Social Model v3.0

[yellowdaddy]

RF = the people no longer have any loyalty to their political system, they want to change it.
Their values and the values of the political system have diverged, the people behave as Pro-Active acters their divergence equates to Seperatism - because they want to seperate from the values and goals of the system.
The divergence (Seperatism) leads to the change.

RRF same thing. This kind of Seperatism you can describe as "Divergence of Values" or "Change of Allegience"

So yes, you can describe it as change; but also more specifically as a form Seperatism or divergence - that tells you what kind of change it is.

ADF = as above, the change in policy it can lead to sounds like a response to this form of Seperatism. ADF sounds like Apartheid, when you get a society forming within a society - this seems like seperatism to me, because they replace state structures with ones of their own; they are Seperating from the state: the political system they feel alienated by, and the leadership that they feel alienated by.

the +4.5 and +1.7 scalar averages come from (+5)+(-1)+(+5)/3 and (+4)+(-3)+(+4)/3

How are modelling NRF at the moment? Can you show me the algorithms/equations? What specific problem do you have with using Edu.&Rights? NRF must have factors affecting how it rises and falls?

That riot-model page seems to say that you can only have one rebellion at a time, and the leader has to invoke it, that sounds a bit naff?

(I don't want that getting a good economics be the only way of reducing unrest.)

Pardonnez moi, je ne comprends pas, repetez s'il vous plait?!

[LDiCesare]

the +4.5 and +1.7 scalar averages come from (+5)+(-1)+(+5)/3 and (+4)+(-3)+(+4)/3

Well 5 -1 +5 = 9, divided by 3 = 3. So that's why I didn't get it.

NRF:

Nationalistic Rebellion Feeling (NRF)
NRF=(1-DN)*Rexp(8,-13,NAT)

where:
DN is a dummy variable taking the value 1 if the govt nationality and the EG's nationality are the same and 0 otherwise.
NAT is EG's Nationalism.

I attach Rodrigo's document with the whole set of equations, as I don't remember whether it's on the web.
My problem with edu and rights is to find them... There's a rights in the government system, so I should be able to use that one. But education is more of a problem. I found some traces of it in the econ code, but don't really know what they relate to. I also believe the current equations are quite complex and think they'd be better off if they were simplified a bit so I want to keep it simple. Education is problematic as it's a tech, whose value goes from 0 to infinity, and there is education infrastructure too whose values (ranges) I don't know well.

That riot-model page seems to say that you can only have one rebellion at a time, and the leader has to invoke it, that sounds a bit naff?

No, that's one civil war. It means you say that a specific rebellion is given the name "civil war" and gives you some options otherwise not available. I haven't coded that yet, nor checked too much what the exact effects are supposed to be, apart from decreasing empire stability.

Pardonnez moi, je ne comprends pas, repetez s'il vous plait?!

Sorry... I ment I want to let the player prevent discontent in several ways: Having a good economics is one, but that's not reliable as you are bound to get some recessions during the game. So I want another way, which is dedicated to the kind of discontent: Lower discrimination for ADF, adjust policies for BPF, etc. For NRF: increasing rights and education.

[LDiCesare]

I multiplied the Nationalistic Rebellion Feeling value by square root(1 - civil rights). That makes it manageable. In the Social Test scenario, Greeks are down to 1% (workers) or 3% (aristocracy). Of course that sounds fishy since they are of the same eg and shouldn't have different values per social class for this event... But it gets something interesting: Reduction of about 30% of the risk of riots if you have 50% civil rights with respect to 0. 75% civil rights mean 50% reduction. The square root rewards extremely high civil rights more than a straight multiplication, and the value can't be set too high because of bad policies feelings and of the lack of total control the ruler has over their civ.

[yellowdaddy]

Well 5 -1 +5 = 9, divided by 3 = 3. So that's why I didn't get it.

Sorry, I didn't explain, it's not an equation, it moves up and down a scale: -5 -4 -3 -2 -1 0 +1 +2 +3 +4 +5

I actually did 5-6+6=5/3=1.7 because it moves between level -1 and back up to level +5, I should have explained, I forgot and did a very sloppy equation!

So the other one goes: 4-7+7=4/3 ('cos the dif. between +4 and -3 = 7)

What does "Rexp" mean?

Surely Education relates to Research and Development?
Rather than in Civ where yuo just adjust taxation, education levels should affect it.

I agree with you about specific feelings for specific areas, I think I jsut mean you need the economic situation as something that has a strong effect on these PAFs.
I think economics usually effects the intensity and rate of growth of feelings, and is often an element of ADF and NRF (associating greed and corruption with or economic hardship on ethnic minorities)

[LDiCesare]

Rexp is a function which is defined as:

Rexp(a,b,c) is equal to 1/(1+Exp(a+b*c))

It's usually used with a = 1 and b a constant so it's 1/(1 + exp(a + variable)). Basically when variable is very negative, Rexp is worth 1, and when it's very big, Rexp is worth 0.

I don't get how your scale works. It looks like the average, I don't know why you're putting 7s here:
You're doing. 4 + (-3 -4) + (4 - -3) = 4.
Actually this is: a + (b-a) + (c-b) = c. Which means only the last value is useful. I still don't get how the 4.5 can be obtained by a division by 3.

Education as it exists in the code right now is something I find weird. It's useful only for econ in order to grant econ sectors more research points, but doesn't affect other techs or activities. Mark may hopefully explain what that's supposed to do and whether it should encompass more activities. Education infrastructure should probably adjust the RPs generated for all activities, but I can't see if/how it's done.

I agree with you that a bad economic situation should be a factor increasing all other feelings.

[Mark_Everson]

Originally posted by LDiCesare Education as it exists in the code right now is something I find weird. It's useful only for econ in order to grant econ sectors more research points, but doesn't affect other techs or activities. Mark may hopefully explain what that's supposed to do and whether it should encompass more activities. Education infrastructure should probably adjust the RPs generated for all activities, but I can't see if/how it's done.

AFAIK / recall education only boosts econ RPs b/c I was using that as a first implementation, and never got any further. The way it's done for econ should be easily extensible to other areas of the game. Education level should also tie into a lot of other game activities, like the effectiveness / productivity of virtually everything from generalship to economic production. The way that I had envisioned for that is that each area would have a formula to define the "expected" level of education (education "infrastructure" which symbolizes how educated the people in a given square are on average). If education is below the expected level it reduces effectiveness, perhaps linearly. If edu is above, it increases effectiveness, but only as a slowly increasing sub-linear function. Maybe like a sqrt or something. As you can see, this approach requires some coding and xml to accomplish, so I don't think it is an immediate goal. Other ideas on how to handle education are welcome.

[yellowdaddy]

I don't get how your scale works. It looks like the average, I don't know why you're putting 7s here:
You're doing. 4 + (-3 -4) + (4 - -3) = 4.
Actually this is: a + (b-a) + (c-b) = c. Which means only the last value is useful. I still don't get how the 4.5 can be obtained by a division by 3.

Absolutely right, my maths = utter bollocks! Dunno what I did or why!

Key:
W/T = Wealth & Trade
E/R = Education & Rights
P/A = Prestige & Allegience

Year 1:
W/T +5
E/R -1
P/A +5
Expectation Scalar: ('+'5-6+6=5/3=+1.7)

Year 2:
W/T +4 (Econ Growth = -1%)
E/R -3
P/A +4
Exp. Sca.: ('+'4-7+4=4/3=1.3)

Exp. Vector = -0.4


The values for W/T E/R and P/A equate to positions on a scale from -5 to +5, to denote a generally positive or negative mood.

You could have:
W/T +3
E/R -4
P/A +2

which equates (on a 0-10 scale) to:

W/T = 8
E/R = 1
P/A = 7

then it makes sense when you do (a+b+c)/3 = 5.3 in this case, converting back to the scale gives +0.3

---------
- does that mean that LdC's NRF coding is at a bottleneck?
with so many goals that aren't immediate, you might want to post an updated list of the top ten that are...

[LDiCesare]

Nationalistic Rebellion Feeling is currently addressable with civil rights, so I guess it's ok for now and we can pour education effects in later to refine the model.

Now I have another problem. I checked Delenda scenario and found out in this one, the problem is Revolutionary Feeling.
I get, with a government profile of:
Private property: 33%
Social policy: 36%
Warfare: 0.8
Ruler: 0.1
Ethics: 0.1.
For the first social group checked:

Check revolutionary feeling
Checking distance
Comparing Regime Divine Monarchy:
Private property: 10%
Social policy: 30%
Ruler: 0.5
Ethics: 0.5
regime support = 0.39485016
distance = 0.18633026
-> paf = 0.39377046
Comparing Regime Athenian Democracy: Regime policies:
Private property: 50%
Social policy: 60%
Capital: 0.2
Warfare: 0.2
Ruler: 0.2
Human: 0.2
Ethics: 0.2
regime support = 0.14280976
distance = 0.18866977
-> paf = 0.14261593
Comparing Regime Theocracy: Regime policies:
Private property: 10%
Social policy: 80%
Capital: 0.2
Warfare: 0.1
Ruler: 0.2
Human: 0.1
Ethics: 0.4
regime support = 0.20040585
distance = 0.24592382
-> paf = 0.20040585
-> total paf = 0.7929027

Which somehow ends up providing around 16% chances of riots.
I have distances of about 0.2, but the equations provide a feeling of almost 1 * support for the regime.
This is because the equation is:
A=Rexp(50,-300,REGD)*S
which is
S/(1+Exp(50-300*REGD)).
The problem is that: With a distance of 0, we have a value of S/(1 + Exp50) = roughly 0, ok.
With a distance of 1, we have roughly S which is ok too.
But for 0.2, we get S/(1 + Exp (-10)).
Exp(-10) is less than 0.001 soit's almost S.
To get S/2 one needs a distance of 1/6 = 0.167. With 0.188 I'm already almost at S. This means the equation is more or less a step: Either the distance is below a threshold and there's a high likelihood for riots, or not, and there's no riot. The problem is the threshold is at about 0.17, which is way too small considering the support shares above. There's no way people knowing the three regimes above can ever be happy with any one of them as there'll always be support for each of them and the mean regime will always be farther than the threshold.
So the model doesn't work for me as is. I'm going to replace the Rexp thingie with a simple multiplication. That should make things manageable.

The more I'm struggling with the model, the more I find it opaque and hard to react to as a player. Each feeling has effects, but most feelings come from a simple basic factor which is difference of policies (negotiated or regime) and power distribution between the people and the ruler/government. Explaining to the player what's wrong is also pretty complex given the accumulation of factors: To explain a PAF, I must go back to the cause of the paf, here the regime distances, and I'd have to show all of them but to that end, I'd have to store that information which is not otherwise useful, so that's a pain.
I'd rather show what every social group wants in terms of policies and regimes, but that's going to end spreadsheet like..

[yellowdaddy]

...NB, NRF needs the amplifying effects of Economic growth (or sudden lack thereof) (as opposed to PWF).

- - - - - - -

It sounds like you think it needs simplifying, looking at the Govt. model I might agree.

I wonder whether you should only have three social classes:
Rentier (RC), Professional (PC) and Operative (OC) (the latter including slaves) as mentioned earlier in this thread.
Your 6 become:

Military = RC (, PC) (and OC at NCO level)
Ethics = PC
Admin = PC
Kapital = RC (, PC)
Labour = OC
Human = OC

Your Military Leaders and Kapital Owners are the Rentier Class;
Your Clerics/Ethics and Admin are the Professional Class;
Your Labour and Slaves are the Operative Class.

with a knock on effect on your PBs.

[LDiCesare]

The number of social groups is not the problem. Most of the government model is ok, but the riots model is really hard to grasp: In fact, reducing the number of social groups would have about zero effect here as there are just 3 social classes in my scenario (aristocrats, workers, warriors).

I just checked my problem with Revolutionary Feeling again. I told the ai to go with whatever policy people want. With that, they riot. If I change by hand and give them what a minority wants (a regime), they don't riot. This means I have to show the player the possible regimes knowing that the player may define his own values, and show each possible value and the support for each share. That's a lot of panels to know what people may want. And the ai would have to pick a regime among these or do lots of computations to find a good solution. I do not like the idea of browsing maybe 10 possible regimes so I could check which one I must mimick in order to get a good result.

Additionally, the equations are so complex that spotting an error in them is almost impossible.

Finally, the way riots happen is a very long chain:
People want policy X.
This means they feel a Revolutionary Feeling and/or Bad Policies Feeling and/or Replace Ruler Feeling: The corresponding "PAF" are computed.
All PAF are then added together, with variopus weights, to give the probability of an event being called. For instance Riots = PWF + RF + ADF + NRF + 0.4*SDF + 0.6* BPF + 0.1* RRF and Army Betrayal is: RF + 0.2*BPF + 0.2*PWF.
So in order to reduce riots I must act upon any one of the various PAFs. I'll only show the biggest one to the player, as showing all six values is a real pain. Particularly for RF which is actually subdivided in several "I'd want Regime X" feelings.
Getting a feedback to the ai or the player is a real pain with that model. I can't find any way of doing it without resorting to an inane amount of spreadsheets.

I'd rather have a simpler model like:
Discontent 1 = (cause 1) * (aggravator).
Discontent 1 leads directly to a bad effect, bad effect is picked depending on the nature and severity of the discontent. It's easier to present and correct.
It's roughly what I did in Demo8 with the limited social model (Policies and Discrimination only).
I'd show message: People are unhappy because of discrimination - severity 40%.
The effect would be:
Nothing happens (60% since severity is 40%).
Riots (33%) Guerrilla forces formation (7%).
The ai and player can easily understand that, in order to counter this, they need to adjust discrimination.

For Revolutionary Feeling, I'd have not one but several events:
Monarchist regime discontent, Republican regime discontent, etc. each with a smaller %. Currently, the model makes it likely that there be a revolution if there are 3 different regimes that have a small share as they'll ally themselves together in order to overthrow the government. I doubt that communists and monarchists would ally themselves in order for the weaker group to help the stronger overthrow a government.

Note that I said cause and aggravator. Simply put, cause is just a cause (discrimination, bad policies, regime different from desired regime), and aggravator may be the unfulfilled economic expectations.
That way players and ai understand that they can try to correct the possible problems either by having a better economy or by correcting the cause of the problem which is celarly exposed to them.
This may lead to more lines in the riot model, but I'd rather have two lines for 2 different causes than 2 for different effects coming from the same cause.

[yellowdaddy]

cause is just a cause (discrimination, bad policies, regime different from desired regime), and aggravator may be the unfulfilled economic expectations.

sounds about right (economics usually aggravates people more than anything else, even religious extremism) - so there's not too many PAFs then?

So you plan to have more causes than effects in total, and more causes per single effect, then?

This means I have to show the player the possible regimes knowing that the player may define his own values, and show each possible value and the support for each share.

Why should the player get to choose? Shouldn't they be swept along by the results of their earlier decisions/actions?

[Mark_Everson]

Originally posted by LDiCesare
I'd rather have a simpler model like:
Discontent 1 = (cause 1) * (aggravator).
Discontent 1 leads directly to a bad effect, bad effect is picked depending on the nature and severity of the discontent. It's easier to present and correct.

Hi Laurent:

I'm comfortable with your proposed approach, since it seems to make a lot of sense. The big decision for the player then would be in balancing the potential negative effects in economic, technology, or other spheres of the changes made to keep the people happy. For example the people in general might want a monarchy, but that could be slower at tech growth and larger in corruption than a republic regime.

My only caution would be to leave the old formulas commented out in the code so that if we decide your new approach doesn't work for some reason, we can fall back to the old ones easily. Another option (I don't know how hard this would be to code) is to leave the old "complex" model as a player-selectable option for those who actively don't want things to be straightforward and linear. Since you already have the old model pretty much coded, putting in the option to have either might not be too much extra work.

[LDiCesare]

Another option (I don't know how hard this would be to code) is to leave the old "complex" model as a player-selectable option for those who actively don't want things to be straightforward and linear.

Although that is an option, I wouldn't be able to provide:
- a user interface to explain the player what to do
- tell the ai what to do.
I'll certainly do it with a command line argument.

[LDiCesare]

I've created another thread dedicated to the riots part, since this thread is becoming quite big and it deals with the social model in general. Let's keep this one for discussing other parts of the social model (social groups, how to represent them to the player, religion, nationalism, ethnic groups, government model).

[yellowdaddy]

On completely diff'rent tack...

How do political systems and religions work in Clash?

I like the idea of having a sort of "division of power" slider or pie chart, and this interface of generalised policies - do you plan to have it so depending on the settings you pick on the interface, you get assigned an "ism"?

In the "cruise mode" do you return to the traditional select "-ism" as your system of government? If so, how many "isms" do you want?


On the subject of the negotiable Gov.Pol.s -

Slavery and Civil Liberties seem the same to me, though I know you want to have slaves, do you need a seperate policy for that?

Some might say Religious and Ethnic discrim. should be conflated sometimes, though I personally agree with having them seperate.

I think there are some policies missing -

What about Economic policy of the government beyond Property Rights? Primitive Civs have to have a religious system, and become more advanced as economic considerations predominate.

You can have a range of options: religious collectivism (Communism, Fascism, Islamism...) and planned economies, mixed economies as well as market economies and participatory economies and subsistence economies.

Can you have Weberian categories of political power and oligarchies?
charismatic domination (familial and religious)
traditional domination (patriarchs, patrimonalism, feudalism)
legal domination (modern law and state, bureaucracy)