AI Diplomacy, not so bad

[dexters]

And yea, generally, attacking a protected City State smack dab in the middle of the map or is an ally of an empire as a kind of 'soft attack' of your enemy, is not a good idea.

This is probably less ridiculous than the vassals exploit in Civ4 where it's almost always gauranteed you can make bank attacking someone's vassal first instead of engaging them directly, since vassals will often put up a pitiful defense and the master civ would usually have a delayed response.

Yeah, if I ever tried at a serious mod for this game, it'd start with AI giving much more information about what it likes\hates about you and other AI (while still allowing for deception).

Making infoaddict's features standard would be a good start. It has a Civ3/4 style 'web' of relations that I absolutely find is the best way to view relations. I hate the lists in the current diplo UI.

I'd add that while I blame player failure generally on a lot of complaints about AI issues, I think Firaxis is guilty of creating the issue by having a sub-par diplo UI; heck we don't even get graphs until replay, even though the game obviously keep tracks of all those stats.

Also I'd go a step further from what is presented on info addict as I don't believe the diplomacy screen shows protected city states by Civ. There should be a list of each civ, the city states they have relations with and all the city states they have pledged to protect.

And either force a pop up (optional) listing civs you would likely to anger if you attack/dow on a CS right before you do it. That way, there's no way players can feign ignorance.

 

[dexters]

Well, he's talking about a Prince/Archipelago game, and Archipelago maps are famous for being love-fests. I'm not trying to be argumentative, and you are absolutely right that the soft-coded flavors dictate who is more likely to start wars. But, the higher up you go, the more likely flavors don't matter and everyone will declare war on everyone else, regardless of the grand strategy each civ rolled. A Deity/Pangaea game often degrades into a total war game early and stays that way until the end. I'm not saying this is bad or good it's just my experience (and maybe I'm doing it wrong).

What would you suggest as a fix? I don't think the devs should be balancing a game on diety either.

And IIRC, AI's in previous Civ games on Diety or Sid in Civ3's game aren't much better and I highly doubt narrative players are in the diety league. Most play prince++ (prince to Emperor) for the right amount of challenge, with slack to story telling. And I'm defining narrative players as those who ascribe stories/intent/grudges and grand arching geopolitical narratives to their play.

Unless you mean narrative players as in the brag posts about OCC on diety and the tales of doing so. If that's the case it's been done with Civ5 too.

The wars, by the way, aren't the primary setback to narrative play for me either. It's the fact that if you play for a challenge and also want a little engaging diplomacy, you're barking up the wrong tree. There are very few positive modifiers that will actually put you in a position to ally another civ & even if you could there would be no point to it. DoF is the closest relationship you can have with any civ AFAIK. That relationship gets you very little (50 turns of peace?). You're "friend" will watch you get nearly obliterated by two other civs in the ancient era, ask you for a resource they could easily trade for fairly in the middle ages, and denounce you in Renaissance because you built Big Ben. The AI only values winning.

I am in agreement about adding more positive modifiers. Which Is why I made this post on the suggestions forums, and a facsimile post on the 2K forums.

The encouraging thing is that they are moving in that direction. The July patch added an additional modifier. If the AI gets a good deal, out of a trade or a peace treaty (they offered 10 cities, you took only 2) you'll likely max out on -30 modifier

In Firaxis fashion, the UI tooltip on the diplomacy page says 'we have traded recently' which is a misnomer. It should really say 'you have bribed me recently'.

I'm not sure how it decays, but apparently, the bonus does decay over time.

Another aspect I think that can be worked out is a lot of grievances appear to have no decay values, so that means grand strategic grievances (wonder spamming, covets your land, competing for same city states, warmonger) tend to build and build. There should be certain conditions where they decay, or there should be a small amout of decay to keep enmity in check. Here's to hoping they add more modifiers in future patches.

I appreciate your perspective regarding previous Civ games (because I never played them) and I can see your point regarding player manipulation of the AI with regard to powerful positive modifiers such as religion. I also accept that this might be considered the best diplomacy model to keep the player engaged in conflict and active while pursuing victory. I just wonder if there could be exceptions without compromising the competitive concept. Sure, an AI that truly allied the player might be putting itself in a position to ignore the fact that the player is outpacing the AI, but in cases such as liberation, should that matter? That civ has already lost the game once. I don't like the fact that they take the competitive spirit to the point of suicide either. Shouldn't the AI value survival and seek out allies that would prevent them from being destroyed? Sure, this takes the nasty competitive edge off a few of the civs in every game. But would it change the outcome of those games?

Sorry I've rambled.

Thanks for your note, I'm generally not the first one to say 'diplomacy is perfect' but when the alternative point of view is 'diplomacy is broken' followed up by trolls who appear to not have played the game in months making generalized comments (not in this thread mind you but it's happened fairly recently in in others); it's often easier to present the opposing view rather than a nuanced view. I actually appreciate this rather productive back and forth with players who clearly want to talk about the way diplomacy works in this game, it brings out the nuanced perspectives.

 

[CoolLizy]

Covet Lands:
"Turn 121, settled New York, disputed territory"

Speaking of which, I'd like it if the game would ask me "Are you sure you want to settle here? You told *insert name here* you wouldn't settle by them," so that I wouldn't have to guess whether or not I'm too close.

 

[nokmirt]

Catherine's extremely devious and nearly always backstabs her neighbors once she runs out of room to expand. Any time we start near each other, I know I have to build up my military to make other civs more attractive targets once she starts lusting after everyone else's territory.

Funny you regard Catherine as a particular AI! God that is funny. The AI does not knows its Catherine, it does not know and so it does not care. It is simply upset at close enemy expansion near its border. That is all. It does not understand the benefit of a good close ally. It does not know how to trust you see. Civ leaders in real life were human beings, not machines. Firaxis should help the AI think about the usefulness of having beneficial friends, and not automatically hate, any transgressor without looking at the situation first. It should need a real reason to feel threatened.

They should program the leaders to act a bit more like real civ leaders. Instead, they all do the same old boring things. None different than any other. They all end up hating you and each other. Some numbers account for more aggression etc. But really they act all the same. I have never played a game where they have not yet.

Keep in mind that the AI code used in this game is leftover from Civ 4, a game of far different parameters. They should code a unique AI for CiV alone, then perhaps many problems would be solved. Civ 4 and CiV are not very interchangeable. You can perhaps fit many of the same ideas into 5, but you have to do it in a unique way built for that partcular game. I believe this is where their problem lies. Essentially they were afraid to spend money to make money by building a real AI, it was a bad business venture from the start. Either they are morons or they simply just don't care. Either way the whole thing is really pathetic.

 

[dexters]

Speaking of which, I'd like it if the game would ask me "Are you sure you want to settle here? You told *insert name here* you wouldn't settle by them," so that I wouldn't have to guess whether or not I'm too close.

Yeah that's the kind of hand holding that you don't need. Settlement stealing should be fairly self explanatory.

Also covet land =! settlement penalty.

They will still covet land even if they are far away. The settlement penalty is specific to spots on the map your neighbours want to settle but you steal/take first.

The former is just a general modifier to say the AI likes your territory/want your cities.

The latter is to stop early rexing exploits that have the human player settling directly in the AI's direction to cut them off early.

 

[King Patrick]

Funny you regard Catherine as a particular AI! God that is funny. The AI does not knows its Catherine, it does not know and so it does not care. It is simply upset at close enemy expansion near its border. That is all. It does not understand the benefit of a good close ally. It does not know how to trust you see. Civ leaders in real life were human beings, not machines. Firaxis should help the AI think about the usefulness of having beneficial friends, and not automatically hate, any transgressor without looking at the situation first. It should need a real reason to feel threatened.

They should program the leaders to act a bit more like real civ leaders. Instead, they all do the same old boring things. None different than any other. They all end up hating you and each other. Some numbers account for more aggression etc. But really they act all the same. I have never played a game where they have not yet.

Keep in mind that the AI code used in this game is leftover from Civ 4, a game of far different parameters. They should code a unique AI for CiV alone, then perhaps many problems would be solved. Civ 4 and CiV are not very interchangeable. You can perhaps fit many of the same ideas into 5, but you have to do it in a unique way built for that partcular game. I believe this is where their problem lies. Essentially they were afraid to spend money to make money by building a real AI, it was a bad business venture from the start. Either they are morons or they simply just don't care. Either way the whole thing is really pathetic.

This doesn't match up with my experience of playing Civ5 at all. Are you sure you're playing the right game?

 

[nokmirt]

This doesn't match up with my experience of playing Civ5 at all. Are you sure you're playing the right game?

Confucius say, "Don't mention it if your not sure." So yeah, I would not have specified the game if I was not playing it. We just see it different thats all. Maybe you can elaborate about how you do see it. I love every viewpoint. I may learn, that I am wrong in the way I think about it.

 

[King Patrick]

Confucius say, "Don't mention it if your not sure." So yeah, I would not have specified the game if I was not playing it. We just see it different thats all. Maybe you can elaborate about how you do see it. I love every viewpoint. I may learn, that I am wrong in the way I think about it.

I do see some personality in how the AIs behave. The differences are pretty slight in a lot of cases, because they are all operating off the same basic code and with the same palette of choices. So, for example, every civ will attack you if they sense weakness & easy pickings. You will almost always have at least some tension with your neighbors, esp. on smaller maps, because you're all competing for the same resources & territory.

But within those limits, you can see some trends in the individual leaders - like how Catherine will always, always, always backstab you while pretending to be your friend. Bismarck will always be at war with everyone before too long. Montezuma will almost always provoke an alliance of everyone else against him by the midgame. Isabella can be a loyal friend if you're not sharing the same continent, but she will constantly badger you to join her in a war on somebody or other - someone once called her 'your bloodthirsty kid sister.' And so forth.

I know that this is just a semblance of personality - it's nothing more than algorithms and different sets of numbers for aggressiveness, deceptiveness, and so forth for each leader. Plus animations & voice acting. But that's true of any computer game. The important thing is that I do feel I know what to expect from the various leaders, and that a game with Monty as a neighbor is going to be somewhat different from one with Gandhi as a neighbor.

 

[nokmirt]

I do see some personality in how the AIs behave. The differences are pretty slight in a lot of cases, because they are all operating off the same basic code and with the same palette of choices. So, for example, every civ will attack you if they sense weakness & easy pickings. You will almost always have at least some tension with your neighbors, esp. on smaller maps, because you're all competing for the same resources & territory.

But within those limits, you can see some trends in the individual leaders - like how Catherine will always, always, always backstab you while pretending to be your friend. Bismarck will always be at war with everyone before too long. Montezuma will almost always provoke an alliance of everyone else against him by the midgame. Isabella can be a loyal friend if you're not sharing the same continent, but she will constantly badger you to join her in a war on somebody or other - someone once called her 'your bloodthirsty kid sister.' And so forth.

I know that this is just a semblance of personality - it's nothing more than algorithms and different sets of numbers for aggressiveness, deceptiveness, and so forth for each leader. Plus animations & voice acting. But that's true of any computer game. The important thing is that I do feel I know what to expect from the various leaders, and that a game with Monty as a neighbor is going to be somewhat different from one with Gandhi as a neighbor.

I understand where your coming from. Perhaps I am thinking about it way too much. The purpose of the game is really to have fun, and I do have fun when I play. It is never going to be perfect, and the designers have made it better in my book.

You know, I never did try a game with random personalities. I want to try that and see how it goes. It would be funny seeing Ghandi trying to be a ball busting warmonger.

 

[Glassmage]

He already is capable of being bloodthirsty with his normal flavors. Once after getting nukes and complete Manhattan Project, he comes to my face and say "our threats are backed with nuclear power" or something and then I quit the game.....

 

[qemist]

I think you mean even the civ which asked you to go to war will consider you a warmonger after, and that's indeed bad. I wouldn't hesitate calling that a bug. All other civs may still dislike you, and should, but not the civ whose request you answered.

I was once of that mind, but now I think the situation is more nuanced. If it was a war of aggression, then it is reasonable for the asker to make a note to self, "This guy's a bit of a warmonger like me, better keep an eye on him". It doesn't make any sense for him to promptly denounce the party that answered his request, which sometimes happens. This could be modelled by defering the warmonger penalty wrt the asker a certain number of turns, or until the war is over, or balancing it with a positive modifier that decays faster than the warmonger penalty.

Come to think of it, these requests are always aggressive. The AI never asks for help when it is the victim. If this changed and the war was defensive (i.e. the asker had been attacked) then the effect of responding to their request for help should be only positive wrt the asker.

 

[King Patrick]

I find that often what gets left out of these 'Gandhi asked me to attack Rome and I did and now he's calling me a warmonger' tales of woe is 'so I conquered Rome's 12 cities and tripled the size of my empire and now my enormous military is right on Gandhi's doorstep.'

 

[The Pilgrim]

I was once of that mind, but now I think the situation is more nuanced. If it was a war of aggression, then it is reasonable for the asker to make a note to self, "This guy's a bit of a warmonger like me, better keep an eye on him". It doesn't make any sense for him to promptly denounce the party that answered his request, which sometimes happens. This could be modelled by defering the warmonger penalty wrt the asker a certain number of turns, or until the war is over, or balancing it with a positive modifier that decays faster than the warmonger penalty.

True. However... and it's a biggie... If the asker is 'testing' you this way then your refusal should give you positive modifiers. That doesn't happen. What happens is AI 'makes notes' for this and for that and somewhere in the mid-late game summarizes them. And what does it see? "Oh no, the human player tries to win! I need to declare him asap!" Seriously, it doesn't take an Einstein to know human tries to win. But 'making notes' in that matter eliminates diplomacy as total. Since every action of yours will bring AI closer to 'realizing' you're trying to win and actually doing something about it. Meaning the diplomacy has no impact on the gameplay and relations are going to deteriorate linearly regardless what you do. Does it sound reasonable? What do we have the diplomacy for in the first place? Just to say we have it?

I find that often what gets left out of these 'Gandhi asked me to attack Rome and I did and now he's calling me a warmonger' tales of woe is 'so I conquered Rome's 12 cities and tripled the size of my empire and now my enormous military is right on Gandhi's doorstep.'

So what human should do is to make a phony DoW just as AI does and not to send a single unit towards Rome. This is how everybody will be happy. Finally we've got the right way to play Civ5! C'mon people, you're kidding me... How on earth anyone can advocate such a system? That's a real mystery...

 

[dexters]

Defensive wars turning into an orgy of expansion has always been the preferred method for humans to expand without incurring any penalties like broken trades or 'dow' penalty, because Civ4 used to only track who DoW on who and Civ3 had a punitive reputation system on broken trades so it's always better to not be the one breaking trades and let the AI do it instead.

The current model is more realistic, but I'm not against toning it down a little or adding a decay mechanism. Granted there may already be a decay mechanism on warmonger penalties as time passes, but it's not obvious.

That said phony wars are probably not the right solution either if asked to join in a military alliance. The AI does keep track of your contribution, vaguely noted by Firaxis as based on 'damage done'. I assume this is aggregate of unit hp damage dealt, and may include cities. So accepting a bribe to go to war and not doing anything will likely incur a diplomatic penalty as well.

 

[King Patrick]

So what human should do is to make a phony DoW just as AI does and not to send a single unit towards Rome. This is how everybody will be happy. Finally we've got the right way to play Civ5! C'mon people, you're kidding me... How on earth anyone can advocate such a system? That's a real mystery...

Fallacy of the excluded middle. Fail.

 

[The Pilgrim]

Fallacy of the excluded middle. Fail.

Is there a middle?

 

[King Patrick]

Is there a middle?

Between 'do nothing at all' and 'conquer their entire empire?' I would think so.

 

[oopsy poopsy]

Has anyone tried to just kill units and raze improvements without ever taking cities to try and get a take-it-all-please peace deal? Do you receive warmongering conquering points against if the cities come through a peace trade?

How about if you knock the capitol down to redline and surround it, but don't take it? Never tried anything like this, but I might give it a shot.

More generally, you can create a nuanced narrative to explain their behaviours, even one based on the underlying numbers of the diplo mechanics, but it still amounts to them acting like a bunch of 'tards, especially for a casual civ player. Diplo in civ 5 is simply ricockulous, still is after patches, and just because it's explainable or understandable doesn't mean it's a good game design.

 

[The Pilgrim]

Between 'do nothing at all' and 'conquer their entire empire?' I would think so.

Well, you think wrong. There is a difference between DoWing AI and getting some advantage from doing so and DoWing and getting nothing. It's kinda black and white. If you've managed to benefit from war, you're bastard. Why to ask in the first place then?

This is a magic cycle. We expect AI acting like a reasonable human, AI, on the other hand, 'treats' us as another AI. Expects us to build tons of pikes, move them back and forth and fail to capture a single city. By AI's terms it's a well played war. We'll never understand each other.

 

[CoolLizy]

Yeah that's the kind of hand holding that you don't need. Settlement stealing should be fairly self explanatory.

Also covet land =! settlement penalty.

They will still covet land even if they are far away. The settlement penalty is specific to spots on the map your neighbours want to settle but you steal/take first.

The former is just a general modifier to say the AI likes your territory/want your cities.

The latter is to stop early rexing exploits that have the human player settling directly in the AI's direction to cut them off early.

It's not hand-holding, it's making a broken game mechanic actually work the way it was intended. If Napoleon comes to me whining about how I'm building too close to him and I tell him I won't settle any more cities near him, then he shouldn't come back later throwing a fit about "breaking my promise" when I build another city on the other side of my territory that he still considers too close even though it's really nowhere near him.

Giving the player more information when making a decision is not hand-holding; unless showing the player how much gold they have when negotiating with the AI is also hand-holding.