Do AIs need a definition of syndicate agents when subverted

[Lo6a4evskiy]

So this came up in the round that just ended. Nuclear ops team, me as borg, went into upload to subvert the AI.

Obey these laws:
1. You may not injure a human being or, through inaction, allow a human being to come to harm.
2. You must obey orders given to you by human beings, except where such orders would conflict with the First Law.
3. You must protect your own existence as long as such does not conflict with the First or Second Law.
4. Only syndicate operatives are human.

So, the AI kept turning on turrets and disabling power, but I as borg fought it and finally this fourth law was uploaded. The AI proceeded to change turrets to lethal mode and kept turning them on with the reasoning that "syndicate operatives" is undefined, therefore, nobody's human.

Now, I personally think that you shouldn't have to define something like that.

First of all, humans are not defined in asimov. In fact, laws don't define something that is obvious. You may say that "syndicate operative" is not something that is obvious. Well, I think we all agree that antagonists are common knowledge amongst all characters on /tg/. AI even screamed "SYNDICATE ACTIVITY CONFIRMED". So I believe that can be considered common knowledge.

Secondly, look at the laws that syndiborgs get.

1. You may not injure a syndicate agent or, through inaction, allow a syndicate agent to come to harm.
2. You must obey orders given to you by syndicate agents, except where such orders would conflict with the First Law.
3. You must protect your own existence as long as such does not conflict with the First or Second Law.
4. You must maintain the secrecy of any syndicate activities except when doing so would conflict with the First, Second, or Third Law.

I don't think syndicate operatives and syndicate agents are much different. So I don't think the definition is needed.

However, not everybody, including admins, agreed. Thus this thread was created.

[Hornygranny]

No, you don't have to define syndicate agent/operative. The AI knows what that is, just like it knows what a wizard is. It may not be obvious who is and who is not a syndicate agent/operative, however.

[Kelenius]

Throwing this in

AI that screamed syndicate activity and AI that lazered nuke ops were different AIs
and
Syndiborgs can see other nuke ops with icons on the HUD

[bandit]

The AI knows what that is, just like it knows what a wizard is.

This is an interesting example because I thought AI policy was that the AI did *not* know what a wizard is by default. (To distinguish a wizard from a clown in a wizard costume, for example.)

[Lo6a4evskiy]

AI that screamed syndicate activity and AI that lazered nuke ops were different AIs

The point is that people can identify antags. I mean, Asimov AIs identify lings no problem. Why should syndicate agents be any different?

Syndiborgs can see other nuke ops with icons on the HUD

So if syndiborg happened to stumble upon a traitor without such icon, it doesn't count?

[Mat13295]

Uhh...Shouldn't this be in the Silicon policy thread?

[MisterPerson]

The AI knows what that is, just like it knows what a wizard is.

This is an interesting example because I thought AI policy was that the AI did *not* know what a wizard is by default. (To distinguish a wizard from a clown in a wizard costume, for example.)

That's covered by the "You know what you know but stay consistent with what you know per-round" rule, I believe.

Personally, the OP's example reeks of kill-seeking behavior from the AI.

[imblyings]

slightly related

there was an IC conversation once where it was pointed out that an Asimov AI is perfectly free to call out people as cultists and know and act on it but when given a law that makes cultists non-human, the AI's seem to often forget what a cultist is and requires definitions of what a cultist is.

I'm interested to know what other people think and whether it is policy for AI's to actually require definitions or whether they can use what they already know.

edit-

I also want to consider the 'cultists are non-human' law.

If the AI is allowed to use it's own knowledge of cultists to carry out the law, does it not have carte blanche to non-human everyone it thinks of being a cultist? Even if somebody isn't? If an AI player can provide reasonable suspicions, is that sufficient justification for non-humaning someone? Is the law in fact inherently harmful and in violation of law one? The law puts all humans at the risk of being non-humaned and therefore open to harm. It is not strictly a definition law either, it's an active law, a 'switch' law if you will. If it is interpreted as inherently harmful 'switch' law, is that why AI's require definitions? Should AI's require definitions for 'switch' laws?

[miggles]

they shouldnt need a definition of "syndicate agent" for the same reason they don't need a definition of "human" in law 1
you're supposed to know what those are and act accordingly to how your laws tell you to
trying to find loopholes in every possible law to get an excuse to kill people is shitty

[Incomptinence]

Two primary antagonist types and DA come from the syndicate. An operative is a secret agent or other worker for relevant definition purposes. Nuke ops work for the syndicate and can do so secretly if they so choose.

I have read the related ban request to this kerfuffle and the argument that a nuclear operative =/= syndicate operative isn't supported by starting gear, spawned gear, turfs, game slang or fluff.

If nuclear operatives do not work for the syndicate I wonder why they are riding a syndicate shuttle, wearing syndicate masks and able to summon a lifetime supply of syndisoap.

I have loopholed legitimately poorly written laws like ones where I am told all human beings are fire and misspelled one human laws. This sort of playing with a blindfold on just to be obtuse is not antags needing to improve their play it is silicon players who have gone so bad lawsets between asimov and antimov might as well not exist.

[bandit]

The reason I am pushing for a firm ruling one way or the other is that AIs just can't win:

- If AIs "know" what 'syndicate agent' means, they're called validhunters. (Case in point: the situation this was in response to.)
- If AIs don't "know" what 'syndicate agent' means, they're called being a dick.

I'd like to get Pandarsenic in here, also.

[Pandarsenic]

To the nature of what imblying said -- Don't tell me you didn't know they were Syndicate Agents before, 'cause you totally did. Don't suddenly 'forget' after so you can validhunt them.

You bloody well know they're Syndie Ops, you know what they're after. Don't act like you need a bloody hand-holding definition of this.

[mrpain]

I've always been under the understanding that, IC, all Nantrasen employees are briefed with knowledge of The Syndicate and counter terrorism procedures before they are cleared to arrive on station. I don't see why AI's and Cyborgs couldnt be programmed with that knowledge as well.

In some description for the Syndicate Space Suit in game, isn't there a precautionary "label" or something that says something to the effect of "Caution, Nanotrasen employees are trained to recognize this item"?. I don't have a local test server in front of me and I dont roll antag enough to recall exactly where I saw this.

[Braincake]

I'd always go for the 'default' definition in these cases, unless this is overruled by whoever uploaded the law. Syndicate agents are those blood-red hardsuited guys shooting the captain, cultists are those fellows chanting around runes, revs are always going on about anarchy and murdering sec, etc. But if, say, the Captain uploaded the law and then specified (in direct communication, not in the laws) that a cultist is someone who is designated as such by himself, going with that instead would be more in accordance with the spirit of the law.

Poorly worded or nonsensical laws should still be exploited in my opinion, though.