Here's a quote from another thread:
I originally planned to respond to it in that thread, but I soon realized that it is such a huge topic that it warrants its own discussion thread.
Diplomacy is one area that has been severely neglected in all of gaming. The only game I've ever played that had good diplomacy was an MMO called Shadowbane (and the diplomacy occurred between real players and involved the real consequence of losing your city).
Metrics are totally the wrong way of doing diplomacy. You can't quantify a real person's opinions or a culture's customs with some silly numbers. You can't quantify hidden agendas or secret agreements.
A few key rules should be followed when designing the diplomacy system of a game. The AI-controlled players should:
All of these rules have been broken by games in the past, to their detriment.
Obviously, the design of a diplomacy system is much more complicated than following a few rules. There must also be a strong strategic foundation for the AI which allows it to formulate realistic goals and take the necessary steps to achieve them. AIs should be able to "see through" a player's gestures towards them and find weaknesses to exploit.
If a player is weaker militarily than the AI but stronger economically, it should not allow the player to bribe it into abandoning its plans for invasion. Rather, it should attempt to suck as much gold, mana, spells and other bribes out of the player that it can to further bolster its position.
This type of system is very hard to design, I admit, but I believe it is worth trying. I hope that the scripts and routines used for the AI are open to the modding community so that modders can fix any glaring mistakes or loopholes in the diplomatic AI.
What are your suggestions to improve the diplomacy of this game and elevate it over the mistakes of the past?
The issue is that the AI is tasked with being the game the human plays. Its meant to be an experience, a challenging one hopefully, but more importantly a fun one. The AI randomly obliterateing you simply because it wants to win isn't actually as much fun as youd think. A game thats too easy can still be fun, why do you think so many people use cheats? A love a challenge, and haven't cheated in a "real" game in 10 years, but if the games goal is to ruin everything I do and press me all the time, I probably wont have a good time. Many of the "hardcore" gamers out there wouldn't in reality either, they just like the idea of a challengeing but winnable fight.
What you want the AI to do is unlikely (have it be sneaky like a human), "unfair" ("WTF this game is sh!t, the AI makes alliances and breaks em the second I leave the border??!!?"), un-fun ("I lost another town, without a fight, to my backstabbing nieghbors. Yay."), and will lead to painful meta-gameing ("The AI never takes alliances seriusly, so I start the game at war and don't make peace until there isn't anyone left").
Your ideas would be great for a military sim, or maybe a future game where the AI can realistically cloak-and-dagger you, but what your talking about would ultimately be a major drag. Maybe im wrong, but past experience says otherwise.
This goes back to my original post, where I listed several key rules for the AI to follow, one of which is:
It's one thing to have an AI randomly stab you in the back after you just signed a momentous trade agreement, it's quite another for the AI to do so after having determined that you are becoming too powerful and threatening to them.
I think it's pretty obvious where I stand here.
I have no problem with an AI that stabs me in the back, in fact that makes the game fun. My problem is with your trading idea, and that alone. A trading system that doesn't allow me to know that when I accept a specific trade it's going to actually happen is fundamentally broken and will never gain widespread acceptance from gamers.
But if the AI signs an alliance with me with the goal of trying to get me to engage another enemy, only to turn around four turns later and attack me (which was its actual intent all along)? Great! The hard level AI should play to win. On that much we agree just fine.
But it would still seem random to the player, and in that case what you get is a steamroller effect; the more powerful a player becomes the more likely he is to be attacked, be forced to conquer the enemy, and become more powerful. The infinite loop this becomes starts to get old after a game or two.
I see what your trying to say, and where you stand, but to make an AI think like that is impossible pretty much anyway, as I showed you in many arguements prior. The other thing is, again, if you want it to be a simple "beat the human because he is more powerful (which he almost always will be) then you have a repetitive wargame that actually goes against a diplomatic model.
I agree with you that an AI that always tries to stab you in the back will make going for a diplomatic victory meaningless. But I do agree with Tridus and ChongLi that the AI should play to win.
Just because an AI plays to win, and has a chance to stab you in the back, this doesn't mean that he will do it. If you have a good relation with the AI, or the AI thinks that you are already too powerfull to stab you in the back without bad consequences, the AI should go for a joined diplomatic victory.
I would even say, if the AI proposed the alliance, and you two are the only ones left, the AI has won the game, not the player because the AI 'started' the diplomatic victory. This could result in some though choices for the player. You and an AI are fighting two other AI's and are winning. If your friend can get the other two AI's to surrender and join the alliance, he wins. So, what do you do, leave the alliance and hope you are strong enough to survive what may become a three on one war, or stay in the alliance for a bit longer until the other two are even weaker. Or you could even try to make a new alliance with the other two and stab your AI friend in the back...
If the AI refuses to attack you because you're too strong, you also get a steamroller effect. Namely that since the AI is afraid of you, you can pick them off one by one at your liesure and the other ones won't do anything about it.
Actually, that's more of an anti-steamroller effect. The more powerful you get, the more the AI pushes back against you to attempt to slow down your momentum.
But the AI's push is likely to simply lead to endless war. You will either A: Perish at the hands of your status quo loving nieghbors or B: tear through them, becoming larger and thus a greater threat to more people.
Even if what you say is right, a game where backstabbing is common and war is a daily occurance would get boring. Didn't you start the thread wanting better diplomacy? Now you seem to be more interested in an AI balance mechanism. Pfeh. Sorry, but the human player will always be better than the AI unless it cheats, hence the AI in such a system would be in perpetual war with the player. Thats not diplomacy.
indeed. So what do you propose?
It depends on what chance the AI has to win. If the AI's only chance to win is to defeat you in a war, what else do you expect it to do?
If the AI has a chance to win without war, then by all means, let it go ahead and do that.
If you read what I wrote earlier...
If something like this gets implemented, the AI can win without a war. If the player will like this is another question off course...
Also, Stardock is implementing a Quest victory where you can win the game by completing some difficult questline. So that's another way the AI can win without war. Imagine an AI offering / accepting an alliance with you so he can spend more time on the quest without having to worry too much about a war with the player.
The AI should play to win, but this should not mean that every AI is a war mongering, back stabbing, player hating AI...
Took the words right out of my mouth. Totally agree!
It's sad that some people don't want a truly competetive A.I. They simply want the best possible A.I that plays along with what the player wants....
I don't know if I fall into that group in your oppinion, but I would rather think of my stance as being more of a "realistic implementation" stance. I WOULD LOOOOVVVVEEE the game you guys are proposing. When I was younger, I wrote long complex rants that sound exactly like you own on the civ4 suggestion forum. However, experience is a bitter teacher, and you quickly learn what you wan't isn't always as fantastic as it sounds.
Most of the ideas I've given can easily be incorporated into the difficulty setting of the game. Personally, I want to play the hardest game possible without having the AI outright given freebies, knowledge and capabilities the player does not have.
I could deal with this if you simply would realize the limitations to the AI. First of all, you aren't playing a seperate program running another copy of the game, you are playing the game itself. It knows things by virtue of it needing to know those things to function anywhere near coherently. If it knew nothing of the world around it, like the human, would you have it explore randomly? That sounds decent, but it simply won't grasp the idea of exploring without a little knowledge as to how to explore. Its hard to explain, but I hope you see what I mean.
Those "freebies" do suck, and the idea of the AI having abilities the human doesn't is sickening, but the AI does need bonuses on higher levels. Extra tech, better start position, cheaper stuff, etc is all fine with me so long as it gives a decent and fun challenge.
You appear to want to press the "hard AI" button and be playing some form of cyborg mind. You also want an untrustworthy, backstabbing AI.
I want a good AI that won't let me roll over it, works with me when it makes sense, and yet doesn't wait in the shadows to cripple me. Perhaps im reading you wrong, but in one breath you want an AI that doesn't focus on the player and is realistic, then in the next you want an AI that works to obliterate the human at every opportunity...?
That sounds neat, though id like to know where you heard it. I also agree that without multiple ways to win, the game will become a war sim and not a fantasy sim. Yet, I still believe the AI must be programmed with the idea that war is innevitable, as it is probably the most common victory type.
So the player will either A: Lose or B: Win and win even more....ok, and what's the actual problem here..?
Even if what you say is right, a game where backstabbing is common and war is a daily occurance would get boring.
Nonsense! It would keep you on your toes at all times and force you to check what your neighbors are doing just like in a MP game!
And are you saying that constantly being at war is boring..? This entire GAME is about WAR! It's about getting a basic economy and production started so you can churn out some troops and go capture that lair/node/shard to get more powerful so you can cast better spells, build better factories/farms/libraries so you can get even better troops and capture even MORE powerful nodes/lairs/shards so you get even MORE powerful magic, troops'n'stuff to eradicate your opponents!
Isn't the point of every single game (atleast TBS?) to destroy your enemies..?
A good article I saw on game AI design was that when people make an AI that is truly ruthless, a lot of players rebel. Players have a tendancy to want an AI that "feels" like a challenge, but that they can beat.
Difficulty settings are a good way of dealing with this. The "beginner" AI should be that type of AI. It's playing to present some difficulties to the player, but not really to beat them. The higher end difficulty should be more ruthless and try to win.
If they incorporate different AI "personalities" into the game, that could factor in too. Some personalities could be more open to diplomacy then others. A peace-loving AI is going to be friendlier with people then a warmonger, who would only ally with anybody if it's out of convenience (common enemy or to backstab later).
There was something said around here at some point about a lengthy quest chain that could win the game if completed. I don't know if it was a suggestion or something that is actually going in, but it's been discussed before.
Well, it says so right on the first page of this website...
So it's something that's planned for the game, if they succeed with this in a fun way is still something to wait for off course...
I would even say, if the AI proposed the alliance, and you two are the only ones left, the AI has won the game, not the player because the AI 'started' the diplomatic victory.
Why would anyone (including an AI) ever agree to lose the game? Diplomatic victory in past TBS games involved electing a ruler. If the AI hopes to win a diplomatic victory, it better be able to command the majority of votes.
Personally, I think diplomatic victories are stupid. In real life, elections never solve anything, there is always another election around the corner. Who actually believes there could be an election to create one world government and that it would somehow result in world peace?
Modern games (almost every single multiplayer game) are designed with a client/server model. The server contains all of the details of the world and the clients connect to it and play through it. Even when you are playing by yourself, you are still connecting to a server. AI players are clients as well, each of them connecting to the server the same way you do. This means the AI players get no special knowledge/data advantages at all.
The server may have difficulty settings, but they act more like handicaps in that they give bonuses (money / food / research / production / happiness) to some players and penalties (in the same areas) to other players. Whom these bonuses and penalties apply to is determined by the difficulty slider (easy tends to give the bonuses to humans whereas hard gives them to AIs). Ideally, the AI should be good enough to play the game at normal (no penalties or bonuses for anybody) and actually be competitive with an average human.
I would be very surprised (shocked even) if SD goes back and designs this game the old monolithic way, especially considering that they are planning to include multiplayer (a client/server model makes multiplayer a lot easier to implement, since it is basically incorporated into the design from the very beginning).
I've always loved AI personalities. MoM had em, Civ has em. They're what truly brings variety and unpredictability to the game. I think the most important part of AI personalities is that they are random each game and that they try to keep their personality a secret. I've always thought it was pretty cheesy in Civ that you knew Monty was the aggressive one.
You guys should stop complaining about an ai being honest and just ask for a role playing trigger. If it's optional, people that want the goody bastards to actually behave like goody bastards won't have to worry about being stabbed in the back from a civilization run by nuns.
Those that abhor behaving like civilized individuals can leave it off and have every AI seek only the best possible option, as opposed to using flavorful things like honesty and bias in the calculations. Myself, when I'm playing a good guy, I try not to be a dick. It's a challenge in itself to behave and only act against threats. When I'm a badguy, I stab my best buddy in the back whether it's prudent or not, just because it feels like the proper thing to do. AI is so much more interesting when you behave in a realistic fashion.
No one has ever managed to conquer the world by being a cut throat conqueror that didn't keep their word. That type is only effective for a short period of time before getting the shit kicked out of them by everyone else.
I am not sure if the AI can even have an agenda apar from meeting a victory condition. Let us suppose that that is the case, that is the AI can have an agenda in the sense that it wants to conquer another one's capitol city.
Now suppose the AI will not reveal any info that will not further his agenda... What info would it release to you? Keep in mind that not releasing any info can also be a telltale sign that the AI is up to something. Ideally you will want the AI to act as if it is not up to anything and hen have it strike when you do not expect it. I highly doubt the rules in the OP will cover this nicely.
Diplomacy in a game like this is extremely important because I do not want to actually wage war to win the game all the time. Waging war in games like these can be fun, but nearer to the end of the game you know you will win and then you will not want to go through the trouble of mopping up some chanceless AI.
Earlier in the game what I would really want to see is not just an AI that will offer a challenge, but I want to be able to forge alliances in order to defend against a common foe, I want to pool spells with it if it is mutually beneficial, and I want to be anle to somewhat rely on an AI being my friend and Help me in time of need. What I find sorely lacking in most games if the option to actually work together with the AI. Of course there needs to be some balance to this, but if Stardock can get this right the game will truly have the potential to be amazing.
Your definition of agenda is different than mine. You cannot just say "I want to meet the conquest victory condition", you have to have a plan of how to accomplish that. That plan is your agenda.
Everything you do, from the units you train, buildings you construct, spells you research and cast to the diplomatic choices you make are part of that. In most typical games, your agenda starts off only vaguely formed and develops and evolves as the game progresses and you begin to get a feel for how things are going.
Designing an AI that has an agenda (rather than one that just moves at random) is important to a competitive game. Every AI opponent in every game you play should have a completely different agenda, unless you are somehow playing on the exact same maps with the exact same starting conditions and playing the exact same way.
As an example, suppose you discover some adamantium in your territory. The AI notices this and adds "get adamantium from player X" to its agenda. This should result in a change in the AI's behaviour, causing it to try multiple strategies to obtain this adamantium from you, including diplomatic and war strategies.
OK this actually sounds very cool. This was also my idea of an agenda, even if I may hae worded it wrong.
I just fear that even an AI obeying the rules in the OP will give things away that it should not. Remember that not wanting to disclose info can also be a dead giveaway of it's intentions. In the end you learn to adapt to that and you will learn what to expect in different situations.
In a game lik Civ IV diplomacy was handled with random factors and arbitrary dice rolls. The difficulties there were not governed by the obscure ways of the AI to handle diplomacy. The system was in fact crystal clear. It was up to the player to understand the system and then learn what the consequences of different actions would be. The difficulties of handling diplomacy in Civ IV was then just juggling with pro's and con's of different decision, choosing the option that was the least bad and preparing yourself for the bad onsequences.
In the end, diplomcy in CIv was all risk management. This was a crystal clear system that worked, and if you wanted to have an unpredictable factor you had an option to play with random personalities. In the end, the Ai would seem like it pursued it's own agenda.
I would very much be in favor of a system that was transparant to a degree. Juggling actively with diplomatic hardships wasso much fun in Civ IV that I would not mind seeing that in EWOM.
Civ 4 was a good game, however, its diplomacy was pretty bad. It was totally transparent in that you could actually learn what a Civ was going to do (before it actually does it) by observing how it responds to certain requests.
The classic example all the high-level players look for is "We have enough on our hands already" which tells you with 100% certainty that they are building their forces in preparation to declare war on someone (likely you). With this knowledge in hand, you can take appropriate action (build more units, gift them, etc) and check their response every turn until it changes. When it did change, you knew you had prevented them from attacking you without firing a single shot.
This, combined with the "Friendly, Pleased, Cautious, etc" and +/- numbers made the diplomacy in Civ 4 a joke, allowing smart players to game the system and win in relative safety.
in order to keep relations up you were however forced to make dificult calls, andonly in rare cases could you play the game start to finish in relative safety, and even then it required a pro-handling of the diplomatic system and an intimate knowledge of it. In no way was it ever so dull as you describe it here as it always required a careful monitoring of the opposition, assessing threats and then taking preparing measures to meet threats you may face.
Maybe you could to a certain extent predict what would happen but in a game where enemies attack each other it is safe to say that avoiding war should be possible in at least some of the games, as long as AI players have another archenemy. If you took proper care you could play a game without war. This worked fine imo.
The warning you speak of was indeed one thing to look out for. It was however entirely possible for an AI to declare war on you without warning so you at least needed to take safety measures no matter what. In the end though I think there is nothing really wrong with a transparent system. In reality countries rarely attack each other out of the blue, there is always an underlying reason an a political tension even though the reason for declaring can be anything that ignites anger. It would be fair to let a player know who he should and who he should not fear because these things are not that obscure in reality either.
In the end it would be horrible to have a system where you cannot ever predict how the AI will respond. An example would be Empire: total war. IN that game countries declare war in a whim and they cancel agreements that are mutual beneficial, factions attack their ally for no reason and small countries declare on nations that can obiterate them with whatever rookie garrison is stationed nearby. This system is complex, vague and lacks any character and therefore it is horrible.
In Civ IV I can see your concerns with the system, but I do not agree with you that the diplomacy was bad. I think in all games that I saw, the Civ diplomacy delivered what it pomised. In reality you can also to a certain degree know there are tentions on certain borders and that both sides are preparing for war. High political officials know these things. In Civ you also know these things, albeit you see these things through some arbitrary means that are abstract. Although those numbers seemed arbitrary it did the job just fine.
In the end if this predictability was really bothering you that much you could always play with random leader personalities on the AI where Ghandi might behave like a madman and attack whenever he has an itch. Even if the numbers were not telling you war was coming some proper scouting and witnessing armies gathering near your bothers were a sure warning...
In the end Civ IV was challenging all the way because on deity level the game was not beatable at times, even by the most competent players. On deity it was always tense and even if the numbers were telling you war might be possible, here was not nessecarily much you could do about it.
Diplomacy in Elementla is going to occupy a much MUCH bigger part of Elemental than anything we've done before by far.
If you've ever read the book "Game of Thrones" you will understand much of our inspiration in that department.
There are many great features available to you once you register, including:
Sign in or Create Account