Diplomacy, how does it work?

[hoverdog]

I'll just repost our newest update:

Diplomacy, theoretically, should be an extremely important field in games that are all about the clash of mighty interstellar empires. However, in most 4X, relations with AI (and even human if there’s no outside communication) factions are overly simplified and, thus, bland. Most of the time, it boils down to exchanging technology, agreeing on non-aggression pacts or alliances, and, of course, declaring war. It’s the shallowest area in otherwise deep and sophisticated games (see MoO2). Playing a diplomatic race isn't that much different from playing a despised one, you are just less likely to be attacked.

Starlife aims to overhaul diplomacy completely. We want to make it into an important factor that you always have to take into consideration, even when not directly parleying with others.

The main element on which the system is based on is Intention. You could also say that intentions are one of the bases of the whole game, because you place them on almost every field (for example, you ‘intend’ to colonize a specific planet, which results in civilian migrations), but its impact is most visible in diplomacy.

Let’s illustrate it on an example. You are at war with another race, and decide to ask for peace. When sending emissary, you have to choose why you’re doing so, like:

1. Intention: War preparation. You want to regroup and continue the war at your leisure.

2. Intention: Improve Relations. You no longer wish to pursue combat with this race, maybe because they’re more powerful.

3. Intention: Espionage. You covet their secrets and plan on stealing them.

Each intention gives an appropriate bonus (like faster ship construction in example 1). Obviously, the enemy doesn't know your true plans, but – if they managed to infiltrate your empire – can learn them and respond properly.

Each intention has a goal you need to achieve connected. If you choose Espionage, you are supposed to send a number of spies to enemy territory; if you decide on War preparation, you are to declare war within specified time gap.

Now you could of course try to meta-game and pick only false goals to stump the AI. However, breaking your objectives comes with penalties. Intentions are what you communicate to your subjects. So, when you choose War preparation, the entire empire commits to constructing a powerful navy. When it turns out that you had no plans of going to war, they are confused and angry at you. The penalties for breaking objectives won’t be too harsh, but enough to make you think twice, especially if you break your promises willy-nilly.

Another example. You’re at war with two AI races (A and B). You make peace with A with the intention to go to war again, and with B to just shake hands and restart trading. You start to build a new fleet. This makes both A and B uneasy for obvious reasons. They both may have spies – let’s say A learns your true intentions but B doesn't. Now, A can start a preemptive war on you, and B joins in because it thinks you’re might be dangerous to them (it doesn't know you want peace with them and won’t believe you).



DISCUSS!!

 

[Destroid]

Woah when did hoverdog become a full blown member of the team?

 

[tiagocc0]

Well, he won the writer contest. And he is our writer/designer now.

 

[tuluse]

I meant to reply to this days ago, but then forgot.

Anyways, I really like the idea of having a secret agenda or at least internal agenda.

However, I don't like your specific example of building ships faster because you choose war preparation. I'd prefer ship construction not be affected in such a way, and only be affected by capacity, resources, and manpower. I would much prefer something like a war weariness rating, which would be affected by your intentions. In turn, man power available could be affected by war weariness. I realize this is more complex and thus harder to do, but I'm just wish casting.

Also, I don't like the goals as explained here either. If I prepare for war, I shouldn't have to declare war, but just get some kind of concession from the enemy. Or maybe even have a defensive oriented plan. Like the UK before WW2. They were re-arming and preparing for war with Germany, but they didn't actually want to fight the war. If that could be simulated somehow it would be p cool.

Just some thoughts on the not complete ideas presented here.

 

[tiagocc0]

I get your point tuluse, it would be a lot less 'arcade' the way you describe. I will see to it, for now it's just examples we wrote on the paper, we haven't coded it yet.

About the prepare for war, I think it should force you to go to war, in the UK case it was probably another intention related to defense that we can also put in the game.

 

[Norfleet]

I don't think this "Intention" idea works well: It's overly simplistic and restrictive: What if you simply have no singular intention, but instead, a multi-option plan which may or may not involve any of these goals? On top of that, it doesn't work well. Say, for instance, my goal is to build a powerful navy. You set a requirement to build an enormous number of ships...but my plan for a powerful navy doesn't include an enormous number of ships at all. Instead, perhaps my vision for a powerful navy involves a small number of highly effective craft that I micromanage intensively: Building an enormous number would be both expensive and ineffectual. Suppose my powerful navy is a single ship capable of shredding anything that it gets in its sights, such as, say, "the entire enemy fleet". The game simply wouldn't be able to comprehend what it is that I have made.

Plus, plans change: Perhaps your goal is to prepare for a second war, but a third player enters the arena and alters that plan. Perhaps these new guys attack you. Perhaps these new guys are more attack-worthy than your previous targets. For whatever reason, something happens to postpone or otherwise change your goals. I don't see why one should be locked into this course of action or heavily penalized for not undertaking a course of action that changing circumstances have rendered undesirable. Perhaps you wanted to attack them again, but since then, you're getting along better and no longer see a need to do so. What then?

 

[tiagocc0]

The bonuses or penalties the intention is going to give you are not to guide your game, nor to block you or force you into a game style.

They are bonuses or penalties that may help or get in your way but not to the point where you lose a game or win a game because of this. It gives you a direction. If used well you will have smooth game, if not used well you will simply have a hard time dealing with the penalties which will disappear quickly.

There will probably be a mechanism to change plans if needed that could have less penalties than breaking an intention, but might give less bonuses as well.

In your case, we just need to added a different intention, one is to build a fleet, your is to build one ship.
They are different and are not mutually exclusive.

The main goal of the intention is not the bonuses and penalties, they are included as for you to not abuse the system.
The main goal is to give your empire a direction, that enemies may or may not learn.

Thus also giving you a way of learning your enemies directions.
It's hard to say that this won't work well before we code and test it. Let's see.


One could simply use intention and break it everytime as to confuse your enemies, it means that you will also confuse your friends/government. Because one needs first to fool their friends as to be able to fool their enemies.
This is perfectly acceptable, you are only setting the difficult bar a little higher than normal because you will have to constantly deal with penalties.

 

[Norfleet]

The main goal of the intention is not the bonuses and penalties, they are included as for you to not abuse the system.
The main goal is to give your empire a direction, that enemies may or may not learn.

Unfortunately, the problem with this idea is that the pool of intentions you can possibly create, and the conditions to test for whether or not the player is following them, is simply always going to be much smaller than the number of things the player could possibly be intending to do. What happens when none of your existing options fit?

Thus also giving you a way of learning your enemies directions.
It's hard to say that this won't work well before we code and test it. Let's see.

I would think you should learn your enemy's direction by observing his actual actions, not by some nebulously stated strategy fixed in stone by the developers. This is especially true in a game like this where plans must necessarily remain fluid with multiple contingency options which may go in radically different directions.

Quite frankly, this resulting system seems more clumsy and less comprehensible to the user than the traditional system of treaties, and seems to be asking the player questions about the future that, quite frankly, he may not even know the answers to.

On the other hand, it would be rather nice for for the player to have some ability to communicate common desires or interests to the AI, which otherwise would not have the ability to discern what the player is doing. It is far too common that the diplomacy system sort of comes apart in the face of an AI that behaves obnoxiously because it is simply unable to understand the meaning of basic things, like "THIS LAND IS MINE, NO YOU CAN'T JUST SNEAK A SETTLER UNIT IN HERE AND COLONIZE IT".

 

[tiagocc0]

It is rather simple, the idea is not for you to use your complex plan (thoughts) to the game. But instead build one from simple pieces.
It must remain basic, with basic ways to check (usually just one checker for each, that will be communicated to the player).
While you do can have complex plans on your mind, you can't say it all to your government. The idea is not to use the system to make all your plans, your plans are still yours, it just gives you small bonuses, the idea is to inform your government of the possible course of actions.

You are reading too much of the system.
It is not a question about the future. If I ask you right now what are your intentions about DU you will probably say to remain at peace.
You don't have to look at the future, quite the contrary, you just have to look at what is happening right now.

The intention is not your real intention, it's the intention that you can communicate to your government.
The same way a player can use it to trick the AI, some AI may use it to trick the player.

 

[Norfleet]

It is not a question about the future. If I ask you right now what are your intentions about DU you will probably say to remain at peace.

Actually, a more accurate representation of my intent would probably a shrug of indifference and a statement of "No comment". Does such a statement exist under your system to grant me no bonus or penalty?

Because it is extremely likely, and generally the case, that I undertake actions for an inscrutable long-range agenda that regards everything else as being annoying distractions from it, and no singular option you can probably put there would in any way summarize my immediate aims, which are basically "continue to carry out the Great Plan, which has basically absolutely nothing to do with any single factor in-game, and wait and see to what happens".

 

[tiagocc0]

No because this is your statement, not the statement you will communicate and mobilize your empire for.

 

[tuluse]

The way I think of it, is that "intentions" aren't your intentions as a ruler, but rather the goals and aspirations you give to your people to accomplish something. I see it as more of an internal policy than anything.

 

[tiagocc0]

Yes, but anybody else looking from the outside will see it as your intention.

EDIT: Let me put it this way.
In a normal diplomacy you just sign peace, war, trade or make an agreement.

This is what you are communicating with their empire. It is your intention to do so.
For example your intention to make peace.

Here we just add another layer.
Now you state your intention to the other race, and another intention to your empire.
And here they may differ.

But there is a third layer.
You can communicate your intention to the other race, communicate another intention to your own race but still have in your mind a different intention.

This third intention is yours, the game can't and won't simulate it.

 

[Norfleet]

The way I think of it, is that "intentions" aren't your intentions as a ruler, but rather the goals and aspirations you give to your people to accomplish something. I see it as more of an internal policy than anything.

Shouldn't this be more tied with your empire's politics than with Diplomacy, though? I could see campaign promises as a major element in the politics of running your own empire, but does this really have anything to do with diplomacy? When I agree to a particular treaty, I don't necessarily have an INTENTION to do anything.

I could see this campaign-statement concept as an interesting part of running an empire, where you state your goals and this affects the morale and devotion of your people correspondingly, but I don't really see this as something that makes a lot of sense tied directly to diplomacy: I can desire to motivate the people into some kind of goal without the presence of any outsiders...alternatively, I may not have any particular intention at all with regards to these tiresome and annoying aliens from the other side of the galaxy.

On top of that, it doesn't actually enhance the diplomacy system at all. The system seems more made to be gamed and exploited arbitrarily, which makes perfect sense in the context of politics, but does nothing to enhance the diplomatic system except add a layer of confusion to it.

 

[tiagocc0]

This system will be tied to other parts of the game.
But because of this system Diplomacy was able to become so much more.

If you make peace with the intention of making peace is the same as having no intention.
So yes, having no intention is an alternative. It just means that you are not lying.

 

[IDtenT]

How about using a normal distribution on the odds of accepting an agreement and bias that distribution based on other things? I feel that the biggest failings in AI design is an over-obsession with determinism.

 

[tiagocc0]

How about using a normal distribution on the odds of accepting an agreement and bias that distribution based on other things? I feel that the biggest failings in AI design is an over-obsession with determinism.

tell me more

 

[IDtenT]

How about using a normal distribution on the odds of accepting an agreement and bias that distribution based on other things? I feel that the biggest failings in AI design is an over-obsession with determinism.

tell me more

We know that RNGs are sometimes used in RPGs and other games, but they are being used directly as representations for a uniform distribution not unlike what is found in say RPG character generation. It is true that these uniform distributions do add up to a a semi-normal distribution - so it's forgiveable (Even though every attribute is completely uniform in outcome, the total outcome of all attributes tends toward the same sum). What I'm suggesting is that a RNG is used in diplomacy in such a way as to represent the area under the distribution, if the offer is within that area it is accepted, if it is not it is rejected.

Now how to determine where the offer lies is up to the designer, but if intrinsic values are applied then one could easily start at 0 (A bit of a misnomer, but unimportant), have the mean at the actual value and have the distribution end at double the value (again misnomer). What one can do then is bias the normal distribution according to relational attributes or whatever (making the mean be equal to actual value +5, etcetera). The normal distribution incorporates diminishing returns after the actual value and an increasing likelihood when approaching from the other end.

Doesn't just have to be diplomacy. It can be called upon in any AI algorithm based on choices.

Hopefully I'm making some sense. I'm slightly drunk.

 

[tiagocc0]

I get what you are saying but I would to do some research to be able to develop it this way. I'm not that good with math
But I will take a look at this

 

[IDtenT]

I get what you are saying but I would to do some research to be able to develop it this way. I'm not that good with math
But I will take a look at this

The most important part is shifting the mean and the variance about. Or well, those are the only things you /can/ care about.

 

[Norfleet]

How about using a normal distribution on the odds of accepting an agreement and bias that distribution based on other things? I feel that the biggest failings in AI design is an over-obsession with determinism.

Disagree here: The idea that the AI will RANDOMLY accept a proposal or not is fundamentally flawed. Firstly, the idea that something like this is simply balls-ass random just promotes save/reload type behavior. Secondly, it makes no logical sense: A proposal should be accepted based on its merits, none of which include any kind of randomness, with the AI always accepting an obviously profitable agreement. The basic test, which so many game AIs fail: I offer them $2 in exchange for their $1. This is a straight-up exchange in which they profit without any drawbacks at all. This proposal should always be accepted, and any AI which fails to accept this is essentially defective.

With this in mind, the AI should evaluate a deal based on two factors:
1. Assessed value of the items on the table.
2. Uncertainty in those values.

In the above example, the assessed value is exactly as given, and the uncertainty is zero because the items are identical and their values are concrete and fixed. Therefore, since the balance of the deal is positive, the AI should always accept this deal, just as a human would. If a deal is of more uncertain valuation, where some of the items on the table have uncertain values, then the AI should either ask for more, or consider it whether or not it likes you enough to do it anyway. But in a deal where the profit exceeds any uncertainty (in this case, none), none of these factors should matter.

 

[tuluse]

You're making it sound a bit too easy Norfleet, because you have to figure out opportunity cost which can be quite complex.