How does the AI "value" the Great Library?

[CerberusIV]

I have just had a situation where the AI's were slow to literature and in building the Great Library. I didn't bother to try for it as I was too busy fighting a war. Also everyone went for researching education rather than the invention path.

The Persians eventually completed the GL but had all the known techs anyway. Three turns later I completed researching education. Xerxes bought education from me for chivalry, 37 gold per turn and 130 gold!!!! I then proceeded to sell education to all the other AI's.

I don't know whether Firaxis can address this but it seems the AI's totally fail to be able to use the GL and stick to their normal rules for buying techs. I suppose it is not a big deal but it is a little disappointing when the opposition can be persuaded to do something this silly.

[HAND]

I consider the most important age for building wonders to be the Middle/Medieval Age.
The only wonders i would build in the Ancient age would be Great Library or Pyramids(only if someone else gets GL first). The reasons for this is, its that gives you a tech boost just at the right point; most of the other civs have discovered each other at this point so techs should flow fairly easily into your hands.
Its now early middle ages you have considerable wonder building options before eduacation comes along ...and if your the building type like me, you'll probably be an Industrious Civ so have the advantage with building mines quickly around your cities plus the extra shield bonus means youu should be ahead with wonder production.

Also in the ancient age i'm REXing like hell, so not much time for wonder building!

I think the reason why AI goes the Education route is that more options for tech trading that way, plus more improvements/wonders. Musketmen aren't really worth the effort at this point..i'd rather stick with pikemen or Numidian Merc since I usually play as Carthage.

[dexters]

The AI builds the great library only if it has to. In my current game, a slipup on my part allowed the AI to switch to the Great Library, (from pyramids) after I beat them to building the Pyramids. This cost me a pre-build in another city.

In anycase,Great Library is less important to AI than to human players. The ones you have to watch out for tend to have parity with you or are in the lead tech wise. (depending on your difficult setting). GL's do nothing for them. And since they don't play like humans, they can't adjust their strategy in the early game to "slack" research with hopes to catch up tech wise after the GL.

So on that sense, GL is really more of a human wonder than it is an AI one. Not to say that AI's will never build it. I've been beaten to it by the AI's in warlord games I've played before. But then again, I probably wasn't playing very efficiently back when I started.

Maybe in Civ 4, they can have AI's that can plan ahead and do things like this. As it is, the AI is quite good and it whooped my ass in the middle ages on regent. I was going to stick it out , but in a fit or rage, I went and deleted my autosaves :/

Edit: On second thought, I'm pretty sure you could do it with Civ 3. And Soren might have already programmed something like it. On occasions, I get the strange feeling that the AI can and does plan ahead. Not in detail mind you, but generally. In the same game I deleted (where I was being whooped), I was essentialy betrayed by my neighbour the Aztecs who I had an ROP with. They didn't use the ROP to attack me, but rather attacked when my forces were at their weakest.

The strange thing was, it seems planned. Earlier in the game, the Aztecs had signed an ROP with Japan so Japan used that to cross their lands to attack me. It then agreed for me to sign an ROP with them so I can cross their territory to take out Japan. They were playing both sides!

Strategically it was a masterstroke. Whether it was coded behavior, emergant behavior or just pure chance, the war that the Aztecs let happen through their ROP with Japan and myself left me in a precarious situation with my Empires split in two (my main empire on one end, and my Japanese conquest on the other). There was really no indication they would attack. The Aztecs were polite right up to declaring war, and I made the mistake of trading them Iron and Saltpeter.

I'll move this into another post when I get back home later today. So for now, sorry if I went off topic. Some food for thought for those who like to think about the AI.

[Nor Me]

Having no special cases seems to be part of the AI design philosophy. If an AI civ would do O.K. without the GL then why should it change its behaviour if it's in a better position? It could be done better but only with more work.

I've tested for the AI avoiding researching techs which make its wonders obsolete and found no such effect.

I suppose the idea might be that avoiding Education might mean that the civ loses an opportunity for a tech lead.

However, I'll leave you with a link to the AU203 spoiler thread. In that game, an AI with the GL seemed to have research most of the way to Military Tradition before Education. This does not appear to have been a bad move:
AU203 spoiler

[Bamspeedy]

Because a tech makes a wonder obsolete does not devalue the AI's priority towards whether or not they research it. What the AI researches is based on various other things and how many turns it will take to research (and obviously what options are available at the time it started a particular tech). The AI puts a value on a tech that has a wonder, not favoring one wonder over the other. The AI does not look several techs into the future. For example, if it has tech A and B to choose from and chooses to go for tech A, but the next turn gets tech B, which allows tech C, which is a great tech, they will keep researching tech A until they are done researching it, or get it by some other means. Once they get tech A, it will re-evaluate what tech it should choose next. The same applies to wonders (they won't change production of wonders, unless the wonder they were building got built by somebody else). Generally, the AI just starts building any wonder that is available as soon as there is a city that has just completed something and meets the requirements. Some circumstances affect this of course (they need a defender in the city first, and it seems the city needs to be producing at least 2-3 shields/turn).

What will the AI research next?

I did some tests of what priorities the AI puts on wonders. If I strip all the wonders of their prerequisites, then 80-90% of the time everyone builds the UN. A few build shakesphere's (this is what most will build if the UN gets built). After those two, then JS Bach's, Newtons, Sistine... But in one test I did, I had contact with all 24 civs and the governor wanted the Great Library more than the UN for me. So it may be possible they put more priority on it if they do have contact with 2 or more civs (otherwise, it still ends up in the top 10, though).
Testing this is frustrating, though as the governor won't allow you more than 14 wonders in a city and the governor doesn't agree with the domestic advisor. The governor chooses a wonder if you complete something, but if someone else built the wonder, then the domestic advisor comes up notifying you that you need to change production of it and suggests a totally different wonder than what the governor would have recommended. I'm thinking which wonder it chooses, may be based partially on turns to complete, similar to which tech it chooses to research. Usually the governor or domestic advisor represents how the AI would respond in a similar situation.

On a map with only me and 1 AI civ, my governor recommended the Great Library as the 6th most valuable wonder. The domestic advisor valued it as the 15th most valuable (but still the most valuable acient era wonder).

Here's a sample test (there is a random factor, so this would change from game to game).
Island map with just 2 civs, I'm the Aztecs. Starting locations are set to be on coast, with river, so all wonders are available.
Governor recommended:
UN
Shakesphere's
JS Bach's
Newton's
Sistine
Great Library
Cpernicus's
Sun Tzu's
Universal Suffrage
Magellan's
Hanging Gardens
Smith's Trading Co.
Oracle
TOE
then refuses to build anything but spearman, despite I told it to never build any offensive or defensive units.
Domestic advisor recommends (this happens when someone else completes the wonder you were working on).:
Shakesphere
SETI
Longevity
Cure for Cancer
UN
Hoover's
Universal Suffrage
TOE
Smith's
JS Bach's
Leo's
Sistine
Sun Tzu's
Newton's
Great Library
Copernicus's
Magellan's
Pyramids
Hanging Gardens
Oracle
Great Lighthouse
Great Wall
Colossus

[CerberusIV]

Thanks for the comments. It seems to confirm that a GW is just a GW to the AI and it really doesn't understand how to benefit from any particular GW. It explains a lot even though it is disappointing.

[dexters]

Looks like the AI does value the Great Library moe. I've always suspect the GL is worth more to the AI since I've seen AI civs consistently gun for it, as opposed to say, Shakespeare's theatre.

And I found confirmation in the Nov 7, 2001 Apolyton Chat with Soren...

Soren_Johnson_Firaxis] the Great Library is one of the Wonders the AI prefers... if they are under the right set of circumstances (they have met at least 2 other civs...)

[Soren_Johnson_Firaxis] the AI likes the Pyramids, the Great Library, the Art of War, the Hoover Dam, Sistine Chapel maybe? I tried to make sure the AI knew to emphasize the choice wonders since that seemed to be a flaw in previous AI's

Wonders like Shakespeare's are almost always second tier, and are the wonders the AI switches to when the current on they are building has been built by somebody else.