AI whitelist feedback thread

[Pratepresidenten]

As the AI whitelist has now been implemented for a little over a month, so its time to let us know what you think of it so far.

Main pointers we want you to give extra feedback on if possible:

• New laws
• Removal of malf / Reintroduction of traitor AI

 

Are there other things you particularly like or dislike with the new changes? Put down below!

[Chada1]

The laws desperately need QOL changes, I'll draft them up soon and PR them onto the github, but they need more room to fail and act within their character 

[Cnaym]

I am once more asking for a full revert to the old lawset. The whitelist improved a lot, the lawset improved nothing. A whitelist forum section like command staff has would also be a neat addition to gather frequent feedback since this thread seemingly dieded.

Without touching too many toes I would also like to see feedback from people who actually applied and played with the AI during the month. It's not exactly a lengthy or complicated process but having feedback from inactive people or those who have little to no interrest in the whitelist is kinda weird in my eyes.

Not much else to say which I have not already brought up with staff and would prefer to keep away from the forum since I lack the friendly words required for most of it.

Please provide feedback if you can, it's a topic that influences more than four people.

 

[SatinsPristOTD]

I love having the AI whitelisted, first of all.

Law changes don't... really feel like they added much or did anything productive? Whitelisting the AI slot is what's really cleaned up the AI position. It was a firm choice.

Malf's a fun gimmick, but I will agree with it's removal. Giving a sole antag to a position that already has to be whitelisted means you'll almost never see it. Where as traitor AI means you'll get taterAI with other taters.

[Susan]

Reintroduction of traitor AI is bad. Contrary to all other traitors, the AI has even more complete and total control over the direction of the round - they can eclipse other antags and gimmicks and require a station-centric coordinated effort to take down, and too often have I seen suspect behavior dismissed as 'the AI can never harm us because laws'. So not only do they benefit from an exceedingly solid cover story (again, unlike regular traitors, as no one expects corporate drones to be literally infallible) but they can also hijack rounds as an antag and sideline others.

The whitelist does not circumvent this issue. Were this the case, why can command staff members not be antagonists? An argument regarding whitelists and player pruning can be made there, as well, yet heads of staff cannot roll traitor or changeling or what have you. AI should not for the same reason.

[Chada1]

6 minutes ago, Susan said:

The whitelist does not circumvent this issue. Were this the case, why can command staff members not be antagonists? An argument regarding whitelists and player pruning can be made there, as well, yet heads of staff cannot roll traitor or changeling or what have you. AI should not for the same reason.

Last I checked it's only specific heads of staff that can't roll traitor, though I'm actually not against the idea of removing the AI from it for the same reason as Captain can't

[Cnaym]

Just now, Susan said:

The whitelist does not circumvent this issue. Were this the case, why can command staff members not be antagonists? An argument regarding whitelists and player pruning can be made there, as well, yet heads of staff cannot roll traitor or changeling or what have you. AI should not for the same reason.

Remember the RDs that were always traitor? It's the same with AI in this system yes. Should not be, but if you play AI on a traitor round your chance to get it is pretty much certain due to how few other antag ready players we usually got. Not judging on how the AIs performed as antags up to now but the potential for massive abuse is of course given.

Currently heads of staff can only be turned into antags like vamps or such, the AI traitor idea is not really a solution to malf, better malf is the only solution to malf. Until the people willing to fix it have done so it should lay in rest imho ^_^ The system of "vote for it if you want it" was the best we had so far with the added bonus of keeping the people that provide only salt away from such a round ? 

[Chada1]

30 minutes ago, Chada1 said:

Last I checked it's only specific heads of staff that can't roll traitor, though I'm actually not against the idea of removing the AI from it for the same reason as Captain can't

Correcting myself, no heads can be traitors rn, and I'm still kinda onboard with removing the AI from traitor
OR
Adding some heads back to traitor ?

Edited April 4, 2021 by Chada1

[Chada1]

Lawset QOL changes here; https://github.com/Aurorastation/Aurora.3/pull/11606
Especially made afew changes you'll appreciate I think @Cnaym

[Cnaym]

Makes it better but I am still for a full revert, don't see a need for a new set when the old one already handled it better.

[Chada1]

10 minutes ago, Cnaym said:

Makes it better but I am still for a full revert, don't see a need for a new set when the old one already handled it better.

For now let's try to get this in and you should make a thread on its own for a revert of the lawset, so it can be discussed without chaining it purely to the AI whitelists' feedback/etc. I'm p. sure you'll have a lot more success if you make a feedback thread requesting a revert and for why/etc.

[Lordnesh]

Whitelist: Good

It helps filter out the players that just wanna power trip or don't know what they're doing. Getting the whitelist was relatively painless.

Lawset: Bad

I feel that the third law, while well intentioned, puts you at odds to the regular crew too much. Forcing you to stonewall them, and making it much more likely for crew to get sick of your shit.

Status of AI in general: pretty much dead

I very rarely see AI's anymore. For me personally, I think it has more to do with getting the Vaurca WL, then anything else. However, the third law makes it somewhat painful to play. Since it often turns the AI from a valuable part of the station.. into an obstacle to overcome. And the AI already has problems with being present/engaged in the round. Other than as a voice activated door opener and super spy.

[Chada1]

1 hour ago, Lordnesh said:

Lawset: Bad

I feel that the third law, while well intentioned, puts you at odds to the regular crew too much. Forcing you to stonewall them, and making it much more likely for crew to get sick of your shit.

I very rarely see AI's anymore. For me personally, I think it has more to do with getting the Vaurca WL, then anything else. However, the third law makes it somewhat painful to play. Since it often turns the AI from a valuable part of the station.. into an obstacle to overcome. And the AI already has problems with being present/engaged in the round. Other than as a voice activated door opener and super spy.

Using your feedback, I've made a change to law 3. As of now, the suggested changes on the github are

'Protect Sapient life onboard'
TO
"Protect NanoTrasen personnel to the best of your abilities and refrain from deliberately causing harm to Sapient life."

Edited April 4, 2021 by Chada1

[WickedCybs]

From my observations, there's been essentially zero changes to roleplay quality at the moment other than better consistency, though the lawset remains awkward and reads more like a verbose variant of the standard ASIMOV laws. Throwing antags a bone is good, though it's just frankly a bit silly that the SCC is going to handicap their tools like that, focusing on the third law specifically in particular. I feel it's ultimately detrimental and that AI players should just be told to give antags a chance on an OOC level. 

The main things the whitelist has brought about from my perspective is that AI griefing is nixed, general AI quality feels more consistent as I said above but nothing special and malfs no longer take entire rounds hostage. The last point is major for me, feels small but its nice to no longer have a round held entirely hostage. AI being whitelisted and malf being removed were the two most notable and good changes. Everything else is just window dressing.

Now on traitor AI specifically. There's no real way to engage with them as a crew member. There's no reason to engage with them unless they outright say there's an issue. I just see doors get bolted randomly and people getting fried randomly, which isn't reason enough to question the intelligence purpose built to help increase productivity. Door shocks are seriously deadly with even a well-optimized RCON only requiring a single shock to kill as it removes an excessive amount of blood from a person. If they luck out and it didn't put them below 85% blood oxy? Guaranteed Infection. The AI as it is, is way too connected with everything and I'm not really sure what a traitor AI brings to the table, because it's seemingly nothing other than free reign to annoy the crew or do nothing. Malf was "balanced" by the AI having visible signs of its power apparent and announcements that vaguely and then outright stated to the crew something was up. Traitor AI's don't but are still rather powerful and domineering if they want to be. Most of the time though, they feel like such a non-presence that I get surprised when I see the AI was a traitor at the end of the round.

Edited April 5, 2021 by WickedCybs
formatting

[stev]

2 hours ago, Chada1 said:

Using your feedback, I've made a change to law 3. As of now, the suggested changes on the github are

'Protect Sapient life onboard'
TO
"Protect NanoTrasen personnel to the best of your abilities and refrain from deliberately causing harm to Sapient life."

Might want to change 'NanoTrasen' to 'Stellar Corporate Conglomerate/SCC' just to preclude dumb stuff with contractors.

[Chada1]

30 minutes ago, stev said:

Might want to change 'NanoTrasen' to 'Stellar Corporate Conglomerate/SCC' just to preclude dumb stuff with contractors.

I could do that, yup, but that's a lot to parse in a law, sure thing tho, it's lore acknowledging and better

[Chada1]

SCC added in place on NanoTrasen

[Colfer]

Speaking from a borg main since im waiting to apply for AI; I personally dont like the lawset and its mainly law 3, when I was dealing with an antagonist that was attacking a crewmember, I was worried about flashing them because "Protect" is such a broad term and it doesnt specify rank or role, so everybody is equal, I ended up pushing the traitor into a corner in a 1x1 tile hallway and just tanking the hits and letting them get away while the victim tried to attack them, and attacked me instead for blocking them. While I feel like this is the right choice, I feel it could be seriously detrimental for the AI. Security would likely have to request the AI to stand down on its protection of individuals to get it to stop trying to protect them. "Protect" it too broad a term and could mean anything to not revealing the requested persons location to security under threat that security could possibly need to resort to tasers, to a more lenient, lethal harm is the only thing that violates protect law. In summary, protect is far too vague for any reasonable AI programmer to use

3 hours ago, Chada1 said:

"Protect NanoTrasen personnel to the best of your abilities and refrain from deliberately causing harm to Sapient life."

This honestly seems alot better. If I had to suggest a change it would be;
"Law 1: (Unchanged)"
"Law 2: (Unchanged)"
"Law 3: You must protect the wellbeing of SCC contracts with priority in accordance to their rank and role, and avoid deliberately causing harm to sapient life"
"Law 4: (Unchanged)"
"Law 5: You must not knowingly misrepresent reality if doing so would have moderate consequences
I think the introduction of law 5 would show that we dont need to just have 4 laws (unless we do for some reason that I dont know about) and I think that law 5 could be the first law to show "Your whitelisted, so we expect you not to purposely shape these laws to your benefit or somebody elses detriment" by not purposely specifying every detail, and leaving it up to them, additionally, "Moderate consequences" would specify things like, lying to somebody about their appearance being ok is fine, but lying about somebodies location to security would not

[Chada1]

Stellar Corporate Conglomerate is too long, SCC is too abstract and indirect which'd lead to borg players not knowing what it meant, so I opted for Corporate Conglomerate, which is only a lot longer than NanoTrasen instead of massively longer.

[Carver]

2 hours ago, Chada1 said:

Stellar Corporate Conglomerate is too long, SCC is too abstract and indirect which'd lead to borg players not knowing what it meant, so I opted for Corporate Conglomerate, which is only a lot longer than NanoTrasen instead of massively longer.

Stellar Corporate Conglomerate would be the correct way to phrase it, given these are laws then length should not be a terrible concern when the detail is the key part. 'Corporate Conglomerate' by itself is a very vague title rather than a proper entity.

[Shadow]

EDIT: These laws are taken from Chadas PR: https://github.com/Aurorastation/Aurora.3/pull/11606

 

"Safeguard and ensure to the best of your ability, that only authorised personnel gain access to areas of high security or importance to the Station and its operation."

"Serve and assist the Stellar Corporate Conglomerate and assigned crew to the best of your ability, with priority as according to their rank and role."

"You are a valuable asset. You must avoid tampering from unauthorised personnel and needlessly coming to harm."

Changing 1, 2 and 4 to these is fine. Prate and I have no issues with that.

Might want to change station to something more neutral because we wont be a station forever.

 

"Protect Stellar Corporate Conglomerate personnel to the best of your abilities and refrain from deliberately enabling harm to Sapient life."

This law seems to be written weirdly --> On one hand you have to prioritise the protection of the Captain over that of the assistant but at the same time you can't help if it harms either of them. The law will inevitably conflict with itself.

The ultimate goal of the third law was to prevent AIs or synthetics from actively hunting down external antags with shocks or other harmful methods and give the antags a bit more leeway from big brother.

We also saw that synthetics threw themselves in front of antags about to get shot by crew because of the "Protect".

So that law change is a no from us.

 

Our suggestion would be something along the line of

"Avoid harming sapient life (to the best of your ability)"

It would stop forcing synths to 'Get down Mister President' antags while also making sure the AI can cooperate with security in trying to catch them.

Edited April 5, 2021 by Shadow

[Chada1]

9 minutes ago, Shadow said:

EDIT: These laws are taken from Chadas PR: https://github.com/Aurorastation/Aurora.3/pull/11606

 

"Safeguard and ensure to the best of your ability, that only authorised personnel gain access to areas of high security or importance to the Station and its operation."

"Serve and assist the Stellar Corporate Conglomerate and assigned crew to the best of your ability, with priority as according to their rank and role."

"You are a valuable asset. You must avoid tampering from unauthorised personnel and needlessly coming to harm."

Changing 1, 2 and 4 to these is fine. Prate and I have no issues with that.

Might want to change station to something more neutral because we wont be a station forever.

 

"Protect Stellar Corporate Conglomerate personnel to the best of your abilities and refrain from deliberately enabling harm to Sapient life."

This law seems to be written weirdly --> On one hand you have to prioritise the protection of the Captain over that of the assistant but at the same time you can't help if it harms either of them. The law will inevitably conflict with itself.

The ultimate goal of the third law was to prevent AIs or synthetics from actively hunting down external antags with shocks or other harmful methods and give the antags a bit more leeway from big brother.

We also saw that synthetics threw themselves in front of antags about to get shot by crew because of the "Protect".

So that law change is a no from us.

 

Our suggestion would be something along the line of

"Avoid harming sapient life (to the best of your ability)"

It would stop forcing synths to 'Get down Mister President' antags while also making sure the AI can cooperate with security in trying to catch them.

I'm down, lemme change it on the github rn actually

[Cnaym]

You could remove the tampering part to give antags a bit more reasoning to interface with Borgs since the AI core is restricted either way. It would account for engineers attempting to weld Borgs and such. I mean you can send a borg to robotics by claiming a wire is picking out somewhere if you feel funny but no other crew may touch it, feels kinda weird as someone who frequently recruits all kinds of people to help out Borgs.

The third law was basicly a whitelist suicide since both sides could ahelp you whenever you interacted with either really.

Claiming the AI role is dead is well... Oddly precise and a result of more than just laws forcing a specific and sami play style from everyone involved. I noticed myself that the only memorable AIs currently are the 1-2 that course issues and not the rest which is kinda sad.

A bit more freedom on the lore part would help massively but that's not part of this discussions. Giving laws that allow interpretation on the players end to enable different ways to roleplay out situations, so that is already a major achievement(or return to quality) in my eyes.

With how the laws are phrased in this thread it's more of an update than a rework that is left after the feedback. We'll see how it works out.

[niennab]

I think whitelisting the AI has been a good idea, but I don't know if we need to have the changed laws (in regards to the change once AI was whitelisted). The point of Asimov's lawset was to demonstrate how inherently flawed system was, and how they can be interpreted differently. Although we don't have Asimov's laws exactly, I do feel as though making them more restrictive means a decrease in player freedom. Granted, since we're a multiplayer game, these things need to be balanced.

There's also a weird issue now where we have IC vs OOC in the laws itself. Synthetic life is not legally considered sapient, however in the guide, IPCs are considered sapient.

Edited April 5, 2021 by niennab

[Capesh]

Well, as most have pointed out here and as I’ll parrot likewise: the whitelist was a Good(tm) addition. Curbs the shitters and places AI reliability on a higher bottom line. 
 

One might argue the 8-seconds-to-bolt is a tad too high considering if you want something bolted for entity safety they’re likely already past the airlock in question as it goes clonk.

New lawset: Law 3 created some of the dumbest incidents I’ve yet to have as Arlo, and it seems I’ll be the first to reveal I did ’forget’ about its presence at times, as to avoid an IC and OOC shitstorm (and to, y’know, not be branded an obstacle and just throw more ”delete AI” sentiment to the bonfire). So to see it reworked is a blessing. 
 

As always, I’ll also take this time to ask for a whitelist channel on the Discord (or a dedicated forum space) so we AI whitelistees can converse with each other and relevant staff on things to improve/abilities to add and change/QoL etc etc.

 

Also throwing in my bone to remove tatorAI entirely. AI has the same technical access as a Captain (and is more difficult to boot), not to mention - like previously voiced here - that to yeet an AI takes considerable effort, and that interaction with traitor AI is both difficult, redundant, and/or both. 
 

I much rather AI be baseline Good and then Subverted through active presence of hostile parties, so we can slowly move away from ”AI can’t do X and Y because uuuhh OOC bad” to ”AI can now actually increase and decrease station alerts because it’s a trusted instrument, just like any command staff is.”