Government Model v. 3.1

[roquijad]

A new thread for a new model version. Just to keep things organized.

The govt model v.3.1 can be found here:

http://clash.apolyton.net/models/Model-Government.shtml

The old forum thread is here:

http://apolyton.net/forums/showthread.php?s=&threadid=12292

More old threads can be found listed in the govt model page in the Clash web site.

At the end of the old thread I said:

I suggest a completely new approach for [the discussion about the 51%rule vs the negotiated system]: Tell me what the model can't do. Much of the discussion goes around the processes inside the model w/o considering the model's output (which is all that really matters). If you show me an historical framework where the model produces a wrong, unrealistic output, then we'll be on the right track for a discussion. If we find there's a long enough list of things the model can't represent correctly, I'm all for changing it. But please, please, please, before you start throwing examples of the model producing wrong outputs, give me a chance to make a post to clarify a couple of things some of you are getting plain wrong. I can't do it now because it's almost 3AM and I want to go to bed. In a day or two I'll do it. ok?

OK, here it is: important things that aren't clear enough and should be kept in mind:

1) The model was conceived to handle "strategic" political decisions. The ruler and the people (by people I refer to all social classes) are called to decide on things like the type of economic system for the civ, a gross level of civ's aggressiveness in the international field, gross levels of religious and ethnic discrimination and a couple more other grand scale decisions. Because they're strategic decisions, the list of decisions is, therefore, small. It was never my intention to allow decision-making over a long list of policies and laws covering a multitude of topics. There're 3 reasons for this: A) to stay in line with Clash philosophy of avoiding micro-management. B) because it'll be simply tiresome for players to have to obliged to deal with a long list of decisions, many of them related to one and other (therefore demanding a more detailed player understanding of how each policy affects his civ); and C) because every time you add a new policy to decide upon, you have to generate a method to define what is what the people think about it, in order to simulate what they'd do with the policy if the current regime allows them to take part in the decision-making. And doing this is tough because there's no "general method" to simulate people's opinion about any possible decision, so each existent policy adds complexity to the model.
Why I'm saying this? because in examples of decisions presented in the thread, some of you have used low-level or very specific decisions that the model is not intended to handle. Although I understand some examples are just that, simple examples, do keep in mind the model wasn't meant to deal with a vast variety of decisions.

2) The interaction between the player and the model (the govt "window") will be, in general terms, infrequent. Don't imagine the player taking govt decisions every game turn or something like that. The ruler uses the govt window to introduce the values for policies and political structure he desires and then forgets about it. He needs to go back to the govt window only when he has changed his mind about the values he initially entered. This is so because once the model knows what the player wants, it simulates, from that moment on, the politics between the ruler and the people leading to the definition of policies and political structure. In other words, the model plays the ruler based on the info provided by the player. The govt model is most of the time working in "cruise mode", demanding player intervention when it's really needed.

3) Changes take time. In a given game turn, the model takes all the info needed for decision-making (polpowers and preferences of social classes and ruler) and computes "final" values for policies and the political structure. But they're not applied immediately. The model smoothly moves the current govt profile to the new one as game turns pass, simulating the actual implementation of decisions. So, if one player wants to increase the PrivateProperty value and assuming he has enough power to do so, the new PP value will be a reality a few turns from the moment of player's intervention, simulating the process of selling State owned economic activities to the private sector.

4) It's important to not misinterpret the concept of "political power" (or just "power"). It indicates one's degree of influence over decisions. Just that. It doesn't represent how much political support or how many "votes" someone has. Imagine a civ where people vote for parliament representatives and where the "democratic faction" gets the majority of seats. You can't, out of that, take any conclusions about polpowers because the parliament in time may have nearly null power over the govt if the ruler is mostly despotic. A ruler may have no support at all from social classes and yet have all political power (and vise versa). The concept of pol.power has a very specific definition and cannot be freely interpreted.

5) To simulate how the ruler and the social classes interact to determine the govt profile, the model uses what I called the "Negotiation System". The NS represents all sorts of interactions between the actors taking part in the decision. All sorts. It includes negotiations ("support me on this one and I'll support you in that other thing"), threats, extortion, alliances, etc. If you try to compare how well the NS resembles a voting situation, for example, you'll find the NS useless. And that's ok, because what the NS is simulating is the voting event plus all what happened between actors before the voting. What I'm saying is the model acknowledge that, as long as there's more than one actor taking part in a decision, there's always some degree of interaction between them leading to some level of compromise between positions, regardless of what is the specific process of decision-making.

6) FSmith asked if the NS could handle "yes/no" type of decisions. It's a good question because as the NS tends to compromises, yes/no decisions admit no compromise. Although the model currently doesn't include binary type policies, it's important to discuss it because we might want to add one. The NS has no problem dealing with binary decisions once you treat variables correctly. For example, if the govt needs to decide on a treaty with other civ facing the options "Peace" and "War", then, in terms of modeling, you have a variable called TREATY that can take a value of zero (peace) or one (war). Since the NS works in a continuos space, it'll return a value between 0 and 1. Assume it returns 0.7 (that would be the case if the actors supporting war have more power than those supporting peace). But since TREATY can't take such value, the number would have to be rounded to the closest integer. In this case, "1" (war). And there you have it.
In other words, you simply need to ensure binary variables take binary values at the end of the process and you can still use the NS. The binary variable is really a particular case of discrete decisions. The NS can work with discrete decisions as long as the options can be sorted using a given criteria. In the case of our example, an "Alliance" third option may exist. Here the sorting criteria should be "friendship level" and the resulting order would be 0:Alliance, 1:Peace, 2:War. By applying the NS you'd get a value between 0 and 2 that rounded to the closest integer would give you the winning option.
However, if the sorting criteria for discrete options is difficult or impossible to define (for example if the govt has to decide on where to spend money in, with options: "building a hospital" vs "investing in tech" vs "buying military hardware"), then the NS fails.

7) Not only the NS can deal with discrete options. It can also produce more realistic results than a voting system. In our last example, let's assume we have 4 actors: the ruler and 3 other actors with powers and preferences:
ruler: 20% power, preferred option: Alliance (binary value=0)
actor A: 10% power, preferred option: Alliance (binary value=0)
actor B: 30% power, preferred option: Peace (binary value=1)
actor C: 40% power, preferred option: War (binary value=2)
A voting system would lead to "war" because actor C has majority. But what would really happen in real life? The pro-peace and pro-alliance factions would had anticipated they were going to lose to the pro-war faction. But together they can beat the pro-war faction. Therefore, the pro-alliance faction, since it wants at least a peace treaty, would had voted "peace" in order to avoid having war declared. The real voting would had been Alliance=0%, Peace=60%, War=40%, leading to a peace treaty. "Peace" would also had been the result if we'd had applied the NS: In the NS you make a weighted sum of positions (excluding the ruler), so a "preliminary decision" is made. The weighted sum in our case is 0*10%+1*30%+2*40%=1.1. Now the ruler changes this decision using a "modifier" (M). Since he wants a value of zero, the NS uses the modifier to reduce the value of the preliminary decision in order to move it as close as possible to zero. Having only 20% of power the ruler has a small modifier. Let's assume M=0.4. Then the final decision is 1.1-0.4=0.7. That's "1" rounded, i.e. "peace".
This is a good example of what I said earlier about the NS simulating interactions between actors (in this case a joining of factions). And that's something the voting system can't do by itself.

8) In the above example you can see that if M is high enough, then the ruler can simply override any preliminary decision and do what he wants. You can compute M any way you want, as long it's an increasing function of ruler's polpower. The higher ruler's polpower is, the higher M is. A rather simple manner to compute M is rulerpower*(U-L), where U and L are respectively the maximum and minimum values the policy can take (U=2 and L=0 in the case of the Alliance/Peace/War example). But note you can define the function relating M and ruler's polpower in other ways to incorporate other things. For example, FSmith has said that having 51% power should be enough to drive policies any way you want. Although I profoundly disagree with this, we can implement this philosophy within the NS if we want. You only need to define the function in such way that when ruler's polpower takes the value 51%, M takes the value U-L.

9) Another NS characteristic is the ruler doesn't play with exactly the same rules as any other actor (the ruler plays at the end, modifying a preliminary decision of others). I think it was Yoav who said this shouldn't be so. That the player should play in equal conditions. Although I agree that would be great, I don't care much about "purity" in models. I only care models give you the right outcomes. In terms of game-play, what the player should experience is that the higher power he has, the more he can impose his view and vise versa. And when he has low power, the resulting govt profile should be more sensitive to those actors with highest power, yet incorporating interactions with less powerful political forces through different degrees of compromises as it happens in real life. The NS provide these things.

10) The govt model tries to simulate politics. And in real life politics you rarely get a white or black solution. You always get shades of gray. Even if the decision itself or the decision-making mechanisms implemented in a given regime seem to polarize things and restrict politician options, there're always interactions between political actors, many of them behind scenes, allowing greater flexibility than it appears to be. And that's valid for a modern democracy and for elder tribal leaders around a fire.

11) Finally, consider how amazingly simple in computing terms the NS is. In order to achieve similar results with other systems like the voting system, you'd have to "teach" social classes how to strategically vote, who to join forces with and when (remember the Alliance/Peace/War example). And that can become a tough work.


Ok. Sorry for the long post. I know some of you may remain unconvinced with the system. Repeating what I said before, I'd like to see all criticisms in terms of examples of historical frameworks where the model would not produce the right outcomes. Let's try to make a concrete list of situations where the model fails. Otherwise this discussion will just go on and on endlessly.

On the other hand, I'd also appreciate comments on the "Administrating the Empire" chapter of the new model version.

[Mark_Everson]

Hi Rodrigo: [Copied over from Govt Model v.2 (contd.)]

I think you did a great job fixing all those old issues in the new government model! I've only got a few minor comments.

You did a really good job with feudalism! I think the flexibility to handle it will help out the game a lot! As a quibble, I think that feudal units should be available for temporary use by the civilization if their use is requested, and granted by the local authorities.

One thing I didn't like is that you have the Province Autonomy Level set by the ruler. I think this should be negotiated like everything else in the system. My concern is... suppose you have a modern democracy. If I have read your system right so long as there are no feudal troops the player can just declare all provinces in the civ to have a 0 autonomy level. That is OK for a dictatorship, but such things don't happen in democracies were there is typically a federalized system sharing the power geographically as well as demographically. I could go on about this, but I want to hear whether I have misread your approach first.

And I really preferred the old name "Ideologies" to Regimes. But I can see where you did a because some of the things are really regimes, whereas others are ideologies.

I am hoping we can get the model coded as soon as possible! Unfortunately, Gary, who is the one who is going to do it, has a lot of other coding responsibilities too. Still, I hope he's as excited about it as I am.

Great Job,

Mark

[Mark_Everson]

[Copied over from Govt Model v.2 (contd.)]

I have a few more comments from a scrap of paper I discovered from a preliminary version of the model a ways back.

I dislike the notion that religions are the sole purveyors of ethics. Can we not simply put in a "religion" called "philosophy" or "humanism" or some such? That could be a catchall that allowed non-religious states to nonetheless get some of the benefits that the model ascribes to religion. I guess philosophy would not have a holy land that people would be willing to die for however .

I also don't like the fact that Regimes/Ideologies don't have a bureaucracy power. Admittedly foremost historical regimes the goal is very small bureaucracy power. However, I don't think that's the case for the Confucian meritocracy in China. I expect there are other examples also. And the fix is easy, just put in the number, and set it very low for most Ideologies.

I also thought a little bit about class power being overcounted in your approach if several of the numbers were very high. FE A = 100, W = 90, K = 90 etc. could you comment on that?

[Yoav Sissman]

roquijad:

5) To simulate how the ruler and the social classes interact to determine the govt profile, the model uses what I called the "Negotiation System". The NS represents all sorts of interactions between the actors taking part in the decision. All sorts. It includes negotiations ("support me on this one and I`ll support you in that other thing"), threats, extortion, alliances, etc. If you try to compare how well the NS resembles a voting situation, for example, you`ll find the NS useless. And that`s ok, because what the NS is simulating is the voting event plus all what happened between actors before the voting. What I`m saying is the model acknowledge that, as long as there`s more than one actor taking part in a decision, there`s always some degree of interaction between them leading to some level of compromise between positions, regardless of what is the specific process of decision-making.

How does the model stimulates negotiations and alliances? It seems that the model considers measures that increase the de facto polpowers of factions, but they don`t actually make `back room` deals, which are hard to consider and require micromanaging.

The riots model allows for unhappy factions to object to government decisions, and the political players are likely to consider that when they make THEIR decisions. This is why, IMHO, the NS itself should be designed to only represent the power that the gov`t itself (including all of its branches and institutions) grants to the political players. Historically gov`t decisions are made through some sort of a rule of majority or some other rate of the relavent voting pool (be that a modern parliament or a bunch of eldery sitting at a campfire). The NS shouldn`t artificially increase the power of majority or of the extremists.

In a situation where the gov`t has the power to enforce its decisions and is willing to risk social upheavel, a group that holds 51% of the polpower can do pretty much everything it wants. FE in 1933 the Nazi party formed a coalition with another extreme-right party, and used a tight majority of the German parliament seats to conduct a full-scale revolution by legal means. The current NS doesn’t allow this; All that the fascist factions will see when they pass the 50% boundary is a slight movment in the government profile towards their idealogy and policies.

A voting system would lead to "war" because actor C has majority. But what would really happen in real life? The pro-peace and pro-alliance factions would had anticipated they were going to lose to the pro-war faction. But together they can beat the pro-war faction. Therefore, the pro-alliance faction, since it wants at least a peace treaty, would had voted "peace" in order to avoid having war declared. The real voting would had been Alliance=0%, Peace=60%, War=40%, leading to a peace treaty.

This is exactly the idea behind my median-based NS – the most moderate determine where the balance point will be.

As in reality, the party that supports a 3% tax-cut but not 5%, and its vote is the deciding one, will have its way. The actual tax cut won’t be the weighted sum of the opinions of all members of the house.

[QUOTE9) Another NS characteristic is the ruler doesn`t play with exactly the same rules as any other actor (the ruler plays at the end, modifying a preliminary decision of others). I think it was Yoav who said this shouldn`t be so. That the player should play in equal conditions. Although I agree that would be great, I don`t care much about "purity" in models. I only care models give you the right outcomes. In terms of game-play, what the player should experience is that the higher power he has, the more he can impose his view and vise versa. And when he has low power, the resulting govt profile should be more sensitive to those actors with highest power, yet incorporating interactions with less powerful political forces through different degrees of compromises as it happens in real life. The NS provide these things.[/QUOTE]

I don’t know what you mean by ‘purity’ in models, of course we only care about getting the right outcome. The current player interface increases the maximal effect of the player with the ruler polpower, but does it in an arbitrary manner. There are foundamental differences between the way the player is treated and the way that other political players are treated:

(1) The player is enabled in some cases, if he’s powerful enough even though not despotic, to push the government profile all the way towards his preferences. I consider this to be a good idea (and this is what will happen in my NS), but this isn’t the case for any of the other polpowers.

FE if the people have 99% of the power and their desired PP policy is 15, while the clergy holds the remaining 1% and wants PP value of 16, the people still won’t get their way, and the gov’t profile will be set to 15.01. This may be neglegable in that example, but depending on the function we use to set M (the ruler factor) the level of discrimination between the player and the other blocks will probably be high.

(2) M depends only on the ruler polpower while the actual effect of any of the other blocks on the preliminary gov’t profile is the result of multiplying its power with the distance between its stance and that of the preliminary profile calculated without considering it.

FE let’s assume that the ruler holds 20% of the power and 4 of the other PBs, including the people, hold 20% each as well. The ruler want PP to be 70, the people want it to be 30, and everybody else wants it at 50. The preliminary gov’t profile will be set to 45, and then, depending on the function we use, that value will be updated to the final policy (it seems you suggest it will be higher then 50). The point is that it doesn’t matter if the ruler sets his preffered value to 70, 80, or 100. In any case the DNP value will be 45+M. OTOH if the people will only lower their optimal PP level from 30 to 10, they can lower the preliminary profile policy from 45 to 40, and hence push the final result down by 5 points as well. Under those conditions, they will be motivated to ‘lie’ to their theoretical machine and claim they want the PP value to be 0.

11) Finally, consider how amazingly simple in computing terms the NS is. In order to achieve similar results with other systems like the voting system, you`d have to "teach" social classes how to strategically vote, who to join forces with and when (remember the Alliance/Peace/War example). And that can become a tough work.

I agree. This is why I object to the voting system.

The only advantage I can think of that the current NS has over my suggested one, is that the weighted sums system is more comfortable for the AI then the less steep median one.


From the model:

a) For optimum administration effectiveness, you need X units of ININ per inhabitant. X is given by the sum of a Base plus values that depend on Social Policies and Private Property. The higher SP is and the lower PP is, the higher X is. That`s because you need a bigger bureaucracy to support social policies and State held economic activities.

I believe it will be better not to implement a maximal admisitration effectiveness, but to allow the player to see dimished returns from his investments no matter how many units of ININ per inhabitant he builds.

Maybe the decision over this doesn`t belong in this model.

[roquijad]

Mark: you have a good point in Province Autonomy Level. Although I don't think the ruler should negotiate the PAL for each individual province, I think it's reasonable to negotiate the maximum PAL (valid for all provinces).

I dislike the notion that religions are the sole purveyors of ethics. Can we not simply put in a "religion" called "philosophy" or "humanism" or some such?

that means you didn't like my new idea of "philosophy" in the new social model? (your criticism goes in the social model, so let's move this conversation to the social model thread)

Changing the names from ideologies to regimes was more because some people weren't understanding ideologies as forms of govt.

On bureaucracy: I disagree. In the model the bureacrats are the ones that perform the administration. They're not supposed to take the big decisions (although in reality they do interfere with them). As far as I understand the case of China, confusianism implemented in administration had more to do with the high morality and professionalism you'd like to see in administration rather than leaving the big decisions to them. In fact powerful public servants often took big decisions for the ruler, but in theory they shouldn't. Remembering regimes/ideologies are an idealization of a govt form, bureaucrats should never have power.
BUT, I'm willing to include bureaucracy in regimes if others think like you do. It's just that I'm trying to leave within regimes only the strictly needed variables because the more variables they include, the larger the pool of regimes needed to cover all the spectrum of possibilities.

I also thought a little bit about class power being overcounted in your approach if several of the numbers were very high. FE A = 100, W = 90, K = 90 etc. could you comment on that?

I don't understand well your worries. First, remember the social roles values are relative. A social class defined like the one in your example would be powerful as long as other social classes are defined with numbers like A=1, W=0, K=20. Per se, numbers don't mean nothing. They have to be seen relative to others.
On the other hand, if the scenario designer indeed set one social class with high numbers and the rest with low values, then he's actually designing the game with a social class that tends to prevail over the rest in all fields, so there's no contradiction.
Finally, remember actual power depends on social roles values AND demographics, so a social class like yours may at the end have little power if other factors determine it has a very small population compared with other classes.

Yoav:

How does the model stimulates negotiations and alliances?

I think the Alliance/Peace/War example is very clear about that.

IMHO, the NS itself should be designed to only represent the power that the gov`t itself (including all of its branches and institutions) grants to the political players. Historically gov`t decisions are made through some sort of a rule of majority or some other rate of the relavent voting pool (be that a modern parliament or a bunch of eldery sitting at a campfire). The NS shouldn`t artificially increase the power of majority or of the extremists.

Why you don't let the model simulate what really happens in politics? Real politics is not just about how much legal power you have. It's not only about how many votes you control. The govt model has to include some way or another what happens between actors beyond the particular mechanism of decision-making. The NS achieves that, so I don't see any use to separate the processes of interactions between actors from the actual process of taking a decision. The bottom line in politics is actors use any form they can to influence a decision. Politicians don't simply go to parliament and vote.
The NS doesn't "artificially" increase power to any body, it just includes all sources of power/influence, legal or not, in the final decision, as it happens in RL.

Your hitler example is good, but I consider it an exception rather than the rule. The US has had times with an elected republican administration and at the same time with a republican majority in congress, yet the country's laws and policies have never been fully republican. If you check similar situations in other countries, past and present, you'll find majority is not that overwhelming force you speak about, able to ignore the rest and get its way in everything.

The point is that it doesn’t matter if the ruler sets his preffered value to 70, 80, or 100. In any case the DNP value will be 45+M.

And you think this is wrong? The ruler is limited by how much power he has. His ability to alter the govt shouldn't depend on how extreme he presents himself, but only on how much power, legal or not, he has.

FE if the people have 99% of the power and their desired PP policy is 15, while the clergy holds the remaining 1% and wants PP value of 16, the people still won’t get their way, and the gov’t profile will be set to 15.01.

A similar criticism has been given in the past by Richard Bruns, if I remember correctly. His opinion was in real life there're some parties with so small power that they can't really affect policies. Although I find it questionable (you yourself have said how important can be the "deciding vote", therefore giving these parties a negotiation power for other things), it's very easy to incorporate it in the NS. Instead of using a weighted sum, you have to use an exponential weighted sum. For example, assuming only two actors, the weighted sum calculation would be:

polpower_A*opinion_A + polpower_B*opinion_B

while the exp. weighted sum would be:

(exp(K*polpower_A)*opinion_A + exp(K*polpower_B)*opinion_B)/(exp(K*polpower_A)+exp(K*polpower_B))

where K is a constant allowing you to control how big polpower differences should be to simply "ignore" the least powerful actor.

With an appropiate K, you'd get in your example a result like 15.000000001, i.e. 15 for any practical use.

Anyway, the ruler is always able to affect the decision because he acts after the exp. weighted sum. If you were trying to say even the ruler shouldn't be able to affect the decision if a faction is powerful enough, then I must disagree. It's a game after all, and the player can't be totally left aside of govt decisions.

You say your system is better. I think a good way to prove it is presenting it here formally, step by step, and applying it to a couple specific examples. In parallel, you can show the results of my NS, pointing the differences, pros and cons. Please include in my Alliance/Peace/War example in your analysis. For the "M function" of the NS you can use the one I proposed in my last post.

[Simon Loverix]

The player should indeed have some influence at all times. When your power is small and traditionalism is high, the game will get stuck. Same goes for the AI.

Extremist and other minorities do have an influence. FE in my country we have an extreme right party. All other parties have repeatedly and publicly vowed never ever to form a coalition with them, yet they are giving considerable attention to their programme, in the hope they will lose influence and votes.

[LDiCesare]

I didn't see any formula about the ruler's modifier, so I propose one:

Here is one that I propose that actually gives the ruler the same power as anyone else, based on their polpower:
Supposing you want to negotiate someting on a 0-100 scale. Consider all other voters have polPower OthersPP (in %), and want a value of V. This was obtained by a weighted sum. The ruler could vote 0 or 100 to machiavellically (?) obtain his goal, so in a vote system he could influence (100 - OthersPP)% up or downwards: If he voted 0, he'd get V * (OthersPP) as a result, thus a reduction on the low scale of V * his pol power. On the higher side, he can get up to (100 - V) * his pol power.
The point in this proposal is that it doesn't give the ruler more power than his share, and, I think, is the only one that can achieve that: otherwise, ruler polpwer cannot be compared to other factions polpowers since they'd not mean the same thing, and thus needn't sum up to 100.
Examples: For 50% ruler pol power, a base value of 50 can be modified by +- 25.
For 50% ruler power, a value of 0 can be modified -0 to +50,
For 50% ruler power, a value of 25 can be modified -12.5 to +37.5
For 20% ruler power, a value of 25 can be modified -5 to +15.
roquijad, does that fit? Yoav, it seems this goes as far as the model allows to not have a different role for ruler and the rest.
What do you think? (now I go back coding combat)

[Yoav Sissman]

I'm sorry it took me a while to respond. I've been tied up.

roquijad:

About negotiations and alliances:

Why you don't let the model simulate what really happens in politics? Real politics is not just about how much legal power you have. It's not only about how many votes you control. The gov't model has to include some way or another what happens between actors beyond the particular mechanism of decision-making. The NS achieves that, so I don't see any use to separate the processes of interactions between actors from the actual process of taking a decision. The bottom line in politics is actors use any form they can to influence a decision. Politicians don't simply go to parliament and vote.

I'm sorry, I simply don't see how your system combines the process of interaction between actors and the actual process of making a decision where my system doesn't.

We can't compare modern politicians with the political actors of the game. Modern politicians act as representatives, and interacting with their voters take a significant part of their time. True, there are other forums where they can make a difference other then the general assembly, but the rule of majority is always valid.

Your hitler example is good, but I consider it an exception rather than the rule. The US has had times with an elected republican administration and at the same time with a republican majority in congress, yet the country's laws and policies have never been fully republican. If you check similar situations in other countries, past and present, you'll find majority is not that overwhelming force you speak about, able to ignore the rest and get its way in everything.

In reality the Nazi revolution may be a unique event, but still it's a good example of what happens when the majority doesn't care about social upheaval and the people are obedient to their government decisions.

In your Republican example a few factors came into play:

(1) The Republicans may be all members of one party, but in reality they vary a lot in their opinions (it's not that there are only two different opinions in the USA). Since the Democrats had a substantial part of the power, the median point belonged to the moderate Republicans.

(2) The administration slowed down the change, as we assume in the model they support the current gov't profile. In a modern state the administrative system is powerful and changes are slow. The Republicans probably weren't in power long enough to convert the system entirely.

(3) The 'real' Republican views aren't necessarily the ones they advocate. In my country an opposition leader recently came into power and failed to comply with many of his pre election promises. To defend himself he stated, "What you see from there isn't what you see from here."

I think that the profile of the different factions should be the one they'll try to implement if they got the chance, not the alternative they present to come into power.

(4) Traditionalism and civil disobedience all slowed the change further, but they aren't part of the gov't model.

And you think this is wrong? The ruler is limited by how much power he has. His ability to alter the gov’t shouldn't depend on how extreme he presents himself, but only on how much power, legal or not, he has.

Now you're just being unfair... I said:

The point is that it doesn?t matter if the ruler sets his preffered value to 70, 80, or 100. In any case the DNP value will be 45+M. OTOH if the people will only lower their optimal PP level from 30 to 10, they can lower the preliminary profile policy from 45 to 40, and hence push the final result down by 5 points as well. Under those conditions, they will be motivated to ?lie? to their theoretical machine and claim they want the PP value to be 0.

I don't object to the way the gov't profile isn't dependant on how extreme the ruler present himself. My objection is in the next sentence (which you left out of the quote), where I show how the model IS dependant on how extreme the OTHER factions present themselves.

My suggestion is that we eliminate that 'discrimination' from the system and make it unnecessary for the other PBs as well to 'lie' to the system.

A similar criticism has been given in the past by Richard Bruns, if I remember correctly. His opinion was in real life there're some parties with so small power that they can't really affect policies. Although I find it questionable (you yourself have said how important can be the "deciding vote", therefore giving these parties a negotiation power for other things), it's very easy to incorporate it in the NS. Instead of using a weighted sum, you have to use an exponential weighted sum.

I too think that the power distribution should be proportional to the polpower, and not exponential. My criticism was directed at the fact that the political players that hold 99% of the power can't have their policies exactly, not at the way that small parties have an effect.


IMHO if there will be 51 different 'right wing' factions, each with 1% power, and one 49% 'left wing' faction, the gov't profile should be very moderate right (and independent on the exact stance of the extremes).

Anyway, the ruler is always able to affect the decision because he acts after the exp. weighted sum. If you were trying to say even the ruler shouldn't be able to affect the decision if a faction is powerful enough, then I must disagree. It's a game after all, and the player can't be totally left aside of gov’t decisions.

Then maybe we should leave a big enough role to the ruler on every ideology. I think that in reality the president/PM or any other historical ruler did had significant effect (assuming he was competent).

You say your system is better. I think a good way to prove it is presenting it here formally, step by step, and applying it to a couple specific examples. In parallel, you can show the results of my NS, pointing the differences, pros and cons. Please include in my Alliance/Peace/War example in your analysis. For the "M function" of the NS you can use the one I proposed in my last post.

The formal procedure I suggest, in case it wasn't yet understood, is to calculate the amount of support each opinion on the DNPs has, using the current method. The power of the ruler and each other PB is given by the regime, and the power of each faction of each social class is computed through his relative influence on each PB.

Once we computed the polpower behind each suggested value to each of the DNPs, we forget all about PBs and chose the median point for each of the policies separately.

For the alliance/peace/war example you presented there is no difference between the final result that the two systems produce, but let me present an alternative one:

(1) The Example:

Italy in WWI is part of an alliance that looks like it's going to lose the war. Initially it's allied with Germany and in war with England (FE). Some of the polplayers (I rather not get into the question of who exactly they were) want to stay loyal to their current allies, while others want to change sides and team up with the allies to avoid defeat and win the spoils of war.

As the odds pile up in favor of the allies, the voices in favor of joining the former enemies in a war against the former ally intensify. I'll assume that at some point the latter win the majority of influence, and let's assume that 60% want war with Germany (at least as a diplomatical measure, even if in reality not much need to be done). The rest want to stay loyal to their Crouch friends. Only few want to declare neutrality, as that is likely to upset both sides.

In my NS:

60% want relations of 0 (alliance) with England and 2 (war) with Germany. 40% want the opposite. The median point is in warring with Germany and joining the Allie's alliance. The gov't decision may or may not be accepted by the army (regardless of the level of influence it has on the gov't) but that is part of the Riots Model.

In your NS:

I'll assume that the ruler has 30% polpower and is one of those that support breaking the old alliance and forming a new one. The preliminary gov't decision regarding the relations with England is: [0*(60%-30%)+2*40%]/[100%-30%]=1.14
M=(2-0)*30%=0.6

The final decision is 1.14-0.6=0.54, round up to 1 (peace).

The decision regarding Germany:
[2*(60%-30%)+0*40%]/[100%-30%]=0.86

The final decision is 0.86+0.6=1.46 round down to 1 (peace).

Italy implements an unpractical policy of neutrality after already being part of the war.

I'd like to state that I'm not sure of the historical truth that's in this scenario... But I think it's at least conceivable.

Also I'm not sure that the decision regarding alliance/peace/war should be made in this manner, but I took it that you've allowed me to assume if for the sake of argument.

Note that a polpower of over 75% will allow the ruler to completely control foreign relations (as M>1.5).

(2) The Example:

This is a theoretical example presented by me in my last post. Since you seem to agree that the de facto influence of a faction shouldn't depend on how extreme it presents itself, I would like to expand it to show what I consider to be the main flaw of the current system.

The people PB have 20% polpower; they really want PP value to be 40. The ruler has 20% polpower; he really wants PP value to be 60 (and input in his preference tool). Everybody else wants it at 50.

My NS:

The PP value is set to the median, 50.

Your NS:

Preliminary gov't profile = (20%*40+60%*50)/80% = 47.5
M = 100*20% = 20.
The final DNP = min(60,47.5+20) = 60

The ruler gets his way, even though is power is supposedly equal to that of the people.

But now what will happen if everybody maintain their power but the people, frustrated from their lack of affect, decide to 'lie' to the engine and demand PP value to be 0?

Preliminary gov't profile = (20%*0+60%*50)/80% = 37.5
The final DNP = min(60,37.5+20) = 57.5

That's a little better for the people. They presented extreme views to increase their de facto influence. Assuming the AI controlled political players aren't blind to the principles of their political system, they will usually set their preferences to either 100 or 0.

Simon Loverix:

Extremist and other minorities do have an influence. FE in my country we have an extreme right party. All other parties have repeatedly and publicly vowed never ever to form a coalition with them, yet they are giving considerable attention to their programme, in the hope they will lose influence and votes.

I agree that they have an influence; I just say that it shouldn't be increased BECAUSE they are extreme.

In my NS both extremes and moderates will have the same impact on the gov't profile, as they both have the same chance to push the median in the direction of their preference. In your example the extreme right party stayed out of the coalition itself, but increased the overall power of the right wing block, and possibly caused a less extreme right party to be included in the coalition.

IMO that's consistent with my approach.

LDiCesare:

I didn't see any formula about the ruler's modifier, so I propose one:

Here is one that I propose that actually gives the ruler the same power as anyone else, based on their polpower:
Supposing you want to negotiate someting on a 0-100 scale. Consider all other voters have polPower OthersPP (in %), and want a value of V. This was obtained by a weighted sum. The ruler could vote 0 or 100 to machiavellically (?) obtain his goal, so in a vote system he could influence (100 - OthersPP)% up or downwards: If he voted 0, he'd get V * (OthersPP) as a result, thus a reduction on the low scale of V * his pol power. On the higher side, he can get up to (100 - V) * his pol power.
The point in this proposal is that it doesn't give the ruler more power than his share, and, I think, is the only one that can achieve that: otherwise, ruler polpwer cannot be compared to other factions polpowers since they'd not mean the same thing, and thus needn't sum up to 100.
Examples: For 50% ruler pol power, a base value of 50 can be modified by +- 25.
For 50% ruler power, a value of 0 can be modified -0 to +50,
For 50% ruler power, a value of 25 can be modified -12.5 to +37.5
For 20% ruler power, a value of 25 can be modified -5 to +15.
roquijad, does that fit? Yoav, it seems this goes as far as the model allows to not have a different role for ruler and the rest.
What do you think? (now I go back coding combat)

I proposed something similar in the previous thread.

Unlike the constant M approach, it consistent with the Weighted Sum System, but it still have the cons of that system as a whole (that is, amount of power is exactly proportional to the level of extremeness).

Also if we chose to implement it with the goal of allowing the ruler and other political players to compete on even grounds, we should allow everyone the benefits of artificially making his policy more extreme. As it is now, the ruler gets a substantial advantage.

[Mark_Everson]

Hi All:

I'm Not stepping into this argument again ...

Gary made a great (IMO) suggestion in a recent email discussion on overall coding planning. His idea was to port over F_Smith's old Govt demo into the new code architecture basically as-is for a first step toward the goal of having govt and social models functioning by D6. Of course much of it (the stand-alone part like map etc.) would be discarded. This could be done quite quickly, and will get much of the govt/social system going so people can play around with it. Then, later in the D6 process the old model could be upgraded to the new one.

What do you think Rodrigo? Is this ok with you? Like I said it seemed like a really good quick-start approach to me. And since the 51% stuff is already in there, maybe it'd get Yoav off your back since he can try it for himself . What do you think Yoav, any chance .

I now return you to your regularly-scheduled fracas.

[Yoav Sissman]

Tee-hee, I'm not familiar with F_Smith's old Govt demo but it sounds like a useful approach.

BTW does anyone agree with me on the NS by now? I'm not here just for theoretical discussions, and if I can't persuade anyone then I don't want to waste Rodrigo's and mine's time.

[LDiCesare]

Actually, Yoav, I rather agree with your viewpoint. I wouldn't discard rodrigo's NS without giving it a try though. His system has a point since the ruler, even with small PP, can influence things, while with a median system, he will probably be unable to switch the vote even an inch towards his direction. Btw, I like your Italy example. I'd like to hear from Rodrigo about it.

[Mark_Everson]

I said I was going to stay out of this one. I lied, I guess.

I prefer the standard method. It gets a lot of effects right in terms of real power politics that the voting method IMO won't. That is because the median effect, to be done right, would require a superb AI with lots of game theory type information embedded in it. We aren't going to go there... And the median approach is unstable, verging on chaotic. Specifically, if a very small change in input can produce a huge change in result then the behavior of the system is very difficult to predict. Writing good AI for it would be almost impossible IMO.

Also, a lot of the radical shifts made possible by the median method will enable dramatic swings in the riots and revolts level for the player. Especially for governments without a long committment to democracy and the rule of law. FE with 51% of the power you can pick any course in the median method. If you pick 100% and the other 49% of the power wants 0% on a critical issue, you will win the vote, and then have your society explode in rebellion. I don't think this is desirable.

For these reasons I think, at least for AI, something like Rodrigo's approach is mandatory. OTOH the median thing is Very simple to put in as an option for Just the player. (But the player wouldn't be able to turn ruler govt activity over to the AI for the reason I mentioned above).

I strongly support Rodrigo's approach since I think it gets a lot of things right on balance, without the possibility of wrenching changes in the government that make good AI almost impossible.

[Simon Loverix]

In reality the Nazi revolution may be a unique event, but still it's a good example of what happens when the majority doesn't care about social upheaval and the people are obedient to their government decisions.

If the people are obedient (and you have a loyal army) it's evident that you don't have to worry about social upheaval.
The Reichstag during the Nazi's gave all people equal votes, so it represents only one social class, one portion of political power. There were also the army, the clergy, the industrials,.. who supported or at least tolerated the regime, but they didn't have votes equal to their actual influence on the country.

It seems that an exponentially weighted sum will get rid of the marginal extremist as well as preventing brusque changes.

[Yoav Sissman]

LDiCesare:

His system has a point since the ruler, even with small PP, can influence things, while with a median system, he will probably be unable to switch the vote even an inch towards his direction.

That really depends on the number of different policies the political players has. The more diverse the system is, the less blunt the effects of the shift of power in the political arena will be.

FE if there are only two different factions then the median system actually becomes an absolute majority system, but if there are 100 different factions then the ruler is likely to see a pretty smooth change in his influence as his power increases/decreases.

Hopefully the many combinations of ideologies and of stances on the various DNPs will create a sufficiently sophisticated political arena, where a slight change in input will not result in a big change of the overall result (there I'm starting to address Mark's comments ).

Mark:

You make some very important points.

As I understood the distribution of power that allows a faction to riot doesn't necessarily represent the political distribution of power in the government (though it have some effect through the de-facto adjustments system). The government can decide on a critical issue with an outstanding majority and still see the social fabric explode in rebellion, because certain classes are severely underrepresented in the government.

In most cases the condition shouldn't be of a strict 100 policy vs. 0 policy. Normally in RL the center of the political map is a lot more highly concentrated then the extreme sides.

In a case where the political strength indeed represents the 'rioting power' and mostly everyone has extreme views (FE if 51% of each class want a value of 100 for a critical decision and the remaining 49% want 0, so not only the government is split around it but the entire civilization is) then by all means I believe that what we should have is a civil war or something of that kind.

In a case that IMO wasn't as extreme - when the Americans couldn't settle their disagreements concerning slavery, they decreed the matter (not to allow slavery in the west) using their standard government decision making methods, only so the losing side can forget all about accepting the decision of the majority and declared independence (and then the matter was resolved through sheer force).

If the player has over 50% of the polpower then IMHO he should be able to dominant the government. If he completely ignore the interests of factions who can still assassinate him, perform a military coup, rebel, or whatever, then that's his problem... I don't see why that should mean that the player will have to micromanage the government model to survive.

I don't know much about the current progress being made over at the AI section (not much going on the AI forums for me to see I'm afraid ), but I would like to see the AI give some consideration to the rioting capabilities of the civilization as well, otherwise it can have problems avoiding riots when we use roquijad's NS as well as when we use mine.

If we find that giving TOTAL control for 51% polpower holders seem extreme, we can introduce an addition to the model that will make it more complicated but may prevent that.

We can decide that after a radical change in the regime the civilization gets a 'constitution' like document, which state that the government profile concerning some of the decisions can never go below or upper certain values, which have a wide consensus in the time the constitution is wrote, unless the gov't have some special majority (75%?).

We can make constitutions an optional part of the game.

Simon Loverix:

It seems that an exponentially weighted sum will get rid of the marginal extremist as well as preventing brusque changes.

A problem I see with the exponential system, at least as I understand you suggest it, is that it won't sit well with the arbitrary nature of ideologies.

FE if we have two similar regimes the people whose views sit well with them will be divided between both, and will see their power reduced.

In reality OTOH there are no discrete regimes so there is no such effect.

[LDiCesare]

I think we should use Rodrigo's model, plus the median as an option for the player, in which we display the NS proposed range to see the difference between the models.

I'll take a shoot at why the NS system may be hard for AI too.
The way you assign values to actions has a huge influence on the system, but none on the voting system:
Having different war, peace, alliance values lead to very different results, whereas the median system doesn't require fine-tuning of the values. For a 3-value, it may seem silly, but what if we model more complex diplomatic states like war/cease-fire/truce/peace/right of passage/defensive alliance/offensive alliance (values 0-1-2 become 0-3-5 in the Italy example) it is enough to change the outcome. You can add or remove however intermediary states you like in order to get a final result of what you want (or change values, like say war = 0, peace = 1, alliance = 10).
So, for things like tax rates, I think the NS is the best system, but for diplomacy, I don't like it because it will require tweaking figures a lot when we only have an order relation. Adding new intermediary states would totally change the AI behaviour unless you add them in between existing values, which will be arbitrary.