EWOM has various Victory point, it is imperative that the AI knows how to achieve that Strategically.   I will even say every move of the AI should try achieving this strategic goal.  

Basically you said to achieve Victory, the AI should define various subgoals and try achieving them.  I agree.    It will be nice if the AI can pick one of the goals available and take multi-turn actions to achieve it, instead of shifting goals every turn.  Say when AI recently scouted an independent iron mine, and it also knows its rival is on the east side, the AI will pick the best cause of action to achieve its Victory, e.g. either try capture the mine or kill its rival & work for it for a few turns untill more important goal appears.  

Instead of metagoals, subgoals, I call them quests instead.   AI will be able to generate a relatively complete list of quests upon seeing new enemy units/new cities/resources/etc.  Then AI do a cost-benefit analysis to see which course of action is most profitable, which one is the right step to its victory.    The AI will follow thorough in many turns to complete the specific quest.  These complete lists of quests should also be generated at other levels like diplomatic, technological/research, and economic level. 

As for mergers and surrenders, the AI should gang up on a dominant player- balance of power should be a factor in AI relations.  The AI should gang up on an overly powerful AI as well though.  This shouldn't overrule alignment though.

I agree. If you are winning, those AIs that like you should suck up to you. However, every one of us here (probably) would think nothing of abusing that devotion and leaving the AI behind with the rubble, so some sort of other system may also need to be added.

Just as long as it actually works. Although I suppose in a foreign policy situation, I could justify it going either way: nationalism does funny things to people.

I understand what you are talking about with the AI "stream-of-consciousness". However, this "marger" thing had better be a last-ditch effort when someone is thruly getting clobbered: otherise, you could just "absorb" AIs that still have some fight in them.

<!-- @page { margin: 0.79in } P { margin-bottom: 0.08in } -->

Goodmorning all,

I've been thinking about this proposed system. and decided it might help make things clear if i gave a direct example.
We don't know enough of elemental to make meaningful concrete recommendations, so instead I'll focus on Civ 4.

In Civ 4 good players almost can not use the AI; for example Automated workers are generally accepted as not being programmed to build enough cottages. But the problem is deeper then that, The automated worker AI doesn't truly understand specialization. Land is upgraded based on the land itself, not based upon what the town it is nearby is being used for.

A good Civ 4 player will have one or two towns dedicated to research, one or two dedicated to building big projects and wonders, 3 -4 cash towns, making enough money to pay for everything else. . . . Towns have functions and excel at those function depending on the other towns to fill in the gaps, so that overall the civ is balanced. For example a Cash town wants to have 40 food production and every other tile in the 20 it can work will have cottages/townships, banks, and as little else as possible. A research town wants as much food as it can make so that it can have lots and lots of population/specialists boosting research, and just enough construction to build the few research wonders and research buildings the game has to offer.

A good AI is going to be able to exploit the power of specialized cities. Workers assigned to improve a town would improve it according to it's needs to become the right type of specialized town. The AI would have to look at the land around a new town, and the needs of it's civ, to decide what type of town to make. It would have thresholds and decision trees [if town has natural food production >35 food, eligible to be a science town or cash town, natural productivity > 20 hammers, possible for production town . . . ]. All of this pretty much has to exist. So what am i proposing??

When a user turns on AI governance of a town (lets say a cash town). the AI would look at the land around the town, the buildings build in the town, and form a 10 - 20 turn plan of how to make that town a Better cash town, more profitable[not to the stupid level of destroy walls and barracks because they cost money/turn, sane moves necessary moves]. The user could then just leave the town safe in the knowledge that the town isn't going to start building buildings that give bonuses to research and super barracks, or produce any super expensive units. The user could also ask to see that 10-20 year plan, and it would have things like 'banks give bonuses to cash flow, -> build construction hall, to unlock building bank [This example is from MOM not Civ 4, but going for obvious examples here], -> build bank. Cash cities need cottages -> Build worker (call free worker)-> assign to upgrade land around town to make cottages'. The AI would be presenting an ordered list of goals, and actions to fill those goals, and at the side a ordered list of which actions it's going to take.
The user then could leave happy, Or change the priority. The user could instruct the AI to build the worker first then the bank, or to not build the bank at all, because s/he knows that there is a magical bank spell which they plan to research and cast in the next 10 turns.

Now how did the town become assigned to be a cash town? Obviously the user could have picked from a list of town types(cash, military, production balanced science, magic, border, undefined) each would have a smaller list of how to achieve that type of town, a (swarm military town, or an elite unit military town, a magic essence production town, a magic research production town, a mana production, a balanced magic town?) if the user wanted to further refine the town's definition. But what if a town is already established and turn 100 you finally get around to turning on the AI because your just going to go to war, and want it to handle some of the nitty gritty. Well the AI should be able to look at a town, it's configuration, location, and guess what type of town it is. (have a Huge library of +3 to big and expensive? probably a magic town(research), have three low level barracks, and armories, probably a military(swarm)). The AI would probably be able to guess which of the 7 archetypal towns it is, (or give up and say undefined), and probably have a better then random chance of guessing the correct subtype[the user may not give a rats ass about the sub type, so all subtype would have a 'general' setting as well].

It wouldn't matter if the AI wasn't perfect, because it would be better then a blanket 'treat all AI controlled towns like this'. That AI's guesses wouldn't have to be prefect because when you turn on the AI it would tell you what it's guessing, and you could correct it if it's wrong.


The same logic can be applied all the way up the decision tree. . . Have a lot of research towns, probably going for a tech or diplomatic win... have a lot of military towns, probably planning to start a lot of wars.

The point is the AI needs to have most of the tools to make the top down decisions, Just add a bit of UI, and you gain a LOT, for not very much.

Hopefully that is much more direct and clear then my previous attempts, and those who understand programming can hopefully confirm that although it is ambitious, the suggestion is not over the top crazy.


Best wishes all
Robbie Price

Both sufficent bata time, but also if the system is inplace, and access is also given to the decision tree, elite moders could develop new script tools, and define new branch points, finner distictions between towns. This would alow moders to build and save AI scripts, that allow for fully automated competitions. 3 users load their AI to govern the towns, and are shown the whole map, as they watch their creations battle it out against eachother and the official AI.

Just as the game has a map editor, it can come with an AI editor. to incurrage this sort of thing.  I would LOVE to see Frogboy's newest creation each week entering the ring agianst everybody else.

 But that's beside the point,  the main point is that the AI needs a UI to be usefull. and that the resulting UI need not be imposibly complex to be highly functional and a real game seller.

a 4E game where you can actually turn on the AI managment of something and not have to hand hold it all the way(just maybe point it in the right dirrection when it guesses wrong).


Take care all
  Robbie Price

I believe an alliance between the AI's is the better solution and requires less work for the developers.

A merger between two or more nations comes with problems which include:

1) This merger means one nation will have two channellers... I highly doubt this would be supported in the game.  I also doubt Stardock would include extra programming to decide which of the two channellers gets downgraded to a hero status and then which city remains as the capital and then what happens between existing differences in other opponent relations... etc., etc., etc., .

2) The merger means one nation might be trading ore for herbs which it needs while the other nation is designed for using ore to be effective.  The point being that each nation has different needs and specific AI programming.  You can't just mix separate AI scripts to handle a combination of two; three or four nations as a single entity.

3) A merger also places themed nations at risk... such as a nation which uses undead merging with a nation of saints/priests.  Sure the developers and modders could design specifically which nations are allowed to merge and which ones are not allowed to merge, but then once again this is extra working time which could be spent on game content or stability instead.

That's an alliance...

Actually i was hopenig that some of those 'problems' would infact be features.



For A, niether, for B, The capital of the stronger empire. (with the option of declairing a new capital town, in the merger discussions)

I say neither for A because the point is that an empire with two channlers is stronger then just one, and merging is only and strictly a end game, last ditch choice to stave off being killed/elliminated(either immideately, or inevitably).

The new empire would inhairit all properties from it's predisessers, wieghted by the strenghts of the empires involved. 

if we imagine a Civ 4 like AI memory (+1, for X, -2 for Y, +1 for changing Z. . . )   Then the resulting relations can just be all the previous elements summed.  Some thinking would be needed for A at war with B peace with C, and B merges with C to become BC.   A simple revaulation of state, aught be enough to arrive at a conclution of war or not.

That actually is the goal, by designing the AI from the ground up to be transperent, flexible, and to share as much code as possible between AI's Merging them becomes less difficult to code.  If the AI is build to interact with a user, it's only going to be easier to have it interact with a fellow AI.  Mergers beome a pheudo-free bonus feature of a transparent, user friendly, and powerful AI.  

If you can make an AI that can take and made suggestions, and can wieght those suggestions, you can make an AI that combines two (or more) AI's.  Taking and making suggestions is a nessisary requirement of a good AI UI. and Wieghing options is obviously a quality of any good AI, UI or not.  

Yes the merger code is more code, and therfore more work, But if you accept that the AI should have a UI, then all the tools you need to make that UI, do most of the merger for you.

I have actually already suggested a resolution for this


An undead team merging with a holier then thou team would result in little gain to the merged empires. making the merger much less likely to happen because it has no reason to do so.
In the case of very differnt empire strenghts, the new AI would favour stronght the choices of the dominating AI (by it's wieghting), resulting in a weak channeler with few benifits being added to the strong empire. 

If the Holier team got crushed under the undead team and was accepted for a merger, the new AI would behave almost idenentically to the old undead team AI, and gain little more then a somewhat strong hero.  The same situation between two undeads, or two holies would yield a very strong hero and possibly a few channler bonuses. (but the new AI would still behave like the stronger AI more then the weaker). 


Best wishes all.

Robbie Price.

Goodmorning again,

I think the actual simplicity of what i am proposing is being lost in translation. 

Firstly non of this is worthwhile unless you set out to make the AI User accessable/ineractable.  The amound of gameplay gain, would be tiny compaired to the code required to power it. But if you DO use a UI inhanced AI, then most of it is free, or nearly free( once you've made the rest of it work. [not impossible, but not easy, and not beyond frogboy]).

Regarding your statements,

Where to put the capital:  really this isn't a stopping block, AOW, Civ, all 4x games allow you to move the capital. if the Chanler(s) can move(and they can), the capital can move. Problem solved.  Really, not a problem.

Multiple Chanlers: Yes.  (Really there aren't huge issuse here, Channler is a unit, now you have two of them. Yes the channler has faction wide bonuses, but i've already explained how to deal with them, otherwise you just have two units.)

No, I think you completely missed the suggesion here. I'll try again.

Empires A and B merge.  Empire C is born.  

Empire C's turn begins.
The AI from A looks at all of C and creates its action plan, meta goals and goals
The AI from B looks at all of C, and does the same.
AI C recieves the plans of A and B, By compairing the Meta goals and goals of A and B, and their relative strenghts  AI C decides what the new empires meta goals are.
AI C sends the meta goals to AIs A and B
Both A and B set goals and subgoals (with thier personality preferances, whatever) to meet the given meta goals. (can nolonger change Metagoals, This is a property of the UI, so if the UI exists, this will already exisistm, free!)
A and B return thier list of Goals and Subgoals to C, C chooses Goals,
Rince, Lather, Repeat till you have a full action list,
Make it happen.

Very little new code is needed. No comparions of troops, no comparions of items spells, all that jazz.  each original AI independently does what it was doing anyway(  Make meta goals, make goals from meta goals, make subgoals from goals, make actions from subgoals...  ) You just add to the code is an export and recieve from each step, and a system to wiegh two list to make one, and the rest is Free.  Now just add game play affects, and your golden!

If you build it right, the rest comes simply.

And that is what i'm trying to propose an alternitive to, because from a gameplay and stratagy point of view surrenders are looses, and boring.



Hence why i'm not even remotely suggesting that anybody should try to merge the personalities.   Each personality remains independent no merging, no setting personality sliders half way between two entites, no mess.  The results, actions the new entity takes, however draws from both independently. In the end you SEE a new AI that feals and acts like some linear combination of the two previous AIs. Without actually MAKING a new AI out of the two old AI's which would be a completely insain amount of fiddly code to get right, and not at all what i am proposing.

As for the need to "List nation conflicts, water/fire...":  Again there is a far simpler solution.

"

"

The same system can be algorithmically applied to any other property,  If the AI's have no spells in common, No spell bonuses. If one AI had lots of diplomacy experiance, and the other non, no diplomacy bonus, if they traded with differnt empires little or no diplocmacy bonus, if they traded with the same empires for the same sorts of things, ++.  if one has all fallen troops and towns and the other all human troops and towns, no bonuses there. 

In short you don't need to manually code in combinatorial relationship factors, you simply derive compatibility from gameplay state, and weather or not the two AI's return the same list of meta goals, goals...

If the  two AI's look empire C and both say, My meta goal would be XYZ, and they say thier supporting goals would be JKL, Then both will agree with Empire C's actions, and bonuses insue. If they dissagree strongly, detriments are incurred.  You make a messurement of how differnt the final decision was from each AI's proposal and if the agreement is high, high bonuses, if the aggreement is low, low bonuses(or dissadvantages for one or both Chanellers).  Let the preballanced personality differences deirmine dynamically the bonuses and detriments aquired by the new nation. 

+ knowing Stardock and GC the empires will already have presupplied hatreds of eachother, which can be additionally used but would not be needed to have obvious and strong gameplay affects. 

NTJedi
Can you outline for me a concreat example  (perhaps using another game (Civ 4, AOW, MOM)) where the simple modle I am suggesting that can't be resolved by a comparing the state of the two empires, and a measure of agreement baced on how close AI C's action plan is to the action plans of the given AI?   Because i can see how when abstracted it still looks overwhelmingly complex.  but i am fairly confident that if you sit down and step be step create a mental experiement of two empires and merge them with the steps i am suggesting you will not encounter unresolveable problems.

Take care.
Robbie Price.

 

Goodmorning all.

Firstly,  Yes, your right, it is possible that the game could be coded in such a way as to have multiple channlers be inpossible. Most/all of the proprties i can think to include in a channler are extendable to a multi channler modle, But that doesn't mean that the code as written is changeable. However only frogboy knows the code, so this we'll have to let drop.

Secondly, some clarifications;

1.  The 'merger' is a suggested replacment for the option to surrender. Thus all the reasons to include a 'don't allow surrender' are valid concerns.  Your suggestion that if a surrender/ merger option is allowed it aught have a 'don't use this for my sanario' option because it does make certain types of mods impossible i 100% agree.

2. There is only one situation where a faction would offer a meger, and only two situations where a faction would accept a merger. 
          If a faction calculates is chance of winning, (numerically or qualitatively) and arrives at the conclution that no conventional tactics can concivably result in it's victory it would then be enabled to propose a merger. (mods or faction by faction thresholds of 'I've lost' are possible). The key is that no mergers are proposed untill at least one team has already lost (but still has units on the board).  
          A faction will only accept a merger proposal if :
                   The proposing faction is very small compaired to itself and the channler  personalities are simular. For all intensive purposes surrender.
                   The recieving faction is also destined to loose (perhaps or very very close to destiend to loose), and good agreement.
          If a faction can not find anybody willing to merger, they may choose to simply surrender,  and loose the normal way. (or fight on till somebody will accept, or they are eliminated forcefully ofcourse.)

3. Because the merger option replaces a surrender option. All the problems of how a lone AI deals with aquiring a significant number of towns which may or may not be apptly adapted  to how it has been playing.  If surrender is impossible to code, meger is more so. Thus I shall assume that surrender is possible and that the AI problems of a surrender situation are 'resolved'.  This assumption may be invalid, depending on the setup of Elemental (Heroes 3, 4 surrender was a silly option, the maps we're too small to mater. for example.)

4. Why someone would ever reject a merger.  For mergers to work, some mesure of aggreement would be needed. (I called it a coefficent_of_agreement( CoA )), given the mechanism i've proposed there would be a CoA for both factions.  The CoA would  be between 0 and 1, and the 'strenght' of a merged channler would be, for sake of simplicity, Their_original_strenght * Thier_CoA.  So when a merger is porposed CoA's would be calculated [to say pricely how would require more understanding of the game mechanics then we know], and both sides would see them. When choosing to or not to agree.
        If one AI by far outstrenghts the other, then the CoA would be almost 1 for the strong AI, regardless of the CoA of the other.  In this situation the merger behaves almost identically to a surrender. The smaller channler has little to no say in the politics/desissions, hence the larger nations CoA is close to 1. The Larger empire might still reject the merger if the weaker CoA is too small, the Small AI's strength is weaker, then reduced sigginficantly , the result is a unit weaker then the strong AI's hero's. . . rejected.   For example Team A may have CoA of .97, and Team B a CoA of .08.  If the big team accepts the merger their channler would be reduced by ~ 3% of it's strenght, and in exchange would only get 8% of a far weaker channeler. . . not worth it.  Better to refuse and demand a complete surrender, or just concur them completely.   If however the two AI's are very simular in how they want to run things, the smaller AI's CoA may be in the .6 - .8 range. Then the AI's would defiinately agree to the merger, 97% of one Channler and 75% of another is much better then 100% of one only.  [Some additional Merger penelties can be included for balance reasons, such as CoA's can not exceed 0.80 for example to make them somewhat undesirable.] 
       If however the two AI's are almost the same strenght, Both AI's will have a CoA which can be anywhere between 0 and 1. The question then becomes, which is better, an new empire with twice the size and strenght, but with two slightly crippled chanellers, or finding a different merger partner. For the merger to go forward both sides have to agree that the merger is benificial.

Now direct responces:

*Shrug* This is a technical problem, not really relivant to good idea or not, it's just a matter of proper coding, and if surrending is allowed then it will have to be solved in the surrending code.



Firstly the pairing is unlikely, If the two don't agree enough they would have found better merger mates or been crushed.  Secondly If you did manage to get a merger between two equal strenght empires with those different personalities then they would rightly suffer the penelties of the miss-match.  The technical question of how one empire of merchant runs a new empire with necromantic troops and a medium-large standing army/production [remember already have to be loosing to accept the merger, so really won't be a successful war monder {if it is a successful warmonger then the strenghts are not equal, and the stronger dominates therefore no problem}]. 

The new nation doesn't loose the old personalities, they are still there, it just chooses actions based on the suggestions of both parties, and thier strenghts. In the worst case sanario, two completely differnt channlers would somehow merge with equal strength, Both would have terrible CoA's. They will probably loose. This makes sense.

I really don't think we've managed to communicate there at all.

The individual AI's take into consideration all the varibles (just as they would if the enemy had surrendered to them, and no mergers exist. See assumption 3.) Both AI's will propose the BEST route to victory they can calcuate. So the situation your worried about isn't a merger problem it's a surrender problem (or just a very very stupid AI problem which is somewhat insulting to the programmer to imply such neglect).  The merger AI then only needs to compare the two best stratagies and by an itterative process find an executable merged stratagie.

It is worth pointing out that if a merged entity starts to win, both AI's will start to agree more and more.  Two turns before casting the spell of making both AI's are going to be agreeing that casting that spell is the best way to win. (unless one of them is dead set against winning by magic, at which point that channers CoA would be ~0, and nolonger relivent.)

The umbrella AI dosn't need to take into account all the details because the individual AI's do that step for it. Very much differnt from what you seam to think i'm going for.

The researching spells problem can be solved at least a donzen ways, study both once half research going to each, sum the total investment to a new spell picked the first new turn, always have studing two spells at once. record the invested RP, finish one, then pick up the second from where the it was left.... just a technical problem that.

Regarding quests. not actually a problem, If you request a merger because otherwise you'll loose, then you've given up on the quest and it's gone.  If you accept a merger, because your stronger then them by a lot, and happen to fufill your quest bonus.  Yes this is somewhat a freebe, but, only in time not in substance.  If they are surrendering, it's because they've lost. If they have already lost, you've already won, Therefor you are as predestined to finish the quest as they are to loose the game. So you get it a few turns early, bonus. Getting them to beg for a merger in the first place was just another way to complete the quest.  The only issue is when two equally strenghted teams merge, and the AI accepting the merger has a quest in the lands of the AI proposeing the merger. . . but if my quest is to capture your towns, we are probably not the best of friends, therefore would have a low CoA, therefore wouldn't merge in the first place, or are going to suffer for the merger quest success or not. 
 In short, asking somebody to accept you as a merger partner looses you the option to complete quests (if the merger goes through) . ..  the stronger party keeps it's quests. 

Both remain in existance, but only the Umbrella AI makes final choises. The desires can conflict, hence the CoA's, but the actions won't because the AI's only suggest things to best meet the Goals they have been told to meet.  AI A may want to trade for ore, and AI B may want to run an army in and steal it. Both would send the suggestions to the umbrella AI, and the Umbella AI would choose one of the two options.  If it chose to Steal then it would tell AI A and AI B that the Goal is to Steal the ore, and AI A would propose the best way to steal, as would B. A would nolonger suggest trading, Because AI's would have the ability to accept limitations and expectations from outside [because mergers are a free bonus for having an AI UI, the code for finding the best way to do arbitary things will exist.] The result is that AI A's CoA would drop, and AI B's CoA would rise, but under no circomstance would the AI's *act* counter-purposes, since only the Umbrella AI acts. 

 
Only if the necromantic AI was stronger in the first place, if the tables were turned the Necromantic troops would be gradually replaced with less anti-capatalistic units and the land undead soils would be revitalized to produce sale-able goods. . . sad ending for the necro.  In either case the one AI gets a smaller and smaller CoA, and eventually seases to be relivent. That's what you get if you choose to merge with a empire you have a poor compatablity with, so they wont  do it.  + with ther small CoA's to begin with they probably would not be strong enough to win anyway and would end up loosing do to picking a bad merger partner.  In short this isn't a problem, it's a feature.


You are of course right about AgeofWonders:SM, the surrender code was not good, and it was easier to turn it off then to work around it.  With an python, moddable, UI enchanced AI. . .  it is doable. It is doable and it would add a lot of playablity to the end game which traditionally is painfully painfully dull, long and tiresome.

Regarding storyline campaigns or guided storyline maps, obviously both merger and surrender would be turned off.

Goodmorning again,

For Surrender i took the GC modle, all towns and units of faction A are 'concured/given' to faction B. nothing in faction B's orriginal empire is changed by the surrender.  Demanding a surrender rather then accepting a merger should always remain an option. IMO

Ya the War/peace status with other nations would be the product of the two early nations, not nessiaralyl at war with everybody either was at war with, because if everybody is attacking one small nation, and they surrender/merge to a large nation, the large nations politics will dominate by far, (and assuming no other wars were going on) all the other wars would stop [but the bigger nation would loose a small amount of popularity, how much depending on the size of the smaller nation, and how much other nations hated the small nation.]



As i see it an AI needs to has a list of probably 10  - 14 meta goals to choose from (modding might add more). Many of which will not make any sence in a given circomstance,  Meta goals like 'explore and expand' for the early game, meta goals such as 'win by victory condition [X, Y, or Z], Routhlessly destroy Faction J by any means possible' only make sence in late game and midgame meta goals like 'control all of Resourse P' only make sence after you know which resourse are rare, but before it's too late to take advantage of the manopoly.  At any given time I could see an AI running with a list of 2 -3 Meta goals, one primary and one or two secondary.  To achive a given list of 2-3 Meta goals,  probably about 10 - 15 goals 'build good relations, spyon, build army of size__ and training ___, achive production of ___ / turn of ____(complex manufactured good)'. . .
The Umbrella AI does not contribute any bias, or preferance, or personality,  It just desides what the teams Meta goals will be by looks at the given lists of the two subordinate AI's. Looking first at the list of goals to find if there are any agreements / contradictions.  It then reduses the goal list to goals (all of the goals that both want, most of the goals both said they wanted but didn't contradict(wieghted choice), and then wieghted selection of goals from the conflicting list.) Given the new combined list of goals, and the list of Meta goals (and the code for guessing what a human wants to do, from a list of Goals[part of the UI interface] it generates a final list of Meta Goals and returns that meta goal list to the subordate AIs). 
No additional personality considerations are needed. The Umbrella AI doesn't need to know about the full compliment of Units/spells/.../etc, because it's only dealing with the meta goals.

The sub AI's then take the given Meta goals as immutable, and generate the best list of goals and subgoals to meet the meta goals. The proccess is repeated generating a list of goals. The sub Ai's recive the goals and make Subgoals and tasks. . . [note that the last step would be the easiest as there are only so many ways to convert a subgoal into a task, (improve the land around this town(type defined), asign worker to build (thing)].  

All the Umbrella AI provides is a mechanism to force agreement between the two AI's and to measure that agreement to detirmine CoA's, such that the two AI's do not experiance conflict over actions and work cross purposes. Otherwise issues of one AI doing one thing, and the other AI being disabled for it would be crippling. The Umbrella AI forces mutual goal makeing.



The AI's would both make plans for both channlers, and would know both channlers CoAs(strenghts), and act accordingly since both AI's make plans for the FULL empire(both Channler units as well).    The Umbrella AI would only imprison the warmongering AI, if the other AI was MUCH strongerat the time of merger. If a war monger, an a normal AI joined, a merged entity would exist which was more violent then normal, but slightly less then the war monger (wighted by orgininal strength).



The new empire would do what was in the best interest of the new empire, (with most of the personality of the merchant AI[because it was stronger]). The channler would not be pointlessly sacrificed because niether AI is going to see that as the best way to win the game, killing off strong units pointlessly.  But if the Necromantic channler would be used as a tool of the merchant AI, and probably the overall empire would change to represent the merchants way of doing things.  Since vampiric troops consume more profit(via happyness) to maintain, they are undesirable to a merchant AI, they may be dispanded or used in front lines[probably being stronger that's a good place for them anyway], whatever the best way forward is useing to full advantage the troops that the newly formed empire has. Such troops would be less likely to be replaced due to cost. But in all moves and all desicions the AI would do its best to win. The two AIs prevent stupidity, the umbrella AI prevents contradiciton of goals / methods.

 
Take care all

Robbie Price