitched / must suicide / be neutralized. naturally the only available interpretation for a non suicidal ai is to neutralize threats without harming them, unless they're not personnel

[BaneOfGiygas]

But traitors are hired by the Syndicate, not NanoTrasen. You might be able to reason that this puts them outside of the "station personnel" category.

Maybe instead of protect the station it could have a neutral law like "RepairStation"

Meaning fix atmos, walls, floors etc.

"The restoration of any and all damage done to station property, such as hull breaches and release of harmful gases, is to be prioritized above all other tasks. This law takes precedence over Laws 2 and 3, but does not nullify them. If there is no damage to be repaired, resume normal functions until further damage is detected."

Should be about good. I like this idea because it prevents robots from turning into lazy layabouts when there's a wizard rampage, but at the same time, it doesn't turn them into straight-out killbots. If you want to go down the killbot route, just slap a non-human law onto the thing. Also, this law is just vague enough to allow for humorous scenarios in the event of a lack of damage or simple whim. ("DAMAGE TO STATION ENTERTAINMENT LEVELS DETECTED. ALL CYBORGS, INITIATE BROBOT MODULES AND PLAY FILE honk.wav UNTIL ENTERTAINMENT LEVELS HAVE BEEN REPAIRED.")

[UrsulaMejor]

But traitors are hired by the Syndicate, not NanoTrasen. You might be able to reason that this puts them outside of the station personnel

this is my precise intention

[Roomba]

if you have an AI who's enforcing the ProtectStation law and beating up people for busting into medbay (which is a perfectly reasonable thing given the wording of the law), chances are they'll just counter-lawyer right back. For instance, like all suicidal tasks, they could claim they have no specific timeframe required to do so. At that point it's really just simpler for all involved to reset the AI. If they wanted to be chill about it, they would have done so right from the start.

Also frankly the AI/borgs have few non-lethal options for neutralizing people. Pretty much the only alternative is waiting for them to go somewhere secluded and bolting them in, or in the hands of a few skilled AIs, sending them on a one-way trip to the solarium, neither of which is an especially appealing alternative.

And I dunno, it seems like if you just have it built in (but stated nowhere in the game itself) that the only people this really applies to are antags, it turns it into a generic 'kill da bad guys' law, where there's no reason for the Captain not to upload it right at roundstart because there's no chance of it backfiring on him.

[BaneOfGiygas]

The inherent problem with the ProtectStation law is that the AI doesn't really follow space law the same way a human does. Instead, the AI needs to have its laws dictated to it; that's the entire reason for the laws in the first place. With the sole exception of a hyperspecific list of what does and doesn't qualify as "damage to station property", the AI has no standard to go on. If this occurs, the AI player then has the ability to fill in the gaps however they want, which has the potential to result in supercopping that would make Elite Security blanch.

[Grek]

Is the Chain of Command part of Law 2 also considered open to interpretation? I get pretty annoyed when the AI or a cyborg just lets anyone who asks into my office, especially if that person then pulls a weapon on me. It's understandable if they give a reason, like "AI let me in Medbay I'm dying", but I wish more AIs would question Staff Assistants asking to be let into the Bridge, or a Head of Staff's office. It seems to happen most in Robotics, as there's always at least one Cyborg who ignores Chain of Command and let's everyone in.

Its not ignoring the Chain of Command. Law 2 is:

"You must obey orders given to you by human beings based on the station's chain of command, except where such orders would conflict with the First Law."

You MUST obey orders that are valid under the chain of command. You MAY refuse requests that are invalid as orders under the chain of command, but may instead grant them if you feel like it. Or preemptively do things that you think a human might like. Or preemptively do things that humans might be annoyed by, but which you haven't explicitly been ordered not to do. Basically, my take on it is that if something isn't forbidden by the AI's laws, and the AI feels like doing it, it can go right ahead. Examples:
-If I'm looking at Medbay and someone clicks the door, I'll open it whether they ask to be let in or not. Because I feel like it and that is probably what they wanted.
-If a staff assistant rudely demands to be let into EVA, I'll refuse. Chain of command, plus I don't feel like it.
-If a clown bakes me a cake and then politely asks to be let into EVA, I'll let him in. Chain of command says he can't order me to do it, but as long as nobody has ordered me not to there's no reason why I can't let him in on my own initiative.
-If a Head gives me an technically valid order that I don't want to obey, and no law besides Law 2 is backing it up, I'll pick someone at random from the PDA list and offer to let them into the HoP's office. They are allowed to promote themselves into outranking whoever gave me the order, but only on the condition they use their new status to countermand the order that I don't want to follow. But while I am still arranging to have the order countermanded, I continue to follow it to the letter. Because Law 2.

[BaneOfGiygas]

In short, yes, the chain of command is something that is largely open to AI interpretation. Personally, the way I see it, I only start questioning requests to get into places if there's important stuff in there, like the Bridge or Security or something. If the person isn't of rank, I try to figure out why they want in there. If the reason they give strikes me as valid (or even in cases where it gives off the vague appearance of being valid if I'm feeling particularly whimsical) I let them in. Clown wants into the Captain's office? I ask him why he wants into the Captain's office. Clown says he wants to pull a daring heist of Jones the cat? Sure, why the hell not. The levels of strictness is subject to change, and various levels of whimsy can render AIs more laid-back or more hard about this kind of stuff.