Limit AI's ability to be extension of security

[NerdyVampire]

I am very much a supporter of Lordnesh' AI mindset. In order to cultivate good AI players it's necessary to teach how giving antags leeway benefits the round in the long-term. The round becomes much more fun for all if the antag succeeds in getting the spare ID, or if they manage to get their hostage, or if they manage to slip your cameras.

I am personally against any further nerfs to the AI, I think we're already reducing it to something unworthy. But policy wise, or law wise, I'm game for changes. I don't mind having to provide evidence in the form of images, if that's what it takes, or cooperate in some other manner. On the whole I think security is already very lax when it comes to evidence, that amounts to more than what they found during the processing search.

 

[Susan]

It already is, in a way. If you disable suit sensors, you can only be tracked while visible. People who have sensors set to full can even be tracked in areas with non-camera coverage because you know their exact coordinates. 

Just buy an agent ID. 

Edited October 18, 2020 by Susan

[Amunak]

22 hours ago, KingOfThePing said:

Dont open emergency shutters if you dont know what is on the other side, they are everywhere. They also almost always go down, so there are not really many reasons to bolt doors except deny access.

True, that was a bad example. But bolting doors because there's, say, a spider in the room would be a legit use case.

Also the heads might want to give the AI an order to block access to some place for whatever reason, and we can't do that now.

Oh, and bolting doors *open*. Like, I just think it's a nice, fun mechanic for the AI that shouldn't have been removed from the game.

What do you think about adding something like a override permission on higher IDs so that they could then unbolt (or even bolt) doors with an alt-click or something?

22 hours ago, KingOfThePing said:

People don't care, they already know that, it's not working

Do they? You could easily play a chef for a week, never even leaving kitchen, and then go on to play AI and play it like shit against antags. I actually considered playing AI and I didn't know; I probably wouldn't have ruined peoples' fun but you never know. I would definitely take more care after reading this topic. Having a message there wouldn't hurt anything.

Obviously there is always the nuclear option of just putting it behind the same whitelist as department heads are.

17 hours ago, Lordnesh said:

I would support linking camera following to suit sensors. Although, agent IDs have a function that make it so the AI can't track you.

Oh, then I guess it doesn't matter much? It would still help antags some, but it could potentially make the AI's job even harder. And regular crew could be shits about it as well.

7 hours ago, Susan said:

If you disable suit sensors, you can only be tracked while visible.

What specifically do you mean by "while visible"? I thought you always had to be visible (except, I guess, in the areas without camera coverage).

Edited October 18, 2020 by Amunak

[Susan]

While visible to station cameras. If you're in maint or anywhere without cameras you cannot be tracked.

[KingOfThePing]

3 hours ago, Amunak said:

Do they? You could easily play a chef for a week, never even leaving kitchen, and then go on to play AI and play it like shit against antags. I actually considered playing AI and I didn't know; I probably wouldn't have ruined peoples' fun but you never know. I would definitely take more care after reading this topic. Having a message there wouldn't hurt anything.

Obviously there is always the nuclear option of just putting it behind the same whitelist as department heads are.

True, a message surely won't hurt, I agree. Also I did not think about bolting doors open, that really was useful in a lot of situations. 

A whitelist for AI was not wanted in the past and I think it is not a good solution to the problem. There must be other ways.

[Amunak]

14 hours ago, Susan said:

While visible to station cameras. If you're in maint or anywhere without cameras you cannot be tracked.

Ah right, I didn't know that was possible before / on other servers.

 

14 hours ago, KingOfThePing said:

True, a message surely won't hurt, I agree. Also I did not think about bolting doors open, that really was useful in a lot of situations. 

TBF it would be possible to allow AIs bolt open doors only, but at that point we should probably allow both. Alternatively (or in addition) there is something I wanted to do for a long time regardless; adding an option to timings in doors that's "timers off" which would make the doors not close automatically at all so they'd have to be clicked (and possibly same with opening? idk). But I do miss that.

I'll look into adding the message.

[Chada1]

8 hours ago, Amunak said:

TBF it would be possible to allow AIs bolt open doors only, but at that point we should probably allow both. Alternatively (or in addition) there is something I wanted to do for a long time regardless; adding an option to timings in doors that's "timers off" which would make the doors not close automatically at all so they'd have to be clicked (and possibly same with opening? idk). But I do miss that.

I'll look into adding the message.

There's an intricate justification for why the AI (And 'borgs) lost the ability to bolt closed and as one of the people who undertook that change in their design I can explain why.

One of the things you've suggested as a potential use turned out to be near the actual only use that was called for. Instead of being called to protect Crew from harm, Command and Station Crew, sometimes even the AI, but not always, were using the AI as a weapon against Antagonists, it's not even only the AIs fault.

A player culture actually formed around requesting the AI to bolt the Antagonists into a room or block it off, so Security could shut the round down. This is completely unpreventable under the AIs' lawset, since if a Command member requests they do a thing, they generally do a thing within reason.

This is what we know as negative utility, utility that is actually negative to the narrative of the round. You can even wonder why a Command member shouldn't request the AI to do this, it's actually not that unreasonable ICly. We want to remove as much of this negative utility from the game as possible and implement as much positive utility as possible.

Any valid, positive utility, use of bolting closed, the AI can just as easily use by pulling the fire alarm for the area, or dropping the fire shutters for afew rooms. It doesn't need bolting closed to maintain those uses. Which is why we reached the conclusion of removal.

Positive utility being stuff that actually contributed to the narrative of the round, like actually protecting the Crew from harm and disasters, instead of shutting down the narrative via bolting in antagonists/etc.

(Edit: Actually no the ability to bolt open was only drawn up as a possibility, it was never added so removing that part.)

The gameplay station-wide has improved like 10fold since the AI lost the ability, too, so practically, it's holding up.

 

Edited October 19, 2020 by Chada1

[Chada1]

Also, for anyone reading, I edited my post there.

[Amunak]

41 minutes ago, Chada1 said:

A player culture actually formed around requesting the AI to bolt the Antagonists into a room or block it off, so Security could shut the round down. This is completely unpreventable under the AIs' lawset, since if a Command member requests they do a thing, they generally do a thing within reason.

This is what we know as negative utility, utility that is actually negative to the narrative of the round. You can even wonder why a Command member shouldn't request the AI to do this, it's actually not that unreasonable ICly. We want to remove as much of this negative utility from the game as possible and implement as much positive utility as possible.

Yeah, that does sound undesirable. I would expect the whitelisted command roles to know better (was the bolting removal before or after whitelisting command roles?) but I guess when you have a legit ICly tool to simply handle a situation there is no good ICly reason not to use it. In that sense I agree with the removal, as it's no longer even really a player problem. Thank you for the explanation.

With that in mind I'll explore options to add gameplay elements for AIs that hopefully don't have this issue.

[Chada1]

6 minutes ago, Amunak said:

Yeah, that does sound undesirable. I would expect the whitelisted command roles to know better (was the bolting removal before or after whitelisting command roles?) but I guess when you have a legit ICly tool to simply handle a situation there is no good ICly reason not to use it. In that sense I agree with the removal, as it's no longer even really a player problem. Thank you for the explanation.

With that in mind I'll explore options to add gameplay elements for AIs that hopefully don't have this issue.

NP at all, and if you find other ways to improve the AI, go for it, DO IT ?
And it was long after the addition of the whitelist system, but even if Command members were careful about it, it's... still very hazy on if a Security member (JUST an officer) requesting the AI to do a thing like blocking off an Antagonist, could actually be refused. I'm p. sure you'd have to really Really push it to actually say no, and it'd be a no from the perspective of OOCly giving the Antagonist room to breath, not an actual IC no... which would still be a problem. I doubt anyone would mind adding open-bolting back into the game, the developer who coded the removal actually has a draft PR up for it, they've just been super busy.

And open bolting is positive utility, not the awful negative utility, so it'd be in the clear.

Edited October 19, 2020 by Chada1

[Lordnesh]

It's difficult, for me, to think of ways to mechanically nerf AI more then it already is. Mostly because it can't really do much anymore. With the removal of the mech, it's greatest strength at the moment appears to be the automatic follow function of the cameras. However, you can't simply remove the feature, because it impacts other things. As I've mentioned, a lot of antags have ways to counter this already. Agent ID's have a scramble function to interfere with the AI's ability already, and lings have a similar evolution. If you remove the camera following feature, or restrict it, those features become defunct.

 

I like the idea of adding a message about the AI's role in a round, something similar to the traitor message they get at start. Because, like with most things, it's not so much a problem with the tools, but about the people that misuse the tools. Either deliberately for power gaming, or accidentally.

 

With that said. I don't like the idea of removing it completely, as some people would wish. I'm my opinion, the advantages it offers the crew, and even antags, outweigh the disadvantages. In my opinion, one of its biggest strengths is the ability to interface between departments. Things like, warning medical to expect patients because security is freaking with a hostile situation. It also offers opportunities to antags. A simple law addition can suddenly give them the advantage.

[Amunak]

I created a PR that concerns some issues brought up in this topic.

• Feedback thread
• PR#10310

[Lordnesh]

2 hours ago, Amunak said:

I created a PR that concerns some issues brought up in this topic.

• Feedback thread
• PR#10310

One thing I've noticed, and haven't seen addressed, is the shield generators near the core. They were added as an extra layer of defense in extreme circumstance. However, with the removal of the mech, they no longer function without outside help. Has anyone considered making them bolted to the floor by default?

[KingOfThePing]

The shield generators have been added ages ago, the remote mech was a fairly recent addition. 

For SOME reason they are not bolted down. They always required outside help of some kind. Making them bolted down would actually make them serve a purpose. Crazy. Please make it so.

[Butterrobber202]

You know we could just reduce Camera Coverage and that alone would give antags magnitudes more freedom.

[Lordnesh]

I just had an idea. What if we link being able to follow people on cameras to their ID/PDA? That way, an antag can just hide/toss their ID/PDA, and they can't be tracked.

[Chada1]

31 minutes ago, Lordnesh said:

I just had an idea. What if we link being able to follow people on cameras to their ID/PDA? That way, an antag can just hide/toss their ID/PDA, and they can't be tracked.

 

2 hours ago, Butterrobber202 said:

You know we could just reduce Camera Coverage and that alone would give antags magnitudes more freedom.

Removing Camera coverage would be p. awful for AIs since their ability to roleplay/even watch is also proportional to their camera coverage.is the problem with that butter.

Lordnesh, that's a v. good idea but I dunno how hard it'd be to code.

[StationCrab]

10 hours ago, Butterrobber202 said:

You know we could just reduce Camera Coverage and that alone would give antags magnitudes more freedom.

Add more tools for antags to disable camera networks. Jammers, electronic stealth modules. Maybe even let Antags emag cameras so they display a still image of the area that doesn't update when people pass by, to fool the AI into believing the area is clear until such time as it figures out what has happened by watching travel from other networks.

Simply blinding it is too hamfisted I think. Recently radio jammers were made to stop station bounds from remotely accessing doors near the jammer, maybe do something similar for cameras? Throw a jammer down the hall and suddenly the AI can't see, or perhaps the image is distorted.

[Butterrobber202]

11 hours ago, StationCrab said:

Add more tools for antags to disable camera networks. Jammers, electronic stealth modules. Maybe even let Antags emag cameras so they display a still image of the area that doesn't update when people pass by, to fool the AI into believing the area is clear until such time as it figures out what has happened by watching travel from other networks.

Simply blinding it is too hamfisted I think. Recently radio jammers were made to stop station bounds from remotely accessing doors near the jammer, maybe do something similar for cameras? Throw a jammer down the hall and suddenly the AI can't see, or perhaps the image is distorted.

this would be epic

[Carver]

On 24/10/2020 at 21:53, Lordnesh said:

I just had an idea. What if we link being able to follow people on cameras to their ID/PDA? That way, an antag can just hide/toss their ID/PDA, and they can't be tracked.

I'm going to say no because as an antag your PDA is fairly integral unless you chose the Headset uplink - and being able to track kidnapped people who had their ID/PDA stolen is valuable. Antags that aren't dragging other people around have decent tools to prevent tracking (agent IDs, electronic warfare module, etc.).

My only suggestion that could help antags more would be to port/recreate this antag item from /tg/: