Alternative "Negotiation System" for Govt Model - opinions needed!!

[roquijad]

I thought and I thought and I thought about Mark's proposal (and Plutarck's almost equivalent system) and I just can't decide. It's simple and it works, but I admit I dislike the ruler being able to get whatever he wants when he has 51% or more power. Although the speed of changes would be slow when near 50% and maybe he won't get it anyway because many things could happen along the way (due to such low speed), I have troubles accepting it. By having 51% power you know you only have to wait to make real your wishes. You just have to wait. Players will see the 51% value as something special (and something worth fighting for) and I don't like that much.

Of course, if everyone else likes it, then there isn't much to add.

Let me propose an idea to adjust Mark's proposal to something I would feel more comfortable with:
Interpret your power as the probability of things going your way in negotiations. A proposal to increase slavery is presented. All agents' powers are summed and grouped depending on their desires. Suppose that gives us:
support increase: 30%
Refuse increase: 70%

We see them as probabilities. The probability of increasing slavery is 0.3.

A dice is rolled to determine what will happen according to those probabilities. If the increase is accepted, it does so in a magnitude that depends on the share supporting the increase. In this case it'd depend on the value 0.3. In particular and for simplicity, the increase could be set to be equal to that share, so in this case slavery would be increased by 0.3.

The same is done for the proposal of decreasing slavery. And that's it.

You can see policies will tend to move in the direction the most powerful agents prefer. And the greater the consensus, the faster the change. So this is pretty much like Mark's proposal, but instead of making changes certain, they are only probable. If you have more than 50% power it's probable things will go your way in the long run, but it's not sure. Even more, if your power is just a little above 50% and the rest of society does not support you in a given policy, then that policy will mostly preserve its current value over long periods of time (the probability of increasing and decreasing is almost the same and the magnitude of change in either directions is also almost the same).

When there's enough support for increasing (decreasing) a policy value (say 70%), then in the next turns the policy will mostly change in that direction. The ocasional decreases (increases), on the other hand, have lower magnitude, so, combining both effects, you'll see a clear evolution in the policy for a given period of time.

And when you're a total despot (100% power), changes are quick and in the exact direction you wanted.

And there's no machiavelism. There's no gain in lying to the interface and it's nearly impossible to grab more power if the rest agents don't want you to.

In essence, it's only an estochastic version of Mark's idea.

Although players don't like randomness because they don't want to save and load to get what they want, I think there's nothing to fear here. Since no policy has an all-or-nothing effect (they all indicate a degree of "something") there's no important harm if a policy goes the "opposite way" temporarily.

What do you think?

[Mark_Everson]

Hi Rodrigo:

I don't have any big problems with your proposal. Thanks for trying to see it my way . On average the results will be the same anyway. Your stochastic approach just would mean things could wander around a bit more. A few points:

1. Its not clear to me that you have a way for a player with 100% power to get exactly what they want immediately. That is a must IMO.

2. It appears, though I may be misunderstanding you, that a policy can actually go in the opposite direction to the aggregate preferences of the political players. If this is so, I think players will hate that.

[LDiCesare]

I don't like the randomness much. It is not for save/reload, we can dedicate a random number generator to policies. I miss the point. 51% IS what politicians aim for, and 100% is another special value. When voting for a president, and thus determining who rules, we use a 51% rule in the real world.

As an additional remark, I think that negociations like that will only handle values with a clear range. Things like budget with lots of components are difficult to describe with that system. How do you propose to model it?

[Lord God Jinnai]

Well there is one problem with that still in that not all descions can be spanned over a period of time if there is a 51-49 power stuggle.

The easiest example i can think of is going to war.

the ruler had 55% and the others have 45%
ruler = nay
others = yea

Well a 55-45 vote means they don't go to war...without a complex mechanism to have 'covert' or 'guriella' attacks on the other nation by the others i see no way this situatation would be different if the ruler had had 100%. The same is true if the situation was revered...how would the others be able to hamper the war effort if they didn't want it and the ruler did?

I'm not saying i'm against this new proposal, its just that you seem to think that all descions can be slowed down if there was a percarious vote, atleast the posts indicatate this.

The other thing is on the compromise situation...and voting mechanism in general...its more of a question though..

What happens if a group who wants slavery at 10 when they originally voted, but 1 turn later do to some unforseen circumstances (a slave uprising) wants it increased to 20? What also happens when a new group enters into the pol. power structure or one's power is drastially shifted before a policy is fully implemented?

[Yoav Sissman]

Hi all. I'll try and carefully go around the arguments I've related to in the past. I basically agree with most of what's been said here.


(1) I don't see any reason why the NS shouldn't be activated every turn (and again in any point of the turn that the player wishes). The computional resources required are IMO negligable.

If we leave the rest of the details the same, and while the parties opinions stay constant, this will mean that the actual profile will close in on the median profile in an exponentially decreasing ratio (similar to the way that the social parameters fluctuate). So FE let us assume that we decide to close 10% of the gap on each passing turn. Then in the extreme case where the slavery policy is expected to eventually go up from 0% (current) to 100% (median of opinions), it will go up to 10% in the first turn, will be set to 19% (=10%+(100%-10%)/10) in the next turn, 27.1% on the third and so on.

I think its realistic to have government policies twist and tweak more gradually as they close on into their 'final' values. Beside, if we, for the example above, decide to have the policy change in a constant ratio (that is determined during the negotiations) we'll have to figure out a mechanisem to protect the system from manipulations by the player; otherwise if he will be enabled to initiate negotiation at will he will be able to slow down the changes by negotiating every turn.


(2) As of the way we are going to decide how much the policies are going to change every turn, I suggest that we don't use the profile that requires the smallest change. That solution would enable the ruler, should he fail to prevent a change, to slow down the rate of change to null simply by 'defecting to the yay side' and changing his profile to just near the existing government profile.

I suggest we ignore the exact 'proposed policy' of the polpowers altogether for that matter. However, I see no problem in having the rate of policy change depend on the amount of imparity in the vote.


(3) The NS suggested by Rodrigo can work fine for policies such as slavery that consist of one value, independant of that of all other policies. In the case of the decision about the political distribution of the polpower between the polblocks, we must consider the inter-relation between them.

Consider the following scenraio: the current power distribution gives polblock A, B and C each 1/3 of the votes. A's prefered condition would be to have 100% of the power for himself, and so would be that of the two others. So in the negotition there would be a majority (2 to 1) for decresing the power share of ... all of the three polblocks, and to increase the polpower of no one.

There are various possible solutions to this, like enabling a power share change only when there is a majority for both a power increase for one block and a power decrease for another; personally I don't think that this solutions is quite satisfying but we can try it out and see how it goes.

How about if I the nominal distribution of power will require some sort of a special majority to be changed? I think it is fairly realistic since practically all civilizations has a de-facto constitution if not a formal one (though it often had to do with more then power distribution). We can decide that the polpower distribution will be reshaped only when 75% of the votes would agree that some power distribution B is preferable to the existing power distribution A. Then ES will decrease and the profile will change to the most popular profile, maybe through some iterative process?

For now its just a thought for the record.


(4) A blast from the past: Rodrigo, why use a median based process to settle negotiations between polblocks but keep the weighted sum system to decide the stance of each polblock in itself?

Won't it be better to use the median system all across the board and have the parties run the negotiations without using the polblocks as an intermidiator? Seems to me that the same logic that applies to the governmet as a whole should apply to the polblocks, only in miniscale.


(5) I rather go without the randomness myself, for now.


(6) LGJ: I don't think that the option of voting about going to war was suggested here. As I see it, voting is used to decide general policies, not specific operational actions.


P.S. its kinda late so please forgive me for the grammer and spelling mistakes.

[roquijad]

Mark: When ruler has 100% it's certain things will go his way. If you were wondering about the quickness of changes (having 100% you say changes should happen immediately) then there's nothing to worry about. Since magnitude of changes depends on support share for changes, then we can control exactly how big is the change if you have 100% power. In the particular case I proposed to determine that magnitude in my post, the change can be as big as 100 (in a 0-100 scale), so you'd get exactly what you want the very next turn. However, I disagree things should change that fast, but that's something we can define in more detail if we want to implement the system.

It appears, though I may be misunderstanding you, that a policy can actually go in the opposite direction to the aggregate preferences of the political players. If this is so, I think players will hate that.

A policy may in some turns go the opposite way the majority wants, but in the long run the net change will be in the direction of majority. I don't think that's too bad...
But if you're really against oscillations, then an alternative system is to make changes to policies only if there's a majority supporting it. In other words, if the probability of decreasing a policy is, say, 0.1 and the dice indicates the policy must decrease, then you don't decrease it, but preserve its current value.

Your stochastic approach just would mean things could wander around a bit more.

hmmm, I think it's more than that. In the deterministic case it's certain you'll get what you want once you have more than 50%, but in the stochastic case it's just probable. That's IMO a significant difference. Although from a "expected value" point of view we can say the final solution is the same, the actual use of dices makes a difference. If a throw a coin 10 times in real life, I may get more or less than 5 heads, even if 5 was the expected value. Something similar will happen in the game. To my taste, that "error" is preferable compared with the certainty the magic 51% value gives you.

Anyway, Mark, since you're coding, I say just go ahead and code whatever you think is worth testing. This discussion about negotiation systems has shown to be hard to kill...

LDi:

51% IS what politicians aim for, and 100% is another special value. When voting for a president, and thus determining who rules, we use a 51% rule in the real world.

Sure, 51% IS what we use in real life in democracies. That's the danger of insisting in viewing the games' negotiation system in terms of "votes" and demanding a "majority-rule" governs the system. The aim of the negotiation system is to represent any type of govt. While the majority rule is perfect to model democracies, I doubt it's the best way to go for, FE, pharaonic Egypt.

That's why I in general don't like the majority rule. It's just too democratic to apply it to all regimes. I initiated this thread with a new proposed system based on the majority rule only because many team members are inclined to use it, not because I like it... I was attacked by the maybe-it's-me feeling!


Could you believe I've got yet another system to propose? Let's see if this one attracts more of you:

The Round Table System
Political agents (political blocks and the ruler) are sitting in a round table. An initial value for slavery is given to one of these agents (any one) and base on that value agents have to work on. For now, assume the initial value just "fell from the sky".

Each agent has "intervention points". The more power an agent has, the more points he has. Intervention points are used to alter the slavery value.

The system goes like this: the first agent receives from the sky the slavery value and uses his points to change that value according to his desires. The new value for slavery is passed to the next agent in the table, who does the same, creating a new value that's passed to the next agent and so on. The process finishes when agents don't change the slavery value any more (either because they run out of intervention points or because the slavery value matches the one they want, so there's no need to change it).

An example:
agent1: 30% power, desired_slavery:0
agent2: 10% power, desired_slavery:25
agent3: 60% power, desired_slavery:65

Slavery from the sky: 50

For simplicity, assume intervention points are equal to power*100, so points held by each agent are:
agent1:30
agent2:10
agent3:60

iteration 1:
agent1, who plays first, uses his 30 points to reduce slavery from its initial value of 50 to 20. He's out of points now for next iterations.
agent2, playing second, receives a slavery value of 20 from agent1. He uses only 5 of his points to increase slavery to 25, which is the value he wants. He's able to save 5 points for future interventions.
agent3, playing last, receives slavery at 25 and wants it to be 65, so he uses 40 of his 60 points, letting slavery at 65 and saving 20 points for next iterations.

iteration 2:
agent1 passes, since is out of points.
agent2 reduces slavery value from 65 to 60 using the 5 points he saved in last iteration. That's the best he can do. And he's now out of points.
agent3 increases slavery back to 65 and leaves 15 points in his pocket.

iteration 3:
agent1 and 2 pass because they're out of intervention points.
agent3 doesn't use any of his remaining 15 points because slavery is already where he wants.

Process ends because no points were used in the last iteration. Slavery is then set to 65. The game turn ends. (note the result will be the same in future turns unless desires or powers change)

In this case agent3, the most powerful of them all, had exactly what he wanted, but in general it's not true the most powerful actor gets his way. If agent3 would have wanted slavery to be 90, slavery would have been resulted only in 80.

That's the method. It has good characteristics: no machiavelism, increasing influence in final value with increasing power and having 51%+ doesn't imply you get what you want.

One flaw of the system is order matters. You can get different results by making agents play in different orders. But there's a way to fix it. I'll detail that only if the proposal gains some support.

What about the "fell from the sky" value? It's clear the result depends a lot on this value, so this is important. A brief analysis:
1) it can't be the current slavery value. If the initial value comes from a prior negotiation (which would be the case if you're using current value), then an agent with 51%+ obtains whatever he wants after a few game turns.
2) one interesting option is make this value the ruler's one, like a ruler's proposal to the rest agents. In this case the ruler wouldn't be sitting at the table, just the polblocks. Sounds good, but encourages machiavelism. Instead of proposing what you really want, you'll lie to manipulate the results.
3) another option is to always start with the same number, but agents wanting a value near that have it easier than the rest... unfair!
4) the best seems to be the use of "moderates/compromisers". As I said some posts away, polblocks views could be merged to build a new political view representing the position of people looking for compromises. These "moderates" have a piece of power subtracted from the original polblocks. In the round table system, it'd be the moderates who present an initial policy value to start "discussing" about in the round table. Polblocks (with their reduced power) and the ruler would then act according to the rules of the round table I showed above.

opinions?

[roquijad]

ooooops! I wrote the last post off-line and in the mean time Lordy and Yoav posted... that's why there's no comment of mine refering to their opinions...

[Lord God Jinnai]

well if we use your latest proposal i don't think pol power should determine points in and of themsleves...population and perhaps ecomic influence.

This negotiation should also be out of the players controll as he could save points till the end if he thought it would help him out more.

i think the number should since its not the current number always be 1/2 way. picking a random number is just as bad as keeping the current number as the base number, but for a different reason.

i'm interested in this, i dunno if its the best situation, espially with choosing which number to start with, but i'd need to see how the ruler plays out and your idea for how to keep the order from being biased.

In either case this doesn't nessarily answer my questions.

[Yoav Sissman]

I would like the hear the method for ignoring the polpowers order. I've got a feeling that once I hear it I won't like the Round Table (RT) system, but I don't want to criticize it until I do.

I agree with LGJ that the process must be automatic (unless we allow the other polblocks to play it wisely as well).

As for the 4 suggestions for the value dropping from the sky:

(1) I agree it can't be the current.

(2) Its a problematic suggestion since having the ruler pick the value and then not sit at the table won't leave room for he size of his power share to have impact.

(3) The constant value sits best with me. The unfairness that Rodrigo mentions is, IMHO, the principle of the RT system, and will always exist if we assume there is a priliminary suggestion that the polplayers have to spend points to change (more points the further away they are).

(4) The details are not clear enough to me.

[Mark_Everson]

Hi All:

Boy, this topic sure is "the gift that keeps on giving"...

The RT system as presented doesn't do much for me. It seems a more complicated way of getting largely the same results (or would be after being tweaked). I could certainly be wrong, but that's the way it seems after just a bit of thought.

However, now I've got a modification to propose. If A Power Block's "Power Points" needed to be split selectively between All Policies in the RT system it could be very interesting and strategic. But also significantly more complicated. It would happen similarly to Rodrigo's RT proposal except that a PB's power points could only be spent once in a turn. (and probably move things more than in Rodrigo's original proposal. You would have to have all policies open for negotiation at the same time. If one PB only Really cared about a particular issue, it could use all its points there if needed. I am assuming the participants would have to set their points dedicated to each issue before the negotiations began, though it wouldn't Have to be that way.

The proposal gives rise to strategic questions like "Is it more important to reduce Slavery, increase Foreign Policy [flexibility], or maintain my power setting"? But it might be unstable since if one side goes to increase its power while others are focused elsewhere, you could get big power swings that really wouldn't happen. My guess is because of this, some special rules would need to be there to allow points allocated elsewhere to automatically be reassigned to maintaining political power.

Such a system could be really intriguing, but could also be a real nightmare for the AI and the player. For that reason I'm not advocating it, but I thought I'd put down the idea here, and see what others think.

[Yoav Sissman]

I like the idea of allowing the polplayers to choose where they would prefer to throw their weight; I like it a lot. However I'm not very enthusiastic about making a gambling game out of the NS.

So here's my idea:
We use some variation of the system suggested in the top of this thread. Each faction is given a global number of votes equal to its power share multiplied by some convenient constant (or better yet, we can use floating-point variables for this).

Next, each polpower chooses how much of its points it would like to dedicate to each of the issues which are up for a vote, and the new gov't policy is determined using the median system for each subject independently.

As a first step we have to define the objectives of each faction in doing this. For simplicity we can define a utility function from the type:

f(GPi) = - (GPi - PPi) ^ 2

Where GPi is the effective gov't profile for policy i and PPi is the party profile for that policy.

Besides that we have to assign each party a vector of scalars - 'A' to determine the relative importance of each policy for the party. The objective of the party would be to maximize the value of:

A(1)*f(GP1) + A(2)*f(GP2) + ... + A(n)*f(GPn)

where n is the number of different issues up for vote.

We may want to make use of this value for other parts of the game as well. (Riots model?)

The purpose is to create complete transparency to all sides, so each will know what will be the political result of his point’s distribution. The paradox is that the result depends on the way others use their points, which is influenced by the way everybody but them uses theirs, and so on.

So to go around this we should find out if passing control between the polpowers in a round table order, and have each redistribute its points on its turn as necessary to optimize his utility function for that condition, will result in an equilibrium after a reasonable number of iterations.

As of the ruler, he may be flesh and blood so we can't force him to participate in the iterative process. Therefore we can have him input his distribution first and allow the other polpowers to view it as constant (skipping him in the RT). If the ruler isn't happy with the results, then he can just change his preferences and hit his 'negotiate!' button again, until he is satisfied.

Of course that this too requires micromanagment from his behalf so we can enable him to choose to leave it as an automatic process, that will input his preferences and treat him like it treats the other polplayer.

Makes sense?

[Lord God Jinnai]

well assuming we do use the RT system, the points should be given IMO as follows:

Political Power : P

5 + ((1.5 * P) + (X - 50) + Y)

X refers to the certain EG modifiers...taken the average of the entire nation/area. FE Religious Importance would be for clerical section for political power. So if it was at 100, they'd get an extra 50 points, but if its at 0, they suffer with 50 lesser points. Aggresiveness could be used for military, etc.

Y represnts certain philosophy settings in place and how they would modify it...as philosophy section seemed kinda scant, i don't have much idea for that area right now.

[LDiCesare]

A RT where each party has a limited number of points is interesting, but will be nightmarish from the AI point of view, in particular because it cannot know what the player wants to do, so the parties may put all their votes where that pleases the ruler and the player, by saving/reloading, may be able to determine how to vote to get the best of the system. So, although I like the idea, I think it would:
1) be very complicated for the ai to handle
2) lead to abuse
LGJ, why do you want something other than political power to have an influence on the number of votes?

Sure, 51% IS what we use in real life in democracies. That's the danger of insisting in viewing the games' negotiation system in terms of "votes" and demanding a "majority-rule" governs the system. The aim of the negotiation system is to represent any type of govt. While the majority rule is perfect to model democracies, I doubt it's the best way to go for, FE, pharaonic Egypt.

I don't know about Egypt. I know however that Louis XVI's France was based roughly on very small ruler power with roughly:
10%ruler, 30%noble, 30%clergy, 30%"Tiers Etat". Nobility, clergy and "the rest" had equal power, and thus the first two always got what they wanted because they voted together. The Tiers Etat never got anything. That is why the Revolution came, because they had no way to influence the policy, even though they had 1/3rd of the votes. Now, admittedly, the ruler's power could have been more important, but it was relinquished by the king in order to avoid riots. I think a model that would allow Tiers Etat to change things with their 1/3rd of votes would be historically inaccurate.

[axi]

Such a system could be really intriguing, but could also be a real nightmare for the AI and the player. For that reason I'm not advocating it, but I thought I'd put down the idea here, and see what others think.

This is in fact internal diplomacy. Perhaps if we code a sucessful model for regular diplomacy, we can adapt it to the internal affairs of the civ. If the AI can handle regular diplomacy *and the civ3 AI seems to have proven that it can) then it can handle internal negotiations.

On Yoav's proposal: Wouldn't be too resource consuming to have several iterations where we maximise a function of several variables?

[Lord God Jinnai]

Originally posted by LDiCesare
LGJ, why do you want something other than political power to have an influence on the number of votes?

Well the reason is beause while their political power may be low, if they have support of the masses, their indirect political power may also influence the situation. This is not obvious with a casual look and in many cases might not even matter. Political power, represents just that, how much they can sway in the realm of politics. While descion making does have much to do with that, it is not 100% politics.