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

[Mark_Everson]

Originally posted by LDiCesare
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. (Snip)

Hi Laurent:

Sorry I don't understand you. If the first two always vote together all that guarantees is 40% of the power. None of the systems we are proposing will allow a side to mostly get its way with 40% of the power. Now if the Clergy always vote with the ruler then what you say makes sense to me. But then its 70% vs 30% and all procedures will give the 70% side most or all of what they want, depending on system.

Everybody:

One point I would like to make on the median system... Even democracies don't decide Many issues on a straight voting basis. There are a variety of 60%-support or 66%-support, or even stronger, thresholds built into the US system (only one I really know about.) These thresholds are there for a reason. Specifically to Prevent a slim majority from "changing the rules" on a substantial majority of the population. In the US system they seem to work fairly well to prevent potentially transient majorities from imposing their will on the minority.

For the big issues like the Government model covers, in many cases 51% would not be adequate to change the governmental "setting". Could we do it in a game? Sure! But don't go saying that its that way in real life.

[Yoav Sissman]

Axi:

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

Normally I would say no, but on this case where the variables are discrete I can't say until I code the whole thing and test it. Since every algorithem I can think of will be expensive in computational resources (though not exponential, I believe) it all depends on the number of polpowers and policies involved. I don't mind giving a shot at coding it, but I want to see how the basic median system works in the demo before I do.

If it'll work fine and relieve the objections, then we probably will be better off keeping it simple. If it'll cause us problems we'll have to consider other alternatives, including my new proposal, which may at least prevent the ruler from retaining total political power if he has 51% polpower.

LGJ:

The political power shares are meant to represent both direct and udirect influence.

[roquijad]

Hi Gentlemen.

A note on AI: If each time the political agent is allowed to use only one "power point" (in a round table sheme), then that can simulate a good enough AI. And at the same time solves the problem of results depending on the sequence agents play.

As a cost, it takes more iterations, but nothing big.

Each agent affects the policy a bit so agents can decide how to spend their points considering the actions of others. So they try to stop one and other or work together changing policies in the same direction.

In Yoav's system, each agent would look at his f(GPi) in his turn, choose the lowest and use one point/vote there. Next iteration things could be different and he might prefer to spend his point somewhere else.

I'm excited. I feel there's something very close to an agreement around here.

Imagine this system:
1) a govt profile falls from the sky to a round table (let's not care for now how that profile is created)
2) one by one, political agents in the table use their power points (given multiplying power shares by some constant) to change policies. In one turn each agent can affect only one policy and only using one power point (to increase or decrease policy value). The policy an agent chooses to change is given by an utility measure such as the one Yoav proposed.
3) iterations go until all agents run out of points or don't want to change the profile anymore.

That has enough AI (agents can react to what others are doing) and involves interesting strategic moves making agents concentrate in what they care the most.

What about the player? He simply puts his desires in the interface and the system treats him like any other agent. So he doesn't have to actually play in the round table. Everything is automatic. Note one could say it isn't fair to model player's utility function because probably the player has different priorities. But players don't need to know how the system works. For all they care, the more power they have, the closest final profile should be to their desires.

If I understood you all correctly, the results from this system would please most of you.

[Lord God Jinnai]

Yes it looks likec it would work, but it could use a lot of time...i'm not sure here so i could be wrong...

The other point is what i'm calling the 'conede factor'.

Policy A: Current 40
Policy B: Current 20

Power X (10%):
-----------
A: 70
B: 30

Power Y (40%):
-----------
A: 10
B: 20

Power Z (50%):
-----------
A: 0
B: 15

Okay in this situation A is the most off for all groups so based on roquijad's last post they all put their points into Policy A. Well Power X who has very little power will loose that battle, but it also wants to raise power B and since it knows it is going to loose that battle could in theory attempt to use its power to raise B while conceding the fact that Policy A is out of its control. This is how politics does work and we are trying to get something that works as closely as possible without detracting from the 'fun' element.

[Mark_Everson]

Hi Everybody:

This looks to be Really shaping up! I just wanted to make one comment. I think a strength of the evolving system is Yoav's suggestion of weights, because it can be used to make Political Blocks act like their real-world counterparts. So an Ethics PB would have a relatively higher weight on ethical issues like slavery and less about private property. The Money-Bags on the other hand will care a Lot about private property

One other quick note. I think the votes need to be distributed roughly evenly in time as the negotiations go forward. IOW if a PB has only 5 percent of the votes, it shouldn't be able to use them all up in the first few rounds, and then have nothing when its most important issue comes under attack. I'm sure we can come up with a good way to do this.

Wonderful effort gents! Lets hope there's not a king-sized fly in the soup that we haven't recognized yet

[Fosse]

Newbie question

Team,
Regardless of the model that gets adopted I am worried that there will not be a fundamental difference in the FEEL of different government types.

In most Civ type games the government you pick just acts as filters for things like unhappieness and economy, which is artificial and feels wrong... and leads to situations in which the most noticable differnce between democracy and communism is that democracies cry when they go to war.

Something I am worrying about with Clash is whether or not I'll be able to tell the difference between being a ruling monarch or an elected official as the leader.

Ruler's influence can of course be altered to simulate how much power is given to monarchs/dictators/presidents/whatever, but will it finally give me the ability to FEEL like I'm actually changing governments instead of picking a different set of filters?

For example, Tudor England. The Church had influence over what the ruler (Henry VIII) could and could not do, so he was able to, as king, simply remove the church and establish a new one that would agree with him. In Clash terms he simply replaced the Ethics polblock that had a Governmet profile he disliked, with one that was more to his liking.

Now, in a modern democracy this is impossible, but as a king will I be able to make such drastic decisions... always bearing in mind that they might come at a cost?

From what it sounds like in the threads and model description is that I won't be able to do things like that... I'll only be able to vote that some blocks get more power while others get less. I know that there may be room for assanations and such (which is great) but Henry was able to do that legally just because he was king.

Another example would be totalitarian regimes of the 20th century.... people with power in Nazi Germany disagreed with Hitler and would even voice that opinion... but it just didn't matter because he did what he wanted.

In short, will the current government I am running under do anything more than change the influence I have over the government profile, which is the filter that my civ's raw resources are sifted through?

I really hope that a democracy will feel like one, while a theocracy will feel like one... am I even making sense?

I hope that this doesn't sound critical, and I hope that I was clear in my question. I applaud the work that is being done and will be happy to comment once I am able to test the current Demos (I'm traveling right now and only able to get on-line in NetCafes... but all I really want is to play the demos!!)

Thanks for your answers, keep up the good work

[Mark_Everson]

Hi Fosse, and welcome to Clash!

I get the impression you're basing your worries on just this thread. This thread is just talking about a small subset of the govt/social model that Rodrigo has done a great job with. Check it out on the web site. I think we already have at least vaguley planned most of what you're looking for. But let's see what the experts say!

[roquijad]

First of all, your signature, Fosse, is outstanding! LOL

Second of all, I feel it's better to answer your questions in the Govt Model Thread, so we can concentrate here on alternative negotiation systems. Please follow this link:

http://apolyton.net/forums/showthread.php?s=&threadid=30817&pagenumber=2

Regarding negotiation systems, I'm worried posts have stopped appearing in this thread. I'm "re-launching" the discussion based on lasts posts to encourage more comments:

Negotiation system #4512
1) An initial (proposed) govt profile is built (formerly known as the "fell from the sky profile"). IMO a good starting point is to take moderate's view (the pro-compromise party).
2) Each political agent (5 political blocks and the ruler) is given a number of "points" according to their power. We could give each agent a number of points equal to 100*power_share, so if you have 30% power, you get 30 points to "play". But a better option is to give each agent 100*X*power_share points, where X is the number of govt variables (policies and power shares). See note at the end.
3) Negotiation starts. All agents play one by one, always in the same order, as if they were playing "monopoly". The order doesn't matter and is built-in the game.
4) Each time an agent has to play, he chooses one of the govt variables (a policy, a power share) and uses 1 (and only 1) point to change it in the direction he wants. The policy is increased/decreased by the same magnitude (1).
5) The govt variable the agent chooses to change in 4) is the one for which the value Wi*Abs(Di - Ni) is maximum, where Di is the value the agent desires for govt variable i, Ni is the value govt variable i has currently in the negotiation and Wi is the weight (importance) the agent assigns to govt variable i.
6) In the case of polblocks, Wi's are game constants, defining the importance of each variable for that particular block. In the case of the player/ruler, Wi's are all equal to 1.
7) When in this negotiation is the player's turn to play, the (human) player does not really play. The computer plays for him just like he was any other agent. Note the computer knows player's Di's because he has entered those values previously in the appropiate interface.
8) Negotiations end when all agents run out of points or when agents don't want to make any further changes.


Note1: In the case of choosing to change a power share, the system will "ask" the agent to play again and use other point to adjust other power share so they all keep summing 100%.

Note2: Wi's represent power blocks' priorities, but it's a little bit unfair for the player when his Wi's are all equal (he can't tell the system his priorities). I don't think this is important, but in case other believe it is, we can come with solutions. The obvius one is to let the player enter his Wi's through interaface, but I'd find it boring and unfriendly (interface-wise). Another option is to let the player choose one govt variable (through a check box, FE) to indicate he wants emphasis on that one. The system would increase the corresponding Wi. Another interesing option is to allow the human player, if he wants it, to actually play in this system, choosing each time the variable to change. That would be the micromanagement version for those who want to control things at the lowest possible level.

Note3: Why 100*X*power_share instead of 100*power_share? Because if we use it w/o X then we can't guarantee players get exactly what they want when they have 100% power.


Characteristics:
1) We need 100*X iterations each time (each game turn). I don't know if that number and the usage of a search method for the maximum each negotiation-turn for each agent is expensive in terms of computing resources. Someone who knows please speak! (we're talking about a search in a list of about 15 elements, btw). (X=15 aprox.)

2) Unless agent's opinions change or power distribution changes, the result is always the same. That's good.

3) The more power you have, the closest the final profile will be to your desires. That's good.

4) If you have total power (100%), then the final profile matches exactly your desires. That's good.

5) There's a little of machiavelism, but IMO negligible. If the player puts extreme values in the interface for a few policies, then he can get for those variables something closer to what he wants, but there's the big cost of getting worse results in the rest variables. That's a cost that doesn't incentive mahciavelism. Note taking extreme positions in all variables has no effect at all.

6) No strange things nor despotism at 51% level. That's good, at least for my taste.

7) In general, the main problem of this system is that the selection of the initial govt profile (step 1) ) needs to be made with care. Agents who have a desired govt profile similar to the initial one, have it easier. The option I proposed above (use a mix of polblocks' desired govt profiles (moderates)) is one I feel comfortable enough with. Other option is to set the initial profile with all govt variables equal to some fix value (always use that same initial profile) and compensate agents with more points the further away from this value they are. For example, if all govt variables are set equal to 50 and agent1 wants a value of 55 for slavery, then he gets 5 extra points to play with as compensation (each variable gives you compensations, of course, and you sum them all to have a total extra points that are added to your "legal" points). I haven't thought much about this solution, but I guess it'd work.


Comments extremely appreciated.

[Lord God Jinnai]

Well for the human players priorities i think we could set a panel where he could select the priority of each one based on the number of dots (say 1-5 dots). Telling the player setting all dots to 0 is the same as setting them all to 5.

[Mark_Everson]

Hi Rodrigo:

This is looking really good, just a few points.

Originally posted by roquijad
3) Negotiation starts. All agents play one by one, always in the same order, as if they were playing "monopoly". The order doesn't matter and is built-in the game.

This still has an obvious defect that I commented on above, you may have missed it...

I think the votes need to be distributed roughly evenly in time during the process as the negotiations go forward. IOW if a PB has only 5 percent of the votes, it shouldn't be able to use them all up in the first few rounds, and then have nothing when its most important issue comes under attack.

In the case of the player/ruler, Wi's are all equal to 1.

Note2: Wi's represent power blocks' priorities, but it's a little bit unfair for the player when his Wi's are all equal (he can't tell the system his priorities). I don't think this is important, but in case other believe it is, we can come with solutions. The obvius one is to let the player enter his Wi's through interaface, but I'd find it boring and unfriendly (interface-wise).

Well, I think our players would find it Very frustrating that they can't use their power where they want. I am with LGJ that we need to give the player the option to enter these. They needn't be numbers. The even settings would be the default ones. These settings would not need to be changed frequently IMO.

Note1: In the case of choosing to change a power share, the system will "ask" the agent to play again and use other point to adjust other power share so they all keep summing 100%.

Cute idea! I'm not sure it'll work, but it keeps things neat.

[roquijad]

I think the votes need to be distributed roughly evenly in time during the process as the negotiations go forward. IOW if a PB has only 5 percent of the votes, it shouldn't be able to use them all up in the first few rounds, and then have nothing when its most important issue comes under attack.

I didn't miss that comment the first time. I don't understand exactly where you see an "obvious" defect. All issues (policies and power shares) are being processed simultaneously. Each agent sees all variables and chooses, each time it plays, where to "spend" points. It can concentrate in only one of them if that's a really important issue for it.
I might be misunderstanding you completely....

Well, I think our players would find it Very frustrating that they can't use their power where they want.

Hmmm, I disagree. A player will be using his points where he "wants": taking an extreme example: if a given govt variable is too far from what he said he wanted (in the interface), the system will focus ruler's spending on that one. Allowing the player to define their own Wi's gives him more control on points spending, sure, but I wouldn't say it'd be Very frustrating not to have that enhanced control.
Anyway, if you all think is really important, let's go with it. To me is a minor thing.

[Mark_Everson]

Originally posted by roquijad
I didn't miss that comment the first time. I don't understand exactly where you see an "obvious" defect. All issues (policies and power shares) are being processed simultaneously. Each agent sees all variables and chooses, each time it plays, where to "spend" points. It can concentrate in only one of them if that's a really important issue for it.
I might be misunderstanding you completely....

Hi Rodrigo:

I tried to talk through an example I thought could make the point. I have come to the conclusion that I was completely screwed in the head. But my trying to work it out did illustrate the system better to me, so Im' going to post it anyways! I hope I didn't screw up the math since I'm getting pretty sleepy.

All Policies at start of negotiation are 50%, I assume the moves are made in order the parties are listed in.

Policies are P1-P10, just using numbers because its simpler
Ruler, with 60% power, 500 points
Desires 75 for all policies except 25 for P6
W = .25 for P1; .10 for P2-P5; .05 for P6-P10

Wombats with 20% power, 200 points
Desires 50 for all policies, except 60 for P6
W = .05 for P1-P5; .25 for P6; .05 for P6-P10

Bozos with 20% power, 200 points
Desires 50 for all policies, except 60 for P6
W = .05 for P1-P5; .25 for P6; .05 for P6-P10

Result:
First 5 rounds result in P1 at 55 due to ruler action and P6 at 60 due to combined Wombat/Bozo action. That was easy.

The next 195 rounds are spent with Ruler and Wombats fighting over P1, to a draw. I assume here that if a PB has all policies where it wants them its allowed to pass and retain its points for later. So the Bozos do a lot of passing since by the time it gets to them everything is perfect. By the end of this period the Wombats have run out of points.
P1 = 55, P6 = 60.

The Next 195 rounds results in the same with Bozos blocking Ruler. At the end of this phase only 205 points remain to the Ruler.
P1 = 55, P6 = 60.

Ruler then in priority order changes all policies in P1-P5 to 75 using 120 points. Then 10 points are spent to bring P6 back down to 50, and the remaining 65 are spread over P6-P10 evenly, giving final values of 63 for each of these, except P6 which now is 37.

The result shows that [b]the Wombats fail to achieve any success on holding the line on their most important issue, even though the ruler didn't care very much about P6. But I guess that's just the way the system works out...

Cya,

Mark

[LDiCesare]

I like the system a lot.

Note1: In the case of choosing to change a power share, the system will "ask" the agent to play again and use other point to adjust other power share so they all keep summing 100%.

in particular is very good.

I think the max(Wi*Abs(Di - Ni)) formula may need some alternatives so we can check what results they give. The only alternative I can think of is (Wi *squareroot(Di - Ni)) which means you will fight in order to attain desired values for the most important Wi's. (I tried square too but it is silly).

About showing Wi's to players, I think we could propose a binary choice: important/don't mind. That allows the ruler to abstain when they want to instead of being forced to side somewhere.

Start values:
If we start at 50 for 15 policies, then 50*15 points for 100% power is enough to be a perfect despot, so maybe the number of points should be 50*X*power share instead of 100.

I think we should have a few tests like Mark's, with changing parameters as follows: Number of parties, formula for throwing one's votes, and number of points.
Mark's test with twice less points for everybody shows a lot of differences as, instead of having 210 points left at the end, the ruler only has 105. This means the ruler cannot set all P1-P5 to 75 and the choice of preference formula kicks in:
With 105 points, starting from P6=60, all the rest = 50 (I think I have slightly different figures than what MArk wrote, but the end result was the same):

Wi * distance: P1 = 71, P2-P5 = 65, P6 = 47, P7-P10 = 53
You can see a big difference with 100*X points, as the ruler doesn't get all they want on most policies because they bickerered on P6.

square root: P1 = 75, P2-P5 = 69, P6 = 56, P7-P10 = 50
Here again it is quite different. I may be 1 point wrong somewhere because I did the sqrt by hand. The Wi is much more important here: P1 is set to its desired value because the big Wi makes a big difference . Also, you can see that P6 has been handled differently from P7-P10 because there were not enough points left to use on small Wi's. With 100*X points, the results would probably have been the same as Wi* distance.

[LDiCesare]

This system requires additional controls in the user interface. Instead of a single slider for the value, we need a slider for the value and a slider/toggle for the importance of each policy.
The result of this is a lot of buttons. In itself it's not a problem, but since the outcome of the ruler's intervention will be differene depending on how many points he puts here or there, I suspect that the player will end up moving the value sliders, and then playing with the priority sliders/toggles in order to see what gives them the 'best' result. I think this may produce a lot of micromanagement.

Anyway currently, the coded system is a weighted average of each political block's political power*desired value. I'll try to put a NS system in instead (the NS only makes things easier for the player from a user interface point of view). Then I'd like to have comments about this user interface/tweaking buttons critic befre I decide to start on the system discussed above.

[Lord God Jinnai]

Hmm...perhaps it might be better to have point changes be weighted depending on how far they are from 50? (or from where they began before negotiations took place...which might be better as the populace would be more used to that)?

Another idea would be that a power base would be able to prioritize a specific issue...if they did, that policy would cost them less to move, but all others would be increased in cost (you couldn't have more than 1 or 2 at the most). Of course for this to work, moving 1 point would need to cost a base of 2 units.