dh_epic's AI Solution

[Aussie_Lurker]

Hmmm, a LOT of interesting thoughts here. I agree with Sirrian about needing a middle ground between Civ1/2 system and Civ3 system! Additionally, though, if we have a 'multiple victory' win system as I have suggested elsewhere, then part of our problem is solved, as it would be nigh impossible for a single civ to achieve victory in all areas.
In addition, though, changing the WAY in which the victories are achieved, and adding new victory types, would also allow all civs (player and AI) to remain competitive right up until the 'end-game'.
Having 'minor civs' might also help the situation, and fits nicely into DH_Epics model for type B civs! As I have mentioned in previous threads, if you gave sufficient incentives for peaceful relations with minor civs, then their lack of competition would not be a great hindernace.
Last of all, perhaps if you tied the way in which a civ tries to achieve victory to its civ traits and, to a lesser extent, its current government type. For instance, militaristic, expanionist and seafaring civs will mostly focus on a domination victory, and will probably do so through a combination of squashing less powerful nations, and attacking their closest rivals before they can threaten them; these civs will also probably eschew military alliances and/or MPP's, except for very short term gain.
Agricultural, Scientific and commercial civs, OTOH, will probably seek more peaceful routes to victory-they will probably strive for an economic and/or diplomatic victory, and will probably seek victory through careful alliances with smaller powers against a very clearly percieved threat, and will try and give a leg up to less powerful civs in return for 'future considerations' trade and tech deals will also feature strongly in these civs reportoire.
Industrial and Religious civs will adopt playing styles somewhere in the middle, with their overall style probably being dictated by their other civ trait. They will be more agressive than their ag and sci counterparts, but will usually go to war if they feel 'boxed in', or in pursuit of a specific goal (like a luxury and/or resource in the hands of another player). They will form alliances and trade pacts, but usually only with those who 'share their values'-all others are merely 'grist to the mill'! They will mostly try to pursue cultural, economic and religious victories, as appropriate!
Lastly, if the 'evolving' civ traits model is adopted, then this could create intriguing situations for the human player. For instance, he sees an AI civ that began as agricultural/scientific, so thinks its a bit of a pushover-what he DOESN'T know, however, is that the civs start point dictated a more ruthless Militaristic approach, changing its civ trait accordingly-and that is how it now plays. The *hunter* might end up becoming the *hunted* in this scenario!
Anyway, this is a great thread, and a subject worth delving even further into!

Yours,
Aussie_Lurker.

 

[dh_epic]

It looks like there's a lot of support for the problem -- which I think is MORE important than support for my solution.

I think most people would agree with this statement: The AI needs to offer a better challenge to the player, but they also need to play realistically and cooperatively as much as it makes sense to do so.

The difference is that some people support my solution, whereas other people are (correctly) pointing out the real problem: gameplay. There IS no reward for playing cooperatively and realistically -- much larger reward for breaking treaties, isolating yourself, and changing your mind with the suddenness of a megalomaniacal psychopath. There is a big reward if you copy Hitler, Napoleon, or Julius Caesar (and even Hitler had a few close allies). This is why the AI will generally play like a Caesar (and so should the player) -- because it's the only way to keep up.

Plastiqe is onto something when he talks about making reputation a much more important factor. In Civ 3, I pretty much disregard reputation, and win (almost) every time. As such, the AI needs to be equally as nasty.

Aussie is also onto something when he talks about pulling victory conditions apart. Domination victory is obviously rewarded by playing like a Napoleon. Cultural victory is supposed to be an alternative to that, but it isn't: Napoleon is actually more likely to have a great culture, too, since conquesting grows his economy, production and population. Pulling current victory conditions apart and/or introducing new victory conditions would be a huge step in the right direction -- it would mean that Napoleon couldn't conquer the world AND be in good position to win all other victory conditions. It would mean that the players AND AI would have to make choices as they go along about which goal to pursue, depending on which they could reach faster.

 

[sir_schwick]

Problem: AI Perception of Other Civilizations

• Only one quantity exists for AI to judge opponents against.

Currently all Civ A has to judge Civ B's future actions on is reputation. Reputation mostly is based on how much Civ A likes Civ B. Unfortunately 'reputation' combines threat probablity, integirty, and a few other different fields into one, creating a confusing quantity.

Solution: Create New Set of Parameters Civs Judge Each Other By

• Concept

Being fond of someone and trusting them are two completely different parameters. Civ A may agree with the ideology and culture of Civ B, but doubts that the massive troop deployment on his border is benign. In CIV various leaders would judge opponents through a variety of filters. The interaction of these filters would help predict probable enemy actions and guide their own actions.

• Parameters

Direct Military Threat - This is a gauge of how dangerous your opponent would be in a direct military conflict, not proxy. Part of this determination woud also include the evaluation of likely allies. Contributing factors include; the borders and terrain that must be crossed, military technology levels, military deployment, military size, current engagements, and potential military production.

Indirect Military Threat - This is how much your opponent could threaten non-domestic targets militariliy. This includes important trading partners, allies, or other interests abroad. Many of the same consderations as direct military threat are factored, of course adjusting for your and the opponents logistical concerns.

Integrity - This is how much your opponent honors reasonable agreements. Opponents who regularily break ROP treaties and trade treaties without the proper notice will be percieved as devious. Turning on your allies in a losing war does not hurt your integirty, but will be noted, as discussed below.

Expansionism - This is a measure of how much expansion is likely of a player. Players who invade often or frequently are expansionistic. Those who turn on allies once they win a war are expansionistic. Those who eventually try to extort or otherwise expand territory are expansionistic. A very high degree of this trait might make an empire powerful but unpopular, even to the point of being allied against.

-------------------------------------------------------------

More explanatino is coming, but I have some work to do now.

 

[Khan Quest]

Dh_epic, and everyone else these ideas are great, this is a great thread.

I’d like to see type B (& C, D) played another way. These other civs would try to win by the other means; Cultural, Economic, etc. (wasn’t it you, dh, that suggested additional methods of winning?). To win economically, for example, you HAVE to play co-operatively to win.

Style could evolve and adapt too. An AI could start out as type A, and never get a chance to attack, thus have a tarnished reputation, but realizes that it has a lot of excess resources, luxuries (and commodities) to trade, and changes it’s play style to match.

The mix of tye A, B, etc. opponents should be more or less random. Or maybe by degree: Opponent 1 this game, is 80% type A, 20% type B, Civ 2 is 33% A, 33% B and 34% C, civ 3 is 100% type C (but may adapt later).

One idea about reputation. Romans captured, enslaved and conquered many Germanic tribes. Later they became allies in WWII and with Democratic systems are allies today. I’d hate to see the Rep thing overdone. This has been said before, but similarities of governing philosophy should also significantly affect relations.

Military victory should remain a viable way to win in Civ IV. There shouldn’t be so many diplomatic changes made that physically conquering the landmasses can’t be done.

A player (or AI) should not be able to always win with the same formulaic approach. Circumstances of starting locations, terrain, resources and discovery of opponent’s “style” should force you to adapt your play style to achieve victory. When playing against several type As, you’ll probably have to play as an A to keep up. Playing as an A amongst mostly type B almost ensures a defeat.

 

[dh_epic]

Looks like the solution alternatives are coming in. By all means, let's keep the discussion rolling.

The Basics: Seperate the AI into two categories, with one category truly playing for victory, with the others as pawns to be manipulated towards certain ends. This is what got this whole thread started.

Random Crisis: The Type B civilizations could undergo important historical events, that stimulate the rest of the world to respond. For example, a Type B Civ is engaging in genocide within its own borders.

New Type A Civs: Certain Type B Civs would become Type A Civs if the other Type A Civs are getting slaughtered.

New Type B Civs: Barbarians can become type B civs by conquering a city or by surviving for a long enough time.

Type B Civs Everywhere: Barbarians ARE type B Civs, or a special kind of AI Civ that has only one real city, with no culture, and are easy to negotiate with / buy off / assimilate.

Overhaul Gameplay: Force a combination of Type A and Type B gameplay styles by rewarding aspects like cooperation, and adding new victory types that cannot be achieved with the overall expansion strategy. Suddenly a Type A civ needs to be a little historically accurate -- keeping a lot of allies, and avoiding atrocities. No need to pull the AI apart, there are now multiple equal strategies.

Many Types of AI: Instead of seperating AIs into those who compete and those who don't, seperate it into those who pursue one goal, and those who pursue another... assuming that there are multiple victory types that are independent (e.g.: expansion doesn't improve your chances to be a cultural powerhouse).

That sums up a lot of the suggestions so far, no?

 

[Aussie_Lurker]

OK DH, I like the general thrust of your most recent post but, as I have put forward in another thread, I would like to add the following 'wrinkle':

Barbarians and goody huts would now be a 'catch-all' Type B, or Minor, civ. They may have no cities, a few cities or quite a few cities, depending on how 'sedentary' they are. Also, whether they are considered 'barbarians' or goody huts will depend on their characteristics and the general 'agression' level you have set. So, for instance, a type B civ with religious/commercial characteristics is liable to settle down early, build a few cities (at most) and try and estbalish peaceful relations with their neighbouring type A, or major, civs (and even other type B civs if possible!). This would be achieved by offering certain bonuses which only type B civs would possess (an idea I got from Minor Races in 'Birth of the Federation', where you are actively encouraged to form peaceful alliances with minor powers, due to the benefits they grant you!)
A militaristic type B civ, OTOH, would have lots of military units, would build very few cities, and would probably optimise said cities to churning out military units to go out and attack neighbouring type A and type B civs. The thing is that even these militaristic civs would offer benefits to a player who could successfully 'tame' them ! Anyway, hope that makes sense.
Another point I wanted to make was that I DO agree with the idea of 'semi-random' events, both positive and negative, which could keep people in the race! What makes this different from simple 'catch up', though, is that semi-random events are in large part based on player/AI actions. So though an economic crash might happen regardless, the chance of it happening would increase significantly if the player was being very 'economically irresponsible'
Again, hope that makes sense!

Yours,
Aussie_Lurker.

 

[plastiqe]

I think that Civ IV should be released in the same pattern as Civ III, the basic game will give you single player and there will be an expansion for multiplayer (and it should work out of the box!!!). I think Siran is right about a different AI for multiplayer, it needs to be way more competitive (probably to the extent of cheating) to keep up with many human opponents. Siran, what aspects of the Covert Action AI would work well in Civilizations?

dh, I strongly disagree with your basics. All civs need to be playing for victory, and they can't be programmed to go for a certain victory right from the start. Imagine starting on an island next a "B1" civ and knowing that they will hold all their treaties and be your "pawn". And I'm sure that players would be able to figure out which kind of civ each one is pretty quickly. I also disagree with Aussie in that civ traits should determine the type of victory a civ pursues. Both of these ideas make opponents too predictable.

AI (artificial intelligence) means trying to play like a human, correct? An AI civ could have a plan of which victory they are going for, based on their early game conditions, but they need to be adaptable. If the AI has a weak civ next to them, they should attack and take them out. If the AI is alone on an island, they should build up their cities and have a strong culture. So the "A1"s and "B1"s would emerge in the game, but would not be programmed that way from the very start.

sir_schwick I like your idea of separating reputation into more factors (you mis-labeled expansionism, it should be aggression). There is a feature in Civ III that could be used for the AI in Civ IV to cover your ideas: the advisors. The human players Greek advisor says (unhappy) "Rome has a bigger military than us, we need to build up our military". If the AI was hearing the same thing from their "advisors" then those factors could influence their behavior. The AI Roman military advisor is saying "attack Greece we are stronger than them!" But the Roman foreign advisor says "Greece is our ally, don't attack them". And the Roman trade advisor says "Greece has incense that they won't trade to us, attack them!" The AI would weigh these factors and choose their actions, in this case to attack.

Now obviously there wouldn't be any actual advisors, it would all be programmed in the AI, but you get the idea. I still think the history of hostility and honored treaties (currently reputation) between two civs should be the most important factor.

Now as to the separating of cultural and military victories, this is very necessary. Maybe captured cities don't grow culture till you've had them longer than another civ, so you would have to rely on your first few cities (homeland) to produce most of your culture.

 

[Aussie_Lurker]

OK, once again you have misunderstood me Plastique! First up I said it should act as a GUIDE to how the AI plays and what victory it pursues! However, a civs starting situation and how things progress through the ages may impact on its final decision! For instance, a militaristic civ trapped in the heart of a mountainous land mass is probably going to eschew a domination victory, and might instead pursue an economic and/or scientific victory! This, of course, will cause its traits to evolve accordingly, which will effect its playing style even more! Later in the game, however, they might encounter lots of rival civs, win a few wars, and decide that a domination victory is best! Again, this might cause yet ANOTHER evolution in its civ trait(s)!! One thing I agree with, though, is that the AI MUST be adaptable to its situation, and be prepared to change its behaviour according to need! Traits SHOULD act as a rough guide, however!
Also, unless I am mistaken, what DH is referring to is 'minor civs' vs major civs! Your class A civs will be your Germany's, Englands, Zulu's etc, wheras your type b civs will be your Hungarians, Poles and other 'good hut/barbarian' civs we know from prior civ games! In 'Birth of the Federation', there was a lot to be gained from forming alliances and membership agreements with your minor or type b civs-and I RARELY tried to crush them with military force-unless I was playing the Cardassians! But I digress ! The point is that you would have MAJOR powers similar to what you have now, but you would fill out the map with a few dozen minor powers whom you can either try and crush or try to woo over to your cause! Type B civs, due to their slow growth and poor development would not try to win-merely try and survive. If they are very aggressive, then they will do this via attacking everyone they can get away with, if they are passive, then they will do it via co-operation and diplomacy-offering their 'gifts' in return for their survival!

Yours,
Aussie_Lurker.

 

[plastiqe]

Must be that Aussie/Canadian language translation, eh.

I think your putting too much emphasis on traits determining a civs strategy. Starting location would by far be the most important factor. If their starting location has/doesn't have a resource. If they are near other civs or isolated. If they have enough room. These are all more important than being commercial or religous, although those traits would make some strategies easier.

I've been disagreeing with there being "A1"s and "B1"s in the five/six/seven starting civs. Barbarians could be programmed differently.. they are barbarians and not civs. If you wanted "minor" civs, why create a whole 'nother AI for them, just have them start later than the starting civs. So if there was an unsettled area at 1000 bc you could have another civ spring up. They'd be way behind in techs and easy to conquer, but thats not to say they couldn't catch up and become powerful. Instead of just minor civs being programmed to co-operate and offer gifts, any civ should do that when they are in a very weak position.

 

[Aussie_Lurker]

OK, to answer your first point-ummm, thats EXACTLY what I said-traits are not the be all and end all, and starting position, resource access, neighbouring civs etc should play a bigger role, but traits should play a role as well-and traits should EVOLVE through the game!

Also, I don't WANT minor civs to appear just late in the game! I want them to be the natural successors to 'goody huts' and barbarians, but as a long term rather than one off effect! In order to achieve this, the AI behind them would have to be less 'competitive'!

Yours,
Aussie_Lurker.

 

[sir_schwick]

I'll go back to my map filling idea. There would be no more static goody huts, instead they would be players. Each game would be populated by many, many players(100+ would be good for normal). Each player would have 'stages of nationalism'(does not described my concept well), some would stay at one stage for the whole game, others not so much. Here are the stages:

Nomadic - These players will wander around and create camps for maybe a couple years at a time(a turn possibly). They presumably follow the herds and occasionally raid. Militaristic nomads will raid a lot more often, while religious nomads will pick up more nationalized religions. Commercial nomads will create tradeable products. Nomads will settle into villages at different points, maybe even early in the game.

Villages - These players form villages(2-4), which act like small, one radius cities of up to pop four. Their technology is usually comparable to other players around them(including civs). They can usually defend themselves and have a strong sense of defence. Some villages even have good road systems. Scientific Villages would tend to be equal or even greater than the region leader in tech. Industrious villages would produce many monuments and other memorial structures, as well as industrial buildngs. Villages might consolidate into a city-state.

City-State - Ruled by whatever means, a city-state controls the territory around it. These players would not expand beyond their initial city. Any settlers they produced(from overpopulation) would form new city-states or colonies. These states do not expand, but are very territorial. Eventually city states will want to expand their territory to a more secure position of Passive Nation.

Passive Nation - These players want to ensure prosperity without being overly expansionistic. They tend to control 8-12 cities and any villages in between these cities, all domestic. They will join in alliances with expansive powers in major wars and keep terriotry gained that way. They will also start wars with similair sized nations over critical resources or territorial disputes. They will only seek to influence their corner of the world, not the entire world. Eventually a successful passive power will want to become an expansive one.

Expansive Powers - These players would seek to increase their influence or direct control over as much of the globe as possible. They will often use espionage, diplomacy, or massiv emilitary force to accomplish their goals. They are the human civ player, minus the exploits. These are the players vieing for the top. An expansive power who fractures or losses badly might become a passive one.

All these players do some research and keep up with the technology around them. They also might eventually evolve or devolve in states. The mix will require you use diplomacy, especially early on.

 

[dh_epic]

Fair enough, some people don't like the whole seperation of AI types and think that the game should be fundamentally altered so the AI doesn't have to choose. But if the game is not fundamentally changed, what it means to play like a human is what it means to be a complete bastard with no regard for history or reputation or anything -- to be vicious and nasty and backstabbing. That's the only way to stay competitive with the best players.

Again, Aussie pretty much nailed it with minor Civs. We're talking about Cuba, Syria, Somalia, El Salvador, Malaysia, Croatia... guys who have no chance at victory but can still change the balance of power. Guys who would be difficult to pick in a world-wide scenario, but who you'd still want to be there. (Playing through an actual game, of course, America could end up being the minor civ that affects the balance of power, and the Poles could be the superpower. The point is the way the world's power is distributed.)

I still think you could sufficiently randomize and control for starting location such that it's no more a factor for your victory than it is now.

But you're right, pulling apart the victory types and rewarding more peaceful, diplomatic, or loyal/consistent behaviors would offer an equal incentive to the traditional route (warlike, isolationist, psychotic).

I think the key for culture, for example, would involve making it more dependent on cooperation. How could you have a culture that is admired around the world if everyone hates you? I think culture should be something you actually cram into peoples' borders with units and luxuries. That way a city no longer just has X points of the home culture, but an additional Y points of alternate cultures. Not only would the Y accumulate from more trade and sending more missionaries/artists/philosophers to their city, but it would also start to subtract when you declare war on that Civ. If they hate you, they start to redefine themselves in a way that is very different from you, spiteful towards you.

I've wanted to do a post about culture for a while. Maybe this weekend or the next.

 

[Aussie_Lurker]

Hey, thanks for that DH_Epic. I also agree that who is a major civ and who is a minor civ should change from game to game. The way I imagine it would work would be: in the setup screen, you select the number of major civs you want in the game (much as you do in civ3), and you can even select who you want it to be (or choose random, as in civ3). Next you click on how many minor civs you want-'None', 'almost none', 'quite a few', 'many' and 'lots'. This will determine roughly how many minor civs will be in the game. They won't play to win, unlike the majors, but they will defend themselves and, if agressive enough, will attack other civs as they see fit. They can grant unique benefits, however, to whomever brings them into their nation (most likely peacefully). These benefits could be a unique unit, wonder and/or improvement, a boost to research in a particular field, an overall research boost, production, food or wealth bonuses, culture bonuses and the like! Also, each minor civ that you deal with in a peaceful fashion might help you towards a diplomatic victory!
Lastly, I do agree that culture victory should work more like the way you said. Though culture points from improvements and wonders should still have an impact on borders, culture victory should depend more on how many of your products and ideas you can export to other nations. Perhaps every time you trade a tech and/or resource, your culture victory points increase!
Anyway, just a thought!

Yours,
Aussie_Lurker.

 

[plastiqe]

I've wanted to do a post about culture for a while. Maybe this weekend or the next.

Heh, me too, but I'll post my idea here since we are talking about it.

Culture Corruption would work the same way as commercial corruption. The further away from your capital city , the more culture a city would lose to corruption. It is pretty simple, and it would reflect that your civs culture comes from your homeland and not from your outlying provinces. The other thing I was thinking is that captured citys should have a delay on when they start generating culture for you, maybe in years or maybe till you have assimilated the population. I don't think it matters if everyone liked you or not, that would be combining cultural and diplomactic victories. Look at the U.S., they have a huge culture but aren't the most popular country by any means.

 

[dh_epic]

Not a bad idea in principle, but somehow I'm worried it could be nearly as annoying as actual corruption. I think the problem with corruption isn't that it exists, but that it's inevitable. Of course you should keep expanding. Of course it's good to have more cities than fewer cities, even if those last few are really corrupt. Thus, you end up having a whole pile of cities with corruption, and you have no real choice. A real choice would either be a way of expanding without corruption, or a way of dominating without expanding. But that's a whole other discussion.

I'm kind of into the idea of dragging minor (type B) civs into your commonwealth / union, and it offering unique / random benefits. Not unlike goodie huts. I think minor civs is way more interesting than goodie huts.

 

[sir_schwick]

What about htis on Culture Corruption. The Culture is not lost, but becomes either 1) culture of another Civ, 2) a local culture. The second could contribute to eventual succession, since they have a strong sense fo regionalism. this woudl be especially true of overseas provinces.

 

[dh_epic]

Coming back to this thread, I definitely tried to offer a deeper meditation on culture.

http://forums.civfanatics.com/showthread.php?t=102134

If you have these Type B Civs or Minor Civs that don't necessarily fight for victory but stick around for realism... adding a culture-spreading model becomes very interesting. You could have a minor Civ that is attacking you because they're just barbaric. But you send a few missionaries their way and infiltrate their culture, and suddenly that Type B Civ feels a certain amount of similarity with you. They even ally with you.

You beat the Type B Civ without even lifting a weapon!

 

[Moonsinger]

Easily said than done! Unless one is in the designer/programmer shoes, it's hard to imagine the problem that one may have to deal with. Is it possible to build a good AI (type A, type B, type C or whatever)? The answer is not "yes" or "no" but "time". On top of that, there are two major types of time: development time and runtime. Development time: If it takes 10 years to develop Civ4, chances are Firaxis won't go for it; on one will. Runtime: If the it takes forever for the AI to make a move, chances are no one will play it.

 

[dh_epic]

Well, the truth is, I kind of threw out this solution not only to show what the AI should strive for... but why it's easier than a lot of suggestions people have put forth for programming "deep blue", that holds loyalties and is still competitive, and overhauling the gameplay to make the AI less exploitable.

Not to say my model will be easy to implement, but easier than the "perfect solutions" people are recommending. They can basically recycle old AI's with some tweaks for the Type A AI. And the Type B AI just has to be there. Again, not to say these will be easy to code, but certainly easier than trying to build an AI that is both realistic and competitive.

I also put in a "shortcut" where the Type A AI catches up using a relatively "dumb" algorithm. It artificially speeds up to stay close to where the player is. Much easier than trying to close game exploits and helping the Type A to play consistently and intelligently. Not everyone likes this one though.

Still, I think this is actually an elegant solution to a problem that could be much more complex.

 

[Jon Shafer]

So you think you know AI programming better than Soren, masters of comp sci from Stanford and Oxford?