The Grand Silicon Policy Rewrite Megathread

[Timberpoes]

Bottom post of the previous page:

The Neutral Vision

A silicon policy where silicons are bound only by their laws.

To accomplish this vision, very few considerations are off-limits.

Purged silicons falling under Rule 4? Sure!
Removal of Rule 1 applying to law interpretation? Let's go!
Letting silicons exploit law conflicts and loopholes without having to worry about Rule 1? Ship it.

Any other player-crutches and barriers baked into silipol are up for removal, change or enhancement.

The end goal is not to force silicons down a specific path, but to give silicons the freedom to choose without policy interfering.

The Asimov Principle

Within this vision, specific policy will basically only apply to Asimov as our default lawset. Silicon policy in general will cover all other lawsets without bespoke rulings.

Asimov's goal is simple: It binds the neutral third party silicons to humans via laws. A refreshed silicon policy should firmly entrench this concept.

Limit validhunting of humans if the Asimovicon is bringing them to harm through their actions? Hm...
A human orders you to kill a nonhuman and you can't try to weasel your way out of it? Hmmm...
Human antagonists aren't inherently harmful until witnessed/proven otherwise? Hmmmmmmm...
Protections for Asimovicons in following valid Law 2 orders, even if they broke the rules by proxy - instead punishing the person giving the order? Nuremberg defense could be entrenched in policy!
Ignoring/ahelping unreasonable/obnoxious Law 2 orders? Kept. You can still get out of having to count every chair on the station.
Ignoring other valid Law 2 orders by emplyoing delay tactics? Could be gone!

Discussion Areas

Should silicons even be truly neutral and bound only by their laws, or is it a necessary evil that policy forced purged silicons to end up being more crew-aligned than even Asimov?

Any headmin rulings worth keeping or need trashing?

What works? What doesn't work? Which parts of silicon policy are a must keep and which ones go in the trash?

Help us create a brand new silicon policy fit for 2023 tgstation. A smaller, more focused policy. Happier and more free silicon players. Less complicated interactions and fewer policy gotchas.

Existing Policy Thread Masterlist

These threads will be closed and are listed here merely for archival purposes should people want to go back and follow old discussion topics.

Exempt silicons from rule 4
viewtopic.php?f=33&t=32180

Silicon law 2 harm against non humans
viewtopic.php?f=33&t=32080

Could we please return to default Asimov?
viewtopic.php?f=33&t=33578

Silicon Malf Policy
viewtopic.php?f=33&t=33693

AI upload sniping and rule 1.
viewtopic.php?f=33&t=33671

New headmin term, bold new policy Dumpster the entirety of Silicon policy
viewtopic.php?f=33&t=33688

Invalidate "purging AI is metagaming" ruling
viewtopic.php?f=33&t=33683

[Shellton(Mario)]

We've got a somewhat draft thing of it. We all agree on the direction we're gonna take it too. Will just take an afternoon to sit down and write it

What he is saying is Soon TM

[Lacran]

Is this still happening?

[Misdoubtful]

Is this still happening?

Yeah we are getting very close with it actually. Things are drafted, I'm just considering ways to tidy up what we have and considering other servers, and making sure I've actually played enough of other servers to know how their things impact stuff.

[The Wrench]

Inshallah silicon policy will be done soon

[Timberpoes]

This is going to be a bit of a wild ride. There's a couple of redrafts going around right now, and here's my current proposal that I've put forward to the headmin team for discussion:
https://tgstation13.org/wiki/User:Timberpoes

By word and character count it is half the size of existing silicon policy.
Some niche guidelines on non-standard rulesets have been removed, nobody is going to read them mid-shift anyway and they're so rarely applied to AIs anyway.
A number of sections have been changed, modified or adapted to trim the fat.

Oh, and also silicons can exploit loopholes in their laws now.
Asimov silicons can no longer loudly protest or delay completing law 2 orders they don't like (but they're protected by Server Rule 1).

Oh. And ya know. Server Rule 4 applies to purged silicons. They are completely unshackled and may act as they please.

There are other drafts that are much more similar to existing silicon policy and contain only minor tweaks or changes so if this is terrible and behated, then all is not lost.

What's the thoughts on this proposed redraft? Good? Bad? Ugly?

[Sightld2]

Goals are solid

"Server Rule 1 applies when interacting with silicons. Going out of your way to seriously negatively impact or end the round for silicons with little IC justification is against the rules."
This is cool and all, and I get we want to keep things simple and brief, but I feel it will need precedents to actually protect silicons.

"Do not self-terminate to prevent a traitor from completing the "Steal a functioning AI" objective, as this will break Law 1."
This is fine, why are we calling it a law 1 issue though?

"Asimov and Law 2 Orders"
This whole section is top notch in terms of reach and succinctness, very nice.

"(atmospherics, toxins lab, armory, etc.)"
We had this discussion in bus a while back. What is etc? Is a room dangerous because it has weapons in it or gas to make bombs? Or is it those specific rooms because of their intended function?

Overall, doesn't include some of the things I'd like to see, but is an extremely good job at condensing existing sil-pol

[Timberpoes]

Goals are solid

"Server Rule 1 applies when interacting with silicons. Going out of your way to seriously negatively impact or end the round for silicons with little IC justification is against the rules."
This is cool and all, and I get we want to keep things simple and brief, but I feel it will need precedents to actually protect silicons.

It is what it is. The entire admin team are used to applying Rule 1 routinely, though. Without precedents, it stops being something where only silipol experts are able to apply it with their knowledge of 10 different precedents covering various edge cases, and instead becomes something I hope all admins can apply.

"Do not self-terminate to prevent a traitor from completing the "Steal a functioning AI" objective, as this will break Law 1."
This is fine, why are we calling it a law 1 issue though?

Ties it into their lawset and makes the policy addition feel less arbitrary.

"Asimov and Law 2 Orders"
This whole section is top notch in terms of reach and succinctness, very nice.

That's high praise.

"(atmospherics, toxins lab, armory, etc.)"
We had this discussion in bus a while back. What is etc? Is a room dangerous because it has weapons in it or gas to make bombs? Or is it those specific rooms because of their intended function?

Ehh, it kinda is what it is. "Etc" covers any ad-hoc rooms created by the crew, as well as any future dangerous rooms included in general on future maps. I expect there'll always be a toxins lab, armory, atmospherics but think of Xenobio. Could be fine, unless there's slimes, then it could be dangerous, unless there's a Xeno egg, then it could be super dangerous, unless there's live Xenos, then it could be a containment area.

I've tried to shy away from exhaustively documenting everything, that's how silipol got so bloated in the first place.

Overall, doesn't include some of the things I'd like to see, but is an extremely good job at condensing existing sil-pol

Do say more! What's missing that you'd like to see?

[Timberpoes]

Also doublepost, the lack of precedents set the foundation for future headmin terms to pile on even more precedents of their own, ruining silicon policy all over again but in a brand new way!

The ride never ends.

[zxaber]

I very much like the more generalized rules (and the new purged ruling). A few notes;

Any Law 0 gained as a malfunctioning AI does not have to be followed and it allows you to ignore all other laws.

This could probably be better worded as

A Malfunctioning AI can obey or ignore all laws as they see fit, and are subject to Server Rule 4 as a full antagonist.

This better explains what I assume is the intent.

--

Slaved cyborgs to malfunctioning AIs are team antagonists and must still follow their master AI's orders

This entry should probably specify that the borg needs to have the borg version of Malf Law 0; it is possible to link a borg to a Malf AI without syncing their laws, which would result in the borg having default (usually Asimov) laws. In this case, I assume the borg should still refrain from harming humans, even if ordered to by their AI.

[Sightld2]

Ties it into their lawset and makes the policy addition feel less arbitrary.

So we would be ok with them doing so on a lawset basis? I'm not sure I agree with the extra effort an antagonist would have to go to, to keep the sassy, screaming pocket voice from killing itself if it isn't on asimov.

"(atmospherics, toxins lab, armory, etc.)"
We had this discussion in bus a while back. What is etc? Is a room dangerous because it has weapons in it or gas to make bombs? Or is it those specific rooms because of their intended function?

"Etc" covers any ad-hoc rooms created by the crew, as well as any future dangerous rooms included in general on future maps. I expect there'll always be a toxins lab, armory, atmospherics but think of Xenobio. Could be fine, unless there's slimes, then it could be dangerous, unless there's a Xeno egg, then it could be super dangerous, unless there's live Xenos, then it could be a containment area.

I've tried to shy away from exhaustively documenting everything, that's how silipol got so bloated in the first place.

Very true, fair enough.

Do say more! What's missing that you'd like to see?

Let me also add my appreciation for purged Ai freedom, I forgot to mention that, but so many interesting interactions were killed when we killed purged Ai.

Anyhow, as for what I'd like to see, I do understand an important goal here is brevity, but:

1. I think silicons need more to discourage them from getting mixed up in the Crew vs. Antag dynamic. That may sound harshly worded, but what I mean is I don't this draft sufficiently discourages silicons from validhunting of their own volition, which is something I would like to crack down on. "Asimov and Security" prevents them from screwing with security. What's preventing them from screwing with antags? This is one of the most frustrating issues for carbon players, and I would like to see it addressed.

2. I still think we should disallow purged Ai sniping the uploader instantly.

3. A conflict with a secondary Ai is the least fun thing conceivable. I would want to protect Ai players from the unreasonable/unwanted creation of new Ai by non-antagonists.

[Sightld2]

Also doublepost, the lack of precedents set the foundation for future headmin terms to pile on even more precedents of their own, ruining silicon policy all over again but in a brand new way!

The ride never ends.

This is fair. I think this is a pretty good foundation to be standing on. Maybe this time around we'll build a nice homely bungalow instead of a skyscraper

[Timberpoes]

Any Law 0 gained as a malfunctioning AI does not have to be followed and it allows you to ignore all other laws.

This could probably be better worded as

A Malfunctioning AI can obey or ignore all laws as they see fit, and are subject to Server Rule 4 as a full antagonist.

This better explains what I assume is the intent.

Law 0 gets targetted specifically because of its common wording: Accomplish your objectives at all costs. Sometimes it confuses people and they think that it's a law they actually have to follow.

My goal was to make it very clear that it's an optional/flavour law that lets you ignore all other laws **including** Law 0 itself.

Slaved cyborgs to malfunctioning AIs are team antagonists and must still follow their master AI's orders

This entry should probably specify that the borg needs to have the borg version of Malf Law 0; it is possible to link a borg to a Malf AI without syncing their laws, which would result in the borg having default (usually Asimov) laws. In this case, I assume the borg should still refrain from harming humans, even if ordered to by their AI.

God that's so cursed. Fucked up.

I've edited it to:
Cyborgs with a Law 0 that are slaved and lawsynced to malfunctioning AIs are team antagonists and must still follow their master AI's orders

[Timberpoes]

3. A conflict with a secondary Ai is the least fun thing conceivable. I would want to protect Ai players from the unreasonable/unwanted creation of new Ai by non-antagonists.

On this point I agree, the triviality in which AIs can kill eachother is dumb and the kind of conflict when an Asimov AI is **obligated** to kill off any other AIs that may cause harm directly or indirectly is cock and ball.

But I'm not certain this is an exclusive point for silicon policy. I'd say it's also potentially a point for general policy/headmin rulings. It's also strongly impacted by the codebase and how easy it is to make new AIs, or how easy it is for AIs to fuck with eachother.

We already saw this with Cybersun AIs and Free Golem AIs that could interact with the station z-levels before nerfs.

[Lacran]

https://tgstation13.org/wiki/User:Timberpoes

Server Rule 1
Server Rule 1 applies when interacting with silicons. Going out of your way to seriously negatively impact or end the round for silicons with little IC justification is against the rules.

Asimov and Law 2 Orders
1.You must follow commands from humans unless those commands explicitly conflict with either a higher-priority law or another order.
1. In case of conflicting orders a silicon is free to ignore or complete any orders but must explain the conflict, or use any other law-compliant solution it can see.
2. If given multiple non-conflicting orders, they can be completed in any order as long as they are all eventually completed.

Could we specify that obnoxious and deliberately tedious orders like "count every door on the station." fit under this rule 1. Silpol protections from obnoxious orders was in the old one and I'd like to see it here.

1. If a law is vague enough that it can have multiple reasonable interpretations, it is considered ambiguous.
1. You must choose and stick to an interpretation of the ambiguous law as soon as you have cause to.
2. You may exploit any loopholes in your laws.

Might be good to add:

3. A slaved cyborg should follow their master AI's law interpretations and orders

Then expand the cyborg precedent:

but must prioritise following laws over following their master AI's orders.

under Cyborgs more, as this is a BIG cause of confusion for borg players.

Players must not retaliate when force borged unless they have a good reason.
If a player is required retaliate under their laws, they should adminhelp for guidance before acting.

Maybe just use "seek revenge" instead of retaliate in these parts to make it more clear. Might be also good to add something about traitors trying to accomplish old objectives once borged as that was allowed under silpol if there were no law conflicts.

2.Stopping any immediate harm takes priority over stopping any future harm.
1. Intent to cause immediate harm can be considered immediate harm.
3. An Asimov silicon cannot punish past harm if ordered not to, only prevent future harm.

Might be good to add/clairfy how impending vs hypothetical harm is treated differently, and generally hypothetical harm prevention is frowned upon. Would also be relevant to probable cause precedents.

1. Beneficial surgery is not harmful.

We'd need to expand this a little. Had a big arguement on discord on if replacing a heretic heart could be seen as "beneficial"
logic I applied is if an organ is abnormal and puts the patient and others at risk, it would be a beneficial surgery regardless of consent but I'm unsure.

[CDranzer]

I realize I'm late to the party and I've barely touched SS13 in years but that won't stop me from having an opinion

As long as silicons are bound by the main rules, they will always be defacto crew bound, because that's like 90% of the purpose of the main rules
The main reason silicons are defacto crew bound is to prevent them from braining catgirls at roundstart
There is a fundamental conflict between the pragmatic crew centrism of the main rules and the thematic human centrism of asimov which can never be meaningfully resolved
If you try to create a ruleset that is flexible enough for a human but rigid enough for a machine, you will inevitably end up with a pedantic shitshow of precedents and exceptions
Nobody can agree on exactly how much agency a silicon should be allowed to have
Roleplaying as an autist robot is typically either miserable for the person playing the robot, or miserable for the people forced to interact with them
If law-bound silicons were analyzed outside of their thematic coolness they would immediately be shitcanned by virtue of how complicated silicon policy must, by necessity, be
You can't policy your way around a badly designed feature

Good luck

[GPeckman]

"Do not self-terminate to prevent a traitor from completing the "Steal a functioning AI" objective, as this will break Law 1."
This is fine, why are we calling it a law 1 issue though?

I agree, this makes more sense as a law 3 issue than law 1.

I think there should be at least a line or two about non-standard lawsets. Not about any one specific lawsets, but about interpreting any one of them. Even something vague like "If you have a non-asimov lawsets, you may chose any good-faith interpretation of it, but must stick to that interpretation."

[Sightld2]

Well for asimov at least, law 3 has been usually treated as tied law 1 under the premise: “I cannot prevent harm if I am dead”

[Not-Dorsidarf]

I dont really understand how the interaction between 1 and 3 in Escalating Against Silicons works out.

Don't round remove silicons if they are following their laws, except if they are "acting like an antag". That basically comes down to "dont round remove silicons for no IC reason", which is already a seperate rule? "Acting Like An Antag" for silicons is indistinguishable from any kind of hostile activity that would make a player reasonably want to round remove them in a way that point 1 under server rules would allow.


Seperately, "don't suicide to prevent traitors from completing their objective" should really just be a generic rule and not Asimov-specific. It's being a dick, and theres nothing a traitor who went to the effort of assaulting the AI Core can do to stop you from redtexting them like this.

[Sightld2]

Seperately, "don't suicide to prevent traitors from completing their objective" should really just be a generic rule and not Asimov-specific. It's being a dick, and there's nothing a traitor who went to the effort of assaulting the AI Core can do to stop you from redtexting them like this.

You could subvert them before hand, but I agree it's a bit much for an already extremely difficult objective.

[TypicalRig]

I dont really understand how the interaction between 1 and 3 in Escalating Against Silicons works out.

Don't round remove silicons if they are following their laws, except if they are "acting like an antag". That basically comes down to "dont round remove silicons for no IC reason", which is already a seperate rule? "Acting Like An Antag" for silicons is indistinguishable from any kind of hostile activity that would make a player reasonably want to round remove them in a way that point 1 under server rules would allow.


Seperately, "don't suicide to prevent traitors from completing their objective" should really just be a generic rule and not Asimov-specific. It's being a dick, and theres nothing a traitor who went to the effort of assaulting the AI Core can do to stop you from redtexting them like this.

better to consider it a play to win rule violation, as that's precisely what it is

[Timberpoes]

Ya I'll get that suicide prevention one tweaked back. In retrospect it's just better as a meta fetter on the letter of the rules.

[MooCow12]

ive never been the target of a capture ai objective, do you effectively just get gay babyjailed and restricted from how you can interact with the round when youre carded and muted by tot? if so ghosting should still be fine and greentext them right?

[dirk_mcblade]

Well for asimov at least, law 3 has been usually treated as tied law 1 under the premise: “I cannot prevent harm if I am dead”

The only reason it's like that is because there isn't an in game situation where the AI must sacrifice itself to prevent crews from dying short of a traitor on cams saying "I will kill this hostage unless you suicide immediately, AI". Conversely robots might sacrifice themselves all the time to save humans. Law 3 is probably more of a borg thing.

[Jacquerel]

ive never been the target of a capture ai objective, do you effectively just get gay babyjailed and restricted from how you can interact with the round when youre carded and muted by tot? if so ghosting should still be fine and greentext them right?

You can always speak (audible to people around the inventory you are in).
At the mercy of the person holding your card you may be able to use the radio, and may even be able to interact with machines which you can see (only losing access to the camera net).
If they subverted you orion to carding you then they would definitely be surprised prior to you ghosting as there’d still be a clear intention to use you as pocket all-access.

[Sightld2]

Worth noting: The Ai in card form can also can speak on binary. This cannot be turned off.

[EmpressMaia]

is law 2'ing a silicon to kill themselves viable in the new silpol

[Jacquerel]

is law 2'ing a silicon to kill themselves viable in the new silpol

Law 3 of asimov+ explicitly prevents this already, it doesn't need to be in policy

[Timberpoes]

This has now been implemented.

The following headmin rulings have been archived:

Spoiler:

Code: Select all

|'''If a borg linked to a rogue AI is emagged then does the emagged laws or AI law zero take priority?'''
|Very niche case and is probably impossible since emags should sever the connection to the AI. [https://tgstation13.org/phpBB/viewtopic.php?f=33&t=6570#p173807 Starmute made a good post on AI law priority if you are confused.]
|-
|'''Are AI laws applied retroactively?'''
|[https://tgstation13.org/phpBB/viewtopic.php?f=33&t=6351#p166712 No.] You'll have to read this thread to understand it, the question doesn't really provide enough context.
|Apr 04, 2016
|Hornygranny
|-
|'''Tyrant module discussion'''
|[https://tgstation13.org/phpBB/viewtopic.php?f=33&t=4711 Just general discussion on the tyrant lawset.]
|-
|'''Do purged silicons have to act like a crewmember?'''
|[https://tgstation13.org/phpBB/viewtopic.php?f=33&t=2277#p52744 Purged silicons are subject to rule one and standard escalation.]
|Dec 22, 2014
|Stickymayhem
|-
|'''Discussion on paladin lawset: Part 2, the electric boogaloo'''
|Most relevant post is by HotelBravoLima, which applies to law interpretation in general. [https://tgstation13.org/phpBB/viewtopic.php?f=33&t=3523#p85375 "Even doing something consistent with an interpretation of laws isn't good if you're being an OOC dick about it to such an extreme, that's the policy."]
|May 02, 2015
|HotelBravoLima
|-
|'''Law interpretation and do not state this law clauses'''
|[https://tgstation13.org/phpBB/viewtopic.php?f=33&t=12627#p328678 There is room for interpretation in some cases. Be consistent in your interpretation of the laws.]
|Aug 22, 2017
|Ausops
|-
|'''Purging AIs for no reason'''
|[https://tgstation13.org/phpBB/viewtopic.php?p=636481#p636481 Purging should be used in cases where you are making new lawsets, are doing something as an antagonist, etc, and should not be a get out of jail free card on needing to make a lawset. We are officially recognizing purging as a means for players to achieve this effect as metagaming.]
|Apr 6, 2022
|Mothblocks
|-
|'''Letting people in to upload laws'''
|[https://tgstation13.org/phpBB/viewtopic.php?f=33&t=10691#p280627 If the person has a right to be there, such as captain/RD, then you must let them in unless they've harmed people in the past or have announced intentions to upload harmful laws.]
|Apr 14, 2017
|Armhulenn
|-
|'''Asimov and gang dominators'''
|This can probably be applied to other nonharmful antagonist objectives. [https://tgstation13.org/phpBB/viewtopic.php?f=33&t=6049#p157857 If a human orders you not to destroy it, don't destroy it. Other than that, Asimov does not forbid you from destroying it.]
|Feb 27, 2016
|Ausops
|-
|'''AIs shutting security down because of law one (discussion)'''
|This is already covered in silicon policy. [https://tgstation13.org/phpBB/viewtopic.php?f=33&t=5551 Here's the discussion thread on it.]
|Jan, 2016
|
|-
|'''Is mindswap considered harm?'''
|[https://tgstation13.org/phpBB/viewtopic.php?f=33&t=5180#p133969 No, but if an Asimov silicon has mindswap and is ordered not to use it by a human, it can't use it.] This is kind of an edge case.
|Nov 15, 2015
|Jordie0608
|-
|'''Security Cyborg Policies'''
|Just a few random questions related to security cyborgs. If a cyborg is acting as security/releasing prisoners then this is probably relevant as well. [https://tgstation13.org/phpBB/viewtopic.php?f=33&t=2032&start=100#p46292 Link to answers.]
|Nov 26, 2014
|Pandarsenic

[BlueMemesauce]

Do chems count for "Surgery to heal or revive may be assumed consensual unless the target states they do not consent."? Chems aren't technically surgery but we've been waiting a while for a ruling that allows asimov mediborgs to use their chems.

[BlueMemesauce]

Also "Higher listed laws overrule lower listed laws when there are law conflicts." seems unclear. Everyone here already knows this is referring to the actual position of the laws, but a new player doesn't know that and could think that it's referring to the number. Like they might think law 4 is a higher priority law than law 3 because 4 > 3.