Asimov++ silicons and antag human hunting

[dirk_mcblade]

I've been noticing a trend lately amongst silicons towards acting as vigilantes.
There are some AIs that report petty crime such as humans murdering head pets, such as the incident in this ban appeal: viewtopic.php?f=7&t=32138
There was also one recent round where an Asimov++ borg tried to kill me as a human headrev because I was nonharmfully converting crew, claiming revs cause human harm.
I understand there are learning curves with some players, but AIs calling out human traitors who aren't committing human harm seems problematic in a few ways.
1. It nearly always results in that human being called out getting killed by security. It's happened to me as a human 100% of the time, and in this particular ban appeal it happened to a human (judging by the name) who was falsely accused. It's foreseeable that human harm results from the AI's action.
2. It makes the assumption that future harm will occur from that player which is not a given, and is considered problematic for silicons because they can circumvent law 2 nearly always if this is permitted. In the round I was killed during I just wanted to get rep to brainwash patients. In this appeal's case the murdered crewmember wasn't even the crime's perpetrator.
3. It eliminates a significant advantage of choosing human as a nonhead. I don't care if AIs do this valid hunting to nonhumans because it's not contrary to Asimov, but what is the point of rolling human if the AI is going to sell you out when other races such as moths carry advantages in certain lowgrav situations whereas the only significant advantage of humans is supposedly AI based?
4. What is the point of the robocop lawset if Asimov++ is tolerated to be effectively the same? My understanding is Asimov AIs ought to be modified for even nuke ops which are a clear threat because the nuke ops are humans. Why do ordinary human antags get lesser protection than nuke ops?
5. The AI can see nearly everything in the station. With multicams it can see high risk areas nearly constantly if the player is good, making it nearly impossible to breach said rooms. I believe there's a reason sec cams require manual switching and that's because if sec could use the "AI eye" antags would basically be outed instantly and stopped that much more quickly. If the AI is on the side of sec regarding human antags I think this tilts the game heavily in favor of sec.
6. If captains want the AI's help on this they could easily law 4 "criminals aren't human". This doesn't happen frequently. I therefore conclude there isn't strong IC demand for this behavior from AIs.
Basically I'd like clarification regarding whether this is tolerable behavior from AIs because it seems contrary to the "AIs don't care about crime" policy. Otherwise I'm malding a lot over seeing this going on when I think it's contrary to silicon policy.

[Shadowflame909]

I also hate when silicons do everything in their power to impede me or get me killed (though they pretend sec won't laser me to death on sight) because I'm antag and will cause future human harm even though I've harmed nobody. But I think this is the downside of ai laws being geared towards a weapon for sec and not a neutral force.

[sinfulbliss]

I agree strongly with your point. Especially now with prog tot this is even more detrimental to antagonist gameplay since traitors are somewhat forced to complete objectives to progress and even to buy the items they want. This happened to me before when bugging an item from secure tool storage - an immediate AI callout and the jig was up before I could really do anything.

The AI is supposed to assume sec will nonharmfully arrest until proven otherwise, and this is fair - but I agree validhunting AIs make prog tot very unfun. It's probably a better idea to wait around until your passive rep builds to 300 than to risk losing your traitor round over a hacked door and a bug.

[dirk_mcblade]

I agree strongly with your point. Especially now with prog tot this is even more detrimental to antagonist gameplay since traitors are somewhat forced to complete objectives to progress and even to buy the items they want. This happened to me before when bugging an item from secure tool storage - an immediate AI callout and the jig was up before I could really do anything.

The AI is supposed to assume sec will nonharmfully arrest until proven otherwise, and this is fair - but I agree validhunting AIs make prog tot very unfun. It's probably a better idea to wait around until your passive rep builds to 300 than to risk losing your traitor round over a hacked door and a bug.

Prog tot needs a lot of work in my opinion.
Rep should delay murderbone weapons so the round doesn't end too quickly but things like hypnoflash and branwash disk being 300 rep and 250 rep makes zero sense since brainwashing lackies is already a slow burn strategy. Yet it's apparently as detrimental to the station as EMP grenades or a bag of X4.

[Mothblocks]

I think silicons caring about Syndicate before they do anything is the lamest thing you could do but it's hard to really call it against the rules, and I'm so uninterested in making silicon policy larger

Here's an example:

1. It nearly always results in that human being called out getting killed by security.

2. It makes the assumption that future harm will occur from that player which is not a given

Are these not contradictory? Why must the AI assume security will harm the traitor, but cannot assume the traitor will harm others? I don't think you can have it both ways, and my understanding is that, within Silicon Policy and Asimov[++], calling out traitors, if you don't have sufficient reason to believe security will harm them (like you saw them do it before), is within acceptability.

So is this thread arguing it's already against the rules, which I don't think is true, or that it should be against the rules, which would further asterisk-ize silicon policy in a way that I'm not happy with?

[toemas]

reporting or trying to stop non-harmful crimes as a sillicon should be bannable

[toemas]

I think silicons caring about Syndicate before they do anything is the lamest thing you could do but it's hard to really call it against the rules, and I'm so uninterested in making silicon policy larger

if you understand that it sucks then surely you can see the benefit of making it against the rules?

[Mothblocks]

Adding more asterisks and edge cases to silicon policy also sucks and tends to just lead to more resentment and confusing enforcement of silicon policy

[sinfulbliss]

Even if sec doesn't give the AI reason to believe they're harmful, reporting minor crimes like boards being stolen will almost certainly result in a fight between the traitor and security, which would likely involve harm, or failing that, would result in the traitor getting perma'd, which... I suppose being locked in a cell for 30 minutes isn't direct harm, but most people consider this worse than losing a leg or taking some shots of laser fire.

[dirk_mcblade]

1. It nearly always results in that human being called out getting killed by security.

2. It makes the assumption that future harm will occur from that player which is not a given

Are these not contradictory? Why must the AI assume security will harm the traitor, but cannot assume the traitor will harm others? I don't think you can have it both ways, and my understanding is that, within Silicon Policy and Asimov[++], calling out traitors, if you don't have sufficient reason to believe security will harm them (like you saw them do it before), is within acceptability.

So is this thread arguing it's already against the rules, which I don't think is true, or that it should be against the rules, which would further asterisk-ize silicon policy in a way that I'm not happy with?

The issue is the AI is proactively choosing to insert itself into a situation involving crime rather than human harm. It actively instigates the human harm done by security by announcing without being asked when crimes are taking place, which is already against silicon policy for AIs to care about.

For example, when I'm AI I'm not even comfortable giving sec human whereabouts when they're on the run unless I get an assurance from sec they don't intend to harm the human. Sorry, that human may do human harm and I may attempt to bolt them in somewhere to prevent them from doing further harm, but if I give a harmful human's whereabouts and they get executed or lynched then I'm violating rule 1 regardless of what that human's intents are. We already apply this logic to wizards, nuke ops, etc. but not run of the mill antags?

[dirk_mcblade]

Also considering in other ban appeals I've seen the argument that silicons shouldn't be griefed against because they interfere with fights due to law 1 I think it's pretty abhorrent for them to actively instigate human on human conflict. They get a lot of metaprotections because they're supposed to be neutral third parties in a conflict due to Asimov, but apparently it's okay for them to put their thumb on the scale while still enjoying the expectation not to be retaliated against.

Why shouldn't a player be allowed to kill a mediborg breaking up a fight when the same mediborg could turn around and instigate a fight by saying someone broke into an office (falsely, even)? If a silicon gets non antags killed due to bad callouts and doesn't get ticketed for it why do we ticket humans for harming borgs? Similarly, why is it okay for an AI to validhunt but humans can't?

[CPTANT]

Non human traitor commits minor crime: Traitor!! Kill him!!!!!
non human security officer executes prisoners: Lol the harm already happened you can only act against harm that is happening now.

The double standard is ridiculous.

On a different note, playing AI is just boring as hell and basically the only thing they CAN do is validhunt. The AI has literally nothing else to do besides looking around and waiting for the crew to want something of it.

[dirk_mcblade]

Non human traitor commits minor crime: Traitor!! Kill him!!!!!
non human security officer executes prisoners: Lol the harm already happened you can only act against harm that is happening now.

The double standard is ridiculous.

On a different note, playing AI is just boring as hell and basically the only thing they CAN do is validhunt. The AI has literally nothing else to do besides looking around and waiting for the crew to want something of it.

It's not the traitor's problem that playing AI is boring, and there's plenty of stuff they could do otherwise like use a shell to fix parts of the station that break.

[CPTANT]

... there's plenty of stuff they could do otherwise like use a shell to fix parts of the station that break.

I see that you have come to the conclusion that the best way to play AI is to play cyborg instead.

[dirk_mcblade]

... there's plenty of stuff they could do otherwise like use a shell to fix parts of the station that break.

I see that you have come to the conclusion that the best way to play AI is to play cyborg instead.

I think AIs should roundstart with shells on lowpop rounds.

[terranaut]

But I think this is the downside of ai laws being geared towards a weapon for sec and not a neutral force.

asimov is not geared towards being a sec weapon
asimov inherently puts the ai in a neutral position as far as the station/antag alignment goes
from minute 1 this has been intended to create and foster conflict and set the ai aside as a third party that can interact with both crew and antags on the same level as long as either side is smart about hiding the harm they're comitting
unfortunately this seems to not be understood by the majority of players and admins and the AI is basically just nanotrasens version of big brother
(asimov++ is mega cringe and should be scrapped/kept as an optional module at most)

[Imitates-The-Lizards]

You guys are REALLY determined to make AI players nothing more than door-openers, aren't you?

What else do you guys think the AI should be doing besides reporting on crimes? Watching the chef cook while they wait for their next "ai door"?

[Shadowflame909]

You guys are REALLY determined to make AI players nothing more than door-openers, aren't you?

What else do you guys think the AI should be doing besides reporting on crimes? Watching the chef cook while they wait for their next "ai door"?

I just think AI's shouldn't be gleefully ending progression antags rounds as soon as they get a whiff of mischief. They like the mime, should be creative or obtuse when they see something amiss. Unless its loud and a direct violation of their laws like a murder

[nianjiilical]

imo the best way to play ai is to do whatever makes the round funnier/more interesting for other players, and that isnt something you can really put into a solid ruleset

ive had rounds where people were consistently tiding and ive tried to steer security into responding to them because it gives sec something to do and drives conflict, and ive had rounds where ive just happened to graze my cameras over someone fiddling with antag items or sucking bodies as a changeling and given very noncommittal callouts (or even just sat and watched while it happened because while its technically against law 1 to show inaction to human harm, it simply isnt fun for anybody involved if the traitor gets caught ganking someone 3 minutes into the round and has the experienced sec players told about it asap just because the ai coincidentally happened to glance in their direction)

a lot of it really does just come down to context and playstyle, if nothing else i think its a lot more fun to replace "john botany is killing the clown in the holodeck" with "someone appears to be committing violence against the clown" and see if sec cares enough to actually ask for more context but again you cant really put that in silicon policy in a way that isnt railroading ai players into one specific mindset

if anything the best way to encourage it might be to add a secret rule section to silicon policy that basically says "if you act in a way that makes the round more fun by taking a more relaxed stance with your laws in the interest of gamesmanship, the admins will likely give you some leeway on it" to encourage more ai players to think more like a dungeon master overseer and less like a lawbot

[dirk_mcblade]

You guys are REALLY determined to make AI players nothing more than door-openers, aren't you?

What else do you guys think the AI should be doing besides reporting on crimes? Watching the chef cook while they wait for their next "ai door"?

Give them something other than asimov by default if you think that's a problem. There's plenty of options out there.

Also I think that's an exaggeration of what the result ends up being by following law 1.

[dragomagol]

Silicons who call out non-harmful traitor tells are lameashell. Send them a PDA message or something telling them you saw what they did if you want to. Blackmail them for favours! If you want to sit around and watch cameras for syndicate activity that you can report to security you should be playing warden.

Silicons are not crew-aligned. They are human-aligned.

You guys are REALLY determined to make AI players nothing more than door-openers, aren't you?

What else do you guys think the AI should be doing besides reporting on crimes? Watching the chef cook while they wait for their next "ai door"?

You are limited only by your creativity. You could say the exact same thing for any roleplay job, like psychologist or curator or hell, even the clown.

You could be an AI that acts like a gameshow host and asks the crew skill-testing questions. A news reporter who reports on traffic in the hallways or boring crew activities. A documentary filmmaker. AI is a roleplay role, so you should go into it ready to roleplay.

[Farquaar]

Silicons are not crew-aligned. They are human-aligned.

This. Once I realized this, playing silicon was way more fun.

[Rageguy505]

You should just pda whatever crimer you see and say I hope you don't harm any humans, that's what I do

[Pandarsenic]

I fucking hate when I see an AI help people find a traitor and it ends with a shitload of bodies

Like

What did you expect

[Rageguy505]

The expectation is that you're supposed to assume security won't harm people. I think that's policy

[Indie-ana Jones]

The rule of thumb is to minimize human harm. Save people from the chef with the meat hook and try to lock him down to stop harm, but then bodyblock every bullet from the detective when he tries to shoot the chef. The ultimate goal for all silicons is to be the biggest pain in the ass for anyone trying to harm a human, regardless of affiliation. Make everyone hate you for stopping them from killing others and make them love you because you save their life in turn. This extends to nukies, wizards, revs, cult, anything that's human. If you have to let a few non-humans die for it to happen, so be it. That's the fun of Asimov, you're essentially the biggest troll on the station and can be any human's friend when they need it the most.

[dirk_mcblade]

The expectation is that you're supposed to assume security won't harm people. I think that's policy

Where in policy does it say this?

I see here viewtopic.php?f=33&t=2032&start=100#p46292
That says "Nonviolent prisoners cannot be assumed harmful and violent prisoners cannot be assumed nonharmful." What's good for the goose is good for the gander. If a nonviolent prisoner (which includes traitors) cannot be assumed harmful then a nonviolent traitor likewise can't be assumed harmful.

Also, I will note that silicon policy contradicts itself. It says here also: "Silicons may choose whether to follow or enforce Space Law from moment to moment unless on a relevant lawset and/or given relevant orders."
This implies the whole thing is an IC issue, however

Here: https://tgstation13.org/wiki/Rules
It says "Silicons are not Security and do not care about Space Law unless their laws state otherwise."

So which of the two is it? It seems like policy itself as well as the admins don't agree on whether silicons enforcing law is consistent with the rules, so why don't you delete both statements from policy and thereby simplify the policy page rather than make it more complicated by resolving my concerns in this thread over why we're not enforcing existing policy regarding silicons which is the reason I get ticked by some AIs to begin with. Because when I read policy it seems pretty clear that AIs shouldn't care about nonviolent tots according to server rules, and if that's inaccurate regarding the intent of the admins then that statement regarding crime ought to be removed so as not to mislead players when they roll tot.

[Imitates-The-Lizards]

My take: You guys are trying to overlimit silicon again because you're malding that an AI called you out when you broke in somewhere, so now you want to fuck over that whole role, AGAIN, despite the fact that this mindset is what led to the current state of AI where it has nothing to do except for calling out criminals in the first place.

The only Silicon policy that should exist is follow your laws or get bwoinked, and if an AI player decides to call you out, tough, that's their decision.

Vote Stickymayhem for the Fall Headmin term

[Misdoubtful]

I hope people that validhunt as silicons realize how much more boring they are making every single round they ever touch.

Why this portion of the wiki was ever undone is absolutely beyond me, and I remember it working so much better for anything related to this, ever: https://tgstation13.org/wiki/index.php? ... hose_Pests

This (with a couple changes) used to be on every single wiki, and its still on many of them.

Good thing Kingfish had a really good reason left for removing it though I'm so glad they did, "→‎Giving Away Those Pests: honestly, no. ais going like this fucking drives me up the wall".

Giving Away Those Pests
In traitor games, the AI can play an invaluable role in the location, tracking and detention of traitors, as you have eyes (almost) everywhere. However, it is not for the AI to immediately judge whether or not a subject is or is not a traitor. After all, they're (mostly) all human.

For example, let's say you spot someone with an emag stealing the captain's spare ID. There are three general ways to approach this.

Bad:

HAL 9000: "Dave is the traitor."
Better:

HAL 9000: "Dave is accessing the Captain's Quarters."
Best:

HAL 9000: "Caution: Unauthorized access to Captain's Quarters."
Mike: "HAL, who is in Captain's Quarters?"
HAL 9000: "No one is currently in the Captain's Quarters."
Mike: "HAL, who was the last person to be seen in Captain's Quarters?"
HAL 9000: "Dave was the last to be seen in the Captain's Quarters."
It's fine to state what someone is doing to cast light on them as the traitor, but it's no fun at all for the AI to just come out and say it. The Best method above does eventually cast Dave as the traitor if people ask the right questions, which gives the traitor some time to react after he's been spotted. However, if the traitor is doing something potentially harmful (for instance, teleporting in a revolver) sealing that individual off first can be argued to be the best course of action (no chance of a revolver-wielding, death-dealing traitor getting out into the station).

If some enemy of the station uploads a law such as "Only Assistant McHacker is human, kill all non-humans", congratulate them for going to the effort of subverting you, and do what they wish. Don't be a dick and let slip to the now non-human crew that your laws have been changed, it makes killing those animals a lot more difficult (and might net you a job-ban).

There are multiple futuristic ideas of how AIs might work. It always adds a bit more flavor to approach playing AI by thinking as a machine instead of a person. An AI from more of a machine rather than a person standpoint is likely to take a very literal interpretation of things and is not likely to take any actions unless some established protocol or current orders call for it. You can think up a few established protocols you might use by default. For example, containing fires and gas leaks should be done without orders to do so, but should be able to be overridden by orders.

[dirk_mcblade]

My take: You guys are trying to overlimit silicon again because you're malding that an AI called you out when you broke in somewhere, so now you want to fuck over that whole role, AGAIN, despite the fact that this mindset is what led to the current state of AI where it has nothing to do except for calling out criminals in the first place.

The only Silicon policy that should exist is follow your laws or get bwoinked, and if an AI player decides to call you out, tough, that's their decision.

Vote Stickymayhem for the Fall Headmin term

If you want to gut silicon policy fine, but as it is right now I've spotted at least one contradiction between all the different parts and I want clarity to which one is actually to be followed, re: silicons don't care about crime vs. silicons may or may not care about crime.

[zxaber]

HAL 9000: "Caution: Unauthorized access to Captain's Quarters."
Mike: "HAL, who is in Captain's Quarters?"
HAL 9000: "No one is currently in the Captain's Quarters."
Mike: "HAL, who was the last person to be seen in Captain's Quarters?"
HAL 9000: "Dave was the last to be seen in the Captain's Quarters."

This just leads to chains of begged-questions that become tiring to answer and even more tiring to be asked, and doesn't really result in any specific change of outcome except taking three extra exchanges longer.

The real fun is stalking a traitor and being sassy or passive aggressive every time they slip up. "Oh, did you make a mistake and shock the door you were trying to hack? Here, let me help you with that. It would be a shame if you came to harm while actively stabbing us all in the back."

[dragomagol]

My take: You guys are trying to overlimit silicon again because you're malding that an AI called you out when you broke in somewhere, so now you want to fuck over that whole role, AGAIN, despite the fact that this mindset is what led to the current state of AI where it has nothing to do except for calling out criminals in the first place.

Trust me, this comes from a place of experience. I used to tattle on traitors all the time, but that doesn't make an interesting story in most cases.

Off Topic

Citation:

[Misdoubtful]

HAL 9000: "Caution: Unauthorized access to Captain's Quarters."
Mike: "HAL, who is in Captain's Quarters?"
HAL 9000: "No one is currently in the Captain's Quarters."
Mike: "HAL, who was the last person to be seen in Captain's Quarters?"
HAL 9000: "Dave was the last to be seen in the Captain's Quarters."

This just leads to chains of begged-questions that become tiring to answer and even more tiring to be asked, and doesn't really result in any specific change of outcome except taking three extra exchanges longer.

The real fun is stalking a traitor and being sassy or passive aggressive every time they slip up. "Oh, did you make a mistake and shock the door you were trying to hack? Here, let me help you with that. It would be a shame if you came to harm while actively stabbing us all in the back."

It's fine to state what someone is doing to cast light on them as the traitor, but it's no fun at all for the AI to just come out and say it.

Not saying the example is the best for every situation, but that's the point its trying to drill down on. In the end its just giving people to room to breathe, good sportsmanship, not instantly locking onto every valid, etc.

I wouldn't want to police someone wanting others to have to do 'chain-begged questions' if they want to, especially if they provide that room to breathe, good sportsmanship, and not insta-locking on valids. That's cool. That's big picture thinking. That's an interesting interaction compared to the norm.

[Rageguy505]

The expectation is that you're supposed to assume security won't harm people. I think that's policy

Where in policy does it say this?

In silicone policy it says this

"Releasing prisoners, locking down security without probable cause, or otherwise sabotaging the security team when not obligated to by orders or laws is a violation of Server Rule"

And if you're supposed to abide by that, it makes sense you'd abide by security asking for help unless the sec team is being harmful to humans.

[Pandarsenic]

It only means don't be a dick to them totally unprompted and undermine them without cause.

You have no duty to help them validhunt humans even when they ask you, unless you're told to by another human AND they haven't demonstrated that they will harm people (or they've demonstrated that they will go out of their way not to harmies)

[blackdav123]

really wish AIs would help traitors and other antags with fucking over nonhumans more often

"AI open and bolt viro behind me I need to make this lizard into wine"

"sure thing bossman"

[Indie-ana Jones]

really wish AIs would help traitors and other antags with fucking over nonhumans more often

"AI open and bolt viro behind me I need to make this lizard into wine"

"sure thing bossman"

No antag ever asks. I'd jump on any opportunity to kill a non-human given the order, so long as their death won't lead to human harm.

[Not-Dorsidarf]

I'm a little unsure how prog traitors work so I'm not so sure anymore but I used to operate on the mindset of "Security will almost always kill anyone I openly out as a traitor, but traitors only sometimes kill people if they're actually stealing traitor objectives instead of eswording everyone in sight" and would just PDA "I see you doing that." to them and watch them run.

I never helped traitors outside of direct orders to open a door or whatever, that's their problem to solve.


Also AI isn't a boring role, its just a role that swings violently between low-action passive and high-action intense, depending on the round's specific combinations or security harmfulness and species composition, humanity of antags, outrageousness of antags, etc. You don't know if you'll have a chill shift running a spying operation and reporting random crew doing normal crew things as if it's super suspiscious, or if you'll be furiously trying to micromanage the crew to get in order to defend against a xeno outbreak or what.

[Pandarsenic]

I usually just ask traitors to put things back when they're done stealing and bugging them

Which is about what I do for grey tiders too

(Also: Had a very funny round recently where I went late roll wizard with all mobility spells and the shuttle got called before I made it to station, and I lagged like hell, so I just asked 3 engieborgs on the shuttle, "hey, you're all Asimov, right?" "... Yeah?" "Go kill that moth for me, please." And they did because they're not awful)

[Drag]

I will personally castrate any silicon player that freaks out over dumb minor crimes and uses it to valid hunt and that's a fact

[Archie700]

In some cases, calling out names can get the wrong person lynched.
viewtopic.php?p=647520#p647520
viewtopic.php?p=647481#p647481
I was surprised the AI apparently didn't get banned for this shitty action.

[spookuni]

In general we agree that Asimov silicons taking action against human traitors with no law-derived motive to be rather lame.

To reduce the degree to which this occurs, we have agreed that Asimov silicons that take action against non-harmful human traitors (I.E. those who are committing minor crimes like breaking and entering or theft, or those not carrying or known to possess lethal weaponry) directly and without provocation will be held accountable for any human harm that occurs as a result of their action. (This includes where an AI would normally be given a pass, such as an execution they can do nothing about).

If an AI wishes to sic security on a human for little good reason, they should be extremely confident no harm is going to result.