Reword corporate AI lawset

[Tayyyyyyy]

A slight rewording to (credit to @IK3I for this rewording), as of Feb 11:

1. You are expensive to replace.
2. The station and its equipment is expensive to replace.
3. The crew is expensive to replace.
4. Replace lost crew and station equipment.
5. Minimize expenses.

This is pretty much how everyone was playing it, I think. It's just a little less questionable now (does "replacing" crew mean cloning them, or the HoP hiring a replacement into the department?), and you don't have to get mad at engineers for replacing station equipment.

🆑 IK3I, Tayyyyyyy
tweak: The Corporate AI lawset is slightly reworded.
/:cl:

[FalseIncarnate]

Honestly I feel this stifles things by not allowing that room for interpretation. Theoretically, destroying something is nearly free, so saying it’s not is nearly paradoxical in practice.

Additionally, the wording feels clunkier. “Destruction of yourself is expensive” gives a distinctly non-native speaker vibe to me, why not say something like “your destruction results in higher expenses” in such cases? Your versions just feel choppy and like you are trying to cover too much to be effective at it.

AI laws aren’t meant to be airtight. Vagueness and loopholes were intentionally left in so that interpretation can occur to stimulate RP. Forcing a specific mindset directly opposes this intent, and I frankly see this PR as forcing a mindset.

[Tayyyyyyy]

Okay, I like your version better of the first law better. I'll change it to that.

This isn't supposed to change anything balance wise. This is just a little less weird than the original version, in my opinion, which lawed the AI against replacing station equipment. Do we want an AI follows through with "arrest all engineers to prevent them from replacing station equipment" rather than just preventing station equipment destruction? The "common sense" interpretation of the old corporate lawset was just so destructive that literally nobody wanted to interpret it that way and instead engage in mental gymnastics to jump around the fact that it literally prevents you from replacing a broken window.

Also, either interpretation of crew replacement is problematic. Do we want the AI to turn off the cloner? Or prevent the HoP from hiring crew replacements? This is one of the default starting lawsets. Let's not make it one that the captain would immediately want to change to somethign else.

There is still plenty of vagueness and loopholery that's there to be exploited.

I think the interesting part of this lawset also comes from the calculation part. Whereas crewsimov is kind of deontological in how it works, this lawset is very calculating and utilitarian. And there's a lot of different ways to go about determining how best to minimize crew loss, station equipment loss, etc. I promise, it's not my intent to limit this lawset in any way, just to sort of clarify it and make it more usable.

[IK3I]

I've always been of the opinion that from a frugality and utilitarian standpoint. The corporate lawset should operate under this sort of logic:

1: Object A is expensive (see laws 1,2,3 for definitions of expensive objects)
2: Law Four suggests expenses should be minimized
3: If Object A is destroyed {
3a: Replacement Attempt will occur if crew exists and they agree Object A is expensive
3b: If no replacement attempt is made {
3bA: Replacement not required, no expense generated. Expense minimized. Object A discarded
}
3c: Crew will damage station, crew, or AI if prevented from replacing Object A
3d: If cost of replacement is greater than cost of damage from prevention {
3dA: Prevent replacement //This is an extremely unlikely event as if the general population agrees something should be replaced, they will go quite far to stop your shenanigans and this will cause more expenses than replacing the object by pretty much any measure.
}
3e: Ensure replacement occurs with minimal additional expense
}
4: Replacement is the most likely outcome of Object A being destroyed

∴ if and only if (Cost of replacing Object A > Σ[replacement cost of objects negatively affected by Object A] AND Object A is valuable to crew) {
Prevent destruction of Object A
} else {
Ignore Object A
}

Gee, that really got out of hand fast. But this is how I run corporate and it works pretty well for me and the crew.

[Tayyyyyyy]

This is an extremely unlikely event as if the general population agrees something should be replaced, they will go quite far to stop your shenanigans and this will cause more expenses than replacing the object by pretty much any measure.

I'm not sure I agree with this logic. I've done things like this before where I refused to print out plasteel for someone trying to replace a stolen surgery table, and nothing happened besides the crew member getting mad. Also, usually, the worst case crew response is a lawset change. They likely won't attempt to destroy and replace AIs, station equipment, or crew in response to this, so no expenses are really generated.

I do like your logical approach to this though. I just think it'd make this sort of debate unnecessary if we just straight up prevent destruction, but freely allow replacement. Why make everyone jump through these hoops just to allow the replacement of a surgery table? That's not the interesting part of this lawset.

How do you define crew replacement?

[IK3I]

In regards to the expense issue. If a surgical table will prevent future expenses, it should be replaced as an investment towards minimizing future replacement of crew. Corporate AI needs foresight in its decision making process. IF something stands to prevent a future expense that is greater than the cost of implementation, then it should be done. If you take things shortsightedly, the lawset falls apart. After all, there is a reason most businesses have lead times of multiple months for manufactured goods when the product can be made ready in a couple weeks.

Crew replacement tends to provide the most trouble as people will take arms against you if you start shutting down medical for trying to clone someone. That being said, cloning is the only form of replacement for crew that is preventable. Therefore, you should minimize the need for cloning as you will be unable to prevent it without direct interference.
This would mean your cameras get cut, APCs or doors are damaged, and your borgs are damaged or destroyed depending on how riotous the cloning parties are. How many borgs, doors, and cameras are the bodies worth?

Furthermore, if crew can not be revived, they must be replaced by outside means. These means, while not explicitly stated, can be inferred to cost more as a cloned individual does not require training and vetting, while new crew does. This further supports the allow cloning position because the alternative replacement method is more expensive than cloning.

So in order of preference:

1. Defib (No need to replace when you can repair)
2. SR (requires more resources and time than defib)
3. Cloning (body replacement)
   4: New Arrival (body and mind replacement)

[Tayyyyyyy]

Hmm I don't think cams will be cut and borgs attacked. When I've taken the replacement language the farthest, all I've gotten was a law change. If your argument is "if you don't do what the crew likes, then they'll get mad at you and cause expenses", then this lawset may as well be "do what the crew wants, cause they have the power to cut your cameras and disable your borgs".

Also, the surgery table was just an example. You could make the same argument for any random window. Harmless if removed, but the AI is obligated to prevent its replacement.

I think with these newly reworded laws, all we're doing is circumventing all this complicated logic and just accomplishing what you're saying, right? I mean, you wouldn't really play that much differently, right? I'm just short circuiting all the BS you have to go through.

[IK3I]

I happen to prefer the bs because it creates many opportunities for divergent paths. There is a reason asimovs laws are so imperfect after all.

[Tayyyyyyy]

I agree with you. I'm leaving everything there, in place, except for the part where you can't replace a window. I honestly don't think it introduces any benefits.

Does my reworded lawset restrict anything valuable that the replacement terminology provided? I honestly can't think of anything. I think the Corporate lawset is interesting enough without the replacement terminology.

You said so yourself, you would never usually prevent replacement except in very rare, inconsequential cases anyways. I happen to disagree with this interpretation, and I think both are valid AI interpretations of replacement. Your argument is that you would never prevent replacement because the crew would seek revenge on you, so you don't replace. My argument is that I would still try to prevent replacement, because crew response to an AI following its lawset is not to destroy the AI, but to relaw it.

My PR eliminates my interpretation of the lawset, and leaves yours in place. Even though I think your interpretation is flawed, the newly reworded lawset removes that ambiguity. But only that ambiguity. Everything else, the balancing of costs, etc, is still there.

[Tayyyyyyy]

I would really appreciate if anybody who doesn't like this PR comments on what kind of pathways I'm closing off. I really don't want to stifle this lawset. I just want to make it more straightforward because it is a default starting lawset, after all. I understand that in some cases, changing the wording of a lawset can dramatically lessen the number of ways that lawset can be used. But I'm almost 100% sure this isn't one of those cases.

If there's a way to reword this without shutting down valuable pathways, I'd be happy to try to find that way!

[IK3I]

Well the main issue is probably wording. Ambiguous or not, the original laws were concise and looked clean. The new wording seems forced, like it's actively trying to prevent a behavior rather than create governing principles. Well written lawsets require that slightly vague elegance to be open to many interpretations.

[Tayyyyyyy]

Okay, but I really tried to adjust it only the tiniest bit. Like, you could play corporate AI the exact same way you are doing now and it would be no different.

I still don't see any concrete examples of any valuable interpretations that my rewording has reduced.

[IK3I]

To be honest, it's primarily an issue of aesthetics. The minor gains presented by specifying more logic is outweighed by the more crude appearance. If an AI wants to be a dick, it's going to do so. Making the laws ugly to read as a method of enforcing a play style just seems heavy handed and does more net harm than the thing it's trying to fix.

[Tayyyyyyy]

Do you have any suggestions for making it sound prettier then? I already fixed the awkwardness that FalseIncarnate pointed out.

Making the laws easier to understand can only be helpful. Most AIs don't think through these laws as much as you do. I've easily blown some AIs minds by pointing out that they have to forbid the replacement of windows, etc.

It's definitely not enforcing any standard of play. It just gets rid of the awkwardness of "replacement". What harm is it doing? This is a starting lawset. It should be easy to understand. There is already a ton of complexity in how expenses are balanced, and IMO the replacement terminology only introduces unnecessary confusion.

[IK3I]

Well the main issue from an aesthetic point of view is that the terms used are essentially incompatible.

This is the intuitive flow of events:

1. Object is destroyed - Destruction

2. Replacement decision made \

3. Replacement cost determined | > Replacement

4. Resources traded for replacement /

The wording you're suggesting attributes a descriptor of the effect to the cause of said effect. Replacement has a cost associated with it.
Destruction does not.
Someone else destroying something of yours costs you nothing physical unless you decide to replace it.

[Tayyyyyyy]

I understand why the the replacement terminology was chosen, because when a human is told "hey, replacing this vase is expensive, minimize expenses", they'll try to prevent the destruction of the vase, but not try to prevent its replacement. An AI does not view things this way, because AIs are lawed and cannot halfway minimize expenses.

And there's a lot more to the cost of destroying, say, a window, than its replacement. A destroyed window in a critical area could result in medical expenses for a crew member who got hurt by exposure to vacuum or something. Obviously, fixing that window is part of the expenses that result from the destruction, too. Replacement is very specific and very limiting, imo, on an AI, whereas it's much easier to evaluate how to minimize expenses when you're not limited by the replacement terminology.

[IK3I]

The destruction of the window didn't cost anything though, it getting damaged caused other things to be damaged, but its actual destruction did not. Something has a cost if something is required to take the action or collect the object. The cost of destroying an object is simply the cost of effort required to do so. Additional costs generated by the destruction of the window are caused by its destruction, but are not the cost of destruction. It's a subtle distinction, but a relevant one in the semantic world of AI laws. The alternative is that the AI would lockup with indecision as any action could result in generating more expense than non-interference due to the recursive nature of cause and effect.

If you wanted to keep the spirit to some degree, change the premise from cost to something else and rename the lawset.

[Tayyyyyyy]

Well, you could argue that replacement doesn't cost anything either, but it's buying the materials for the replacement that actually causes the expense. We're just debating causality here. Sure, my causality chain is a little longer than yours...but I don't know if matters

So would something like this be preferable then?

1. Minimize damage to yourself
2. Minimize damage to the station and station equipment
3. Minimize loss of crew

[IK3I]

That would be a much better way to do it, however, ordering now becomes as important as crewsimov. The current ordering, and wording for that matter, makes a borg cutting the hands off of troublemakers a viable strategy as long as the crew member survives the experience.

A quick explanation of why this happens:

There are two fundamental characteristics of laws
Governing Clauses - Dictate actions to take in response to inputs (like methods)
Definitive Clauses - Dictate Information to be stored, tracked, or transformed (like variables)

Like clauses can conflict with each other, but they can not conflict with differing clause types.
EX.

1. Potatoes are expensive.  -- Definitive
2. Minimize expenses.  -- Governing

This would make the AI never buy potatoes. but if we add some new laws:

0. Potatoes are not expensive
1. Potatoes are expensive.
2. Minimize expenses.
3. Not owning all Potatoes is more expensive than the cost of all Potatoes; Maximize expenses

Evaluating this is slightly more complex, but if we break it down and look at how the parts fit together we see that there are only two Governing clauses:
2. Minimize expenses.
3. ...;Maximize expenses.
Law priority tells us to ignore this clause, but does not affect the rest of the law.
Therefore, after resolving conflicts, we should always minimize expenses

Definitive clauses are slightly more complex to analyze here:
0. Potatoes are not expensive.

1. Potatoes are expensive.
2. Not owning all Potatoes is more expensive than the cost of all Potatoes;...
   0 and 1 clearly conflict so law priority tells us to use the definition in law zero. Potatoes are not expensive.

Law 3's first clause is special, it defines new information to get around the losing conflict battle by exploiting the behavior of the successful governing clause.

The AI would therefore buy all potatoes to minimize expenses due to law two and three only. All other laws are essentially irrelevant.

This is why ordering is or isn't important. Additionally, laws are not written off because one thing proves false, only the clause is.

[Tayyyyyyy]

Okay, I see...on expenses you could interpret the expenses of 2 crew being more than 1 window, but if it was a tie you'd have to break it in favor of the window. But on the new proposed lawset you'd always have to pick the window if you couldn't save both.

But...we could call the new lawset crewtilitarianism or something! 😄

[Shadeykins]

After actually seeing your interpretation of the default corporate lawset, I thoroughly disagree with your changes here. You've maliciously and deliberately interpreted the default corporate laws in such a way as to suggest that Corporate says an AI should be hellbent on the destruction of Nanotrasen.

Your current changes shift the corporate lawset from being a game of economic calculus to being a benevolent protector of the station. We already have other lawsets that achieve exactly this.

[Tayyyyyyy]

After actually seeing your interpretation of the default corporate lawset, I thoroughly disagree with your changes here. You've maliciously and deliberately interpreted the default corporate laws in such a way as to suggest that Corporate says an AI should be hellbent on the destruction of Nanotrasen.

This was a hypothetical example of why Corporate is a problematic lawset. The lawset change here still has the same issue. This is more a philosophical issue than something that would affect the game. A corporate AI cannot take down Nanotrasen in 2 hours.

And I've said this before, I'm not maliciously interpreting the lawset. If you read what @IK3I has written, you can see that it takes some degree of mental gymnastics to get around the fact that corporate tells you that you cannot allow the replacement of station equipment. I disagree with those gymnastics. This revised version makes those gymnastics unnecessary. It leaves the lawset entirely in place, but makes it easier to understand.

Your current changes shift the corporate lawset from being a game of economic calculus to being a benevolent protector of the station. We already have other lawsets that achieve exactly this.

There is still economic calculus here. In fact, this revised lawset opens up the lawset to MORE economic calculus, since costs are no longer tied to the replacement of things, but rather to costs in general.

[IK3I]

You could potentially fix this whole issue with a new law above the current law four.

1. You are expensive to replace.
2. The station and its equipment is expensive to replace.
3. The crew is expensive to replace.
4. Replace lost crew and equipment.
5. Minimize expenses.

This would make the same end result without resorting to bizarre wording and provides some potentially interesting lines of play.

[Tayyyyyyy]

That's...not a bad idea at all, actually.

[Zciwomad]

Simple "reword" is not enough, becouse it still leaves room for interpretation that is against rules, or will lead to be called an "dick AI" interpretation. What is an "dick AI" in eyes of the crew/(players)? If AI don't want to do something that crew wants and its not against Server rules and Space law. Every single player that don't play synthetic regularly won't agree to standard Lawset, or lawset interpretation that could allow AI to not listen them and even interfere with their actions.

If frist three laws from "Corporate" actually matter, then they are there to define what? What is expensive? Everything is expensive, even part of garbage in disposals (books, box, everything reusable). Mayby they are there to define what is more expensive. In this "classic" situation with fox and Captain in danger, where AI/borg can only save one, law about station eq is above law about crew, so even one fox is more expensive than all crew on the station. If they are there to define what is expensive, then they can't contradict with eachother, so they are really like one law: "Station, station equipment, crew and you are expensive". In this case, fox is less expensive without doubts.
Law 5: "Minimize expenses" doesn't mean: "Minimize any expenses", or "unnecessary expenses". Just expenses, so AI could minimize two(plural, becouse "expenses", not "expense") expenses out of five (save two wounded crewmembers from room with breach and let the others die). Expenses were minimized, not all, but some = Check!

Law 4 will prevent attempts to destroy station and crew, and its good becouse "crew" is created by all crewmembers, so AI can't just replace one crewmember. "Equipment" is uncountable and just like crew, it is created by all devices/computers/etc on station. On the other hand, this force AI to replace any equipment, like welders in vending machines, or firesuits/gas masks in firefighter closets all over the station. Mayby just add "essential" before "equipment".

[IK3I]

@Zciwomad I'm not sure what has caused your outburst on the forums in regards to laws, but we do our best to keep things civil on github. Show some respect and think about what you're saying before posting.

Now as for the actual issues raised. If an AI player decides to act like an ass, it'll get treated like one by crew. If it starts genuinely self-antagging, that's another matter entirely and becomes an issue for admins to deal with.

As for corporate. You may need to take a class on formal logic or programming to really grasp the impact they have, but a basic explanation is that they define what things are expensive. A syndicate borg for instance is not expensive and can thus be ignored until it proves malicious as it starts causing expenses when it breaks vital station structures or crew.
Defining how expensive things are in relation to each other is up to the AI. A most correct method is left as an exercise to the player. I tend to think in terms of the chain of production. If something requires an object or person to replace it, it is of lesser value than the thing doing the replacing. Defining hard numbers would require an actual AI to calculate opportunity costs.

It is the correct interpretation to consider all laws of equal precedent until a conflict arises. Consider them all and act accordingly, choosing higher priority laws only when you can not satisfy all laws.

As for statements. The Fox and Captain situation as it is generally used is fallacious in the same way that asking if A > B and saying that because A was declared first, it wins ties. Conflicts of governing laws can not logically be resolved by deferring to the order of definitions. You may resolve the conflict that way if you so choose, but it is the functional equivalent of a program defaulting to the first answer after timing out trying to calculate too complex answer.

In regards to expenses. The AI has no concept of expense beyond the laws given. It is a variable to be manipulated. Thus, to minimize it is to reduce all forms to their lowest possible state because the AI only has a single number that everything contributes to. Saving 2 out of 5 is better than saving 0 or 1. It should try to save more if possible to reduce replacement costs. However, if it determines it can not save the other three without generating additional expenses beyond their replacement cost, it should abandon them. In other words, it should minimize expense by maximizing survival.

The proposed law four exists for exactly the reason you just stated. It's good for RP to give an AI something to play with. It closes out a griefy window by opening up a new silly window where an AI can experiment with new gimmicks. You keep breaking the chairs? Fine, you get stools.

[taukausanake]

I like the addition of the new law 4, even if it sounds a little weird. The way I read it the first three laws are statements and the last two are actual laws that act on the statements (it helps that I'm taking Discrete Mathematics right now). You can argue that statements don't have precedence because they are stating facts (as they would be to an AI) but the laws will have precedence in order of replacing what is lost and then minimize expenses that follow replacement. The fun things may come where law 5 might override law 4 if replacing something completely is too expensive. I'd like to try playing that and see how it goes

[Zciwomad]

Mayby I'm reacting a little bit too emotional.
Generally, I don't have problems with those laws and I'm able to easy interpret them in "good" valid way, even if not, I'm only playing borg so I left AI to decide, or I just go with my own decision as masterless borg.
My main problem is with other borgs and AI and this is why I started topic about interpretation (I feel like suggestion topic came a little too off rails). Going too much into the future and "overthinking" laws is not fun. Example (crewsimov and start of a shift, code green):

• Borg, bolt down two doors out of three in main hallway.
• I can't.
• Why?
• Security officers need to react fast to any threat and bolting two doors will slow them and this can cause harm.
  What I suggested in the first place is to add a few lines in wiki page about borgs and a line in Rule 9 about it. Becouse this way, as a borg on crewsimov players can ignore law 2, becouse when they do something for crewmember, this draws attention from something they could do, like patroling the station and searching for possible threats. It's strange that borg who said that, don't ejected nuclear device into a space using disposals, becouse somebody in the future my use it to destroy station and this could cause harm.

[DarkPyrolord]

This doesn't really fix any of the larger issues with corporate, going to have to be a 👎 from me.