Does the AI know how to get Eureka bonuses?

[mantis2007]

Based on recent "Let's Play" style videos of the pre-release "review" build, I noticed that the AI tends to fall far behind the human player in technology. I am concerned that this suggests a balance issue, and possibly a problem with the AI's inability to take advantage of the new Eureka system.

The Eureka (and Inspiration) system, of course, allows a player to gain 50% of the progress toward unlocking a new Technology (or Civic) by completing a specific objective. For example, building a Quarry triggers the Eureka bonus for Masonry. In my opinion, this system will be very difficult for the AI to use properly.

In part, this is difficult for the AI simply because it's new. Even if the AI has a capability to recognize and take advantage of this system, it is likely that capability is rudimentary at this stage in Civ VI's development. More crucially, however, is the real possibility that it is very difficult or impossible to program an AI to prioritize effectively Eureka moments alongside other objectives.

Consider the Eureka objective for a late-game tech, like Advanced Flight: Build 3 biplanes. It's not at all clear whether it's "worth it" to go for this Eureka for any given player. Should a player who is not focused on military forego other choices to build biplanes just to rush Advanced Flight? That is a very interesting and challenging question for players, and one that the AI almost certainly cannot adequately comprehend.

Because human players will almost certainly have a better understanding of how to use the Eureka system than the AI, this gives human players a major advantage in gaining technology. This can be balanced with various handicap bonuses for the AI, but that may not be satisfactory. Again, based on videos of players using the "review" build, it seems that at Prince difficulty the AI cannot keep up in tech with competent human players.

tl;dr: The AI probably doesn't understand the Eureka (and Inspiration) system in the "review" build, which means science (and culture) might be significantly imbalanced in the release build.

 

[Ryika]

I'd assume it won't understand Eurekas, but due to their bonuses they'll probably get many of them anyway, especially on higher difficulties.

And if not, then just increasing their Culture/Science Modifiers further would do a good-enough job.

Don't see it as much of a problem.

 

[ilker]

I think instead of giving extra bonuses to AI, exempting the AI from Eureka system (by simply giving all eurekas to AI by default) is much fair and competent.

edit: Lets take it that way: I guess all the eureka system is about to give human players more of a feeling of use different strategies for each game and use all different aspects of the game according to their environment, game situation etc. (to increase diversity and fun) which is already there for AI, although not playing the best moves (or one of the good choices from the infinite selection of what can be done in that turn, which is a necessity of a competent AI ). And also it does not mean directly an AI science advantage in this way. Thus I really prefer this over an AI having flat bonuses of extra settlers, builders, more combat strength for same units (which I guess I will be hating most in the difficulty levels), more default amenities, and ability to grow larger cities, more production etc. Similarly I would prefer if AI is not good using the units very efficient, an AI that can accomodate more units is more fair than an AI can produce units faster, or stronger units.

I also worry for the AI for the new agenda system, in multiplayer we as human players dont have likes or dislikes apart from concerning victory conditions, so I think new agenda system is another handicap for AI. Maybe it could be better more dynamic agenda reveal, where we can see for example how AI is thinking (programmed) for that specific game circumstances. That could be more difficult to implement but very cool I guess

 

[mantis2007]

Thanks for the replies -- I'd like to ask that we keep this thread about the AI and the Eureka system, and leave discussion of the (mostly unrelated) agenda system for other threads.

I'd assume it won't understand Eurekas, but due to their bonuses they'll probably get many of them anyway, especially on higher difficulties.

And if not, then just increasing their Culture/Science Modifiers further would do a good-enough job.

I agree that balancing as you suggest is possible, but I still worry that the initial release won't yet have the tech and civic systems balanced well--because of Eurekas.

I disagree with ilker that the AI should use different tech and civics trees that do not have Eureka bonuses. That would require completely rebalancing the AI for those AI-only trees, which is probably far more difficult for the developers (and more likely to cause balance issues) than the alternatives.

 

[KrikkitTwo]

I think instead of giving extra bonuses to AI, exempting the AI from Eureka system (by simply giving all eurekas to AI by default) is much fair and competent. I also worry for the AI for the new agenda system, in multiplayer we as human players dont have likes or dislikes apart from concerning victory conditions, so I think new agenda system is another handicap for AI. Maybe it could be better more dynamic agenda reveal, where we can see for example how AI is thinking (programmed) for that specific game circumstances. That could be more difficult to implement but very cool I guess

No, that is a very bad idea, since the Eureka system is supposed to shape the civ.

Ideally
1. AI would prioritize things that could give them Boosts for tech/civics they like
2. AI would prioritize researching tech/civics they have the Boost for

Either one of those would be good, by themselves... or together..but neither is really necessary.

Rather than altering game mechanics for the AI, if the AIs X is too slow give it a boost to X directly.

 

[Todd Hawks]

Not only is the AI unable to work with it, I think the whole concept of the Eureka bonus is highly flawed. It's a very gamey idea and it makes players hunt the bonuses (kind of like in-game achievements) by often making them do things they usually wouldn't.
I'd rather have them remove it completely or reduce its effect by a lot.

 

[BrokenBokken]

Not only is the AI unable to work with it, I think the whole concept of the Eureka bonus is highly flawed. It's a very gamey idea and it makes players hunt the bonuses (kind of like in-game achievements) by often making them do things they usually wouldn't.
I'd rather have them remove it completely or reduce its effect by a lot.

Honestly, I'd rather have them remove 'beakers' as it is entirely and have technologies completely advance by things you do. We didn't discover irrigation because of sitting in a library. We discovered it because farmers naturally progressed that way to fit their needs. I think it would be amazing if a game could actually implement this type of growth and make it fun and balanced.

 

[KrikkitTwo]

Not only is the AI unable to work with it, I think the whole concept of the Eureka bonus is highly flawed. It's a very gamey idea and it makes players hunt the bonuses (kind of like in-game achievements) by often making them do things they usually wouldn't.
I'd rather have them remove it completely or reduce its effect by a lot.

The idea is that "chasing the Eurekas" Should be a waste of time Unless you were doing something like that anyways... ie building 3 Spearmen is a waste unless you are going military, building an early wonder is a waste unless you are going cultural.

Basically, getting the eureka should be too expensive enough to be worth Just the Eureka.. it should only be worthwhile if it is the Eureka+ something else.

If they do that... then it is working well.

Honestly, I'd rather have them remove 'beakers' as it is entirely and have technologies completely advance by things you do. We didn't discover irrigation because of sitting in a library. We discovered it because farmers naturally progressed that way to fit their needs. I think it would be amazing if a game could actually implement this type of growth and make it fun and balanced.

So basically 100% Eurekas...that would be gamey

 

[Danzo_87]

The idea of eureka and inspiration are to make you play to your map, if your surroundings demand you to play more militaristic you will be flying through the military part of the tree, its how they execute this concept that will determine whether this system will work or not. If doing the optimal play like what player did in civ 5 is on the same speed as a player who utilize eurekas, then this system has failed

 

[UncivilizedGuy]

Honestly, I'd rather have them remove 'beakers' as it is entirely and have technologies completely advance by things you do. We didn't discover irrigation because of sitting in a library. We discovered it because farmers naturally progressed that way to fit their needs. I think it would be amazing if a game could actually implement this type of growth and make it fun and balanced.

This will likely be possible to mod. Simply increase the research costs to an absurd number then give Eurekas a 100% boost. Thanks for the idea!

Although the beakers are symbolic of knowledge accumulation. The beakers are realistic in one aspect by simulating tech diffusion through trade.

 

[Big J Money]

I agree this isn't a big deal. If Firaxis is able to collect game data analytics from the game (like a lot of companies do these days) then they can calculate the average Eureka deficit the AI tends to be in, and give them a commensurate bonus to science and culture acquisition.

They probably already done some of this in their internal testing, but since I imagine the state of the game has constantly been in flux, they'll get much better data after release.

 

[Ornen]

I agree that the AI should be given a flat boost in place of Eurekas. I'd be happy if the AI knew how to consistently pursue eurekas, but 0% concerned if we simply disabled them and gave the AI a bonus equivalent to however often players reliably hit eurekas. Then scale that bonus up/down from Prince as part of the AI's regular bonuses.

I like the idea of shifting research entirely off beakers, but I'm completely fine with the research system, which works well. If you don't want abstractions, you don't want civ – pretty much every mechanic is an abstraction.

 

[Danzo_87]

https://www.reddit.com/r/civ/comments/5853nb/civ_6_producer_sarah_darney_says_a_ton_of_work/

from this thread in reddit it seems like they used all the preview lets plays as their data source to fine tune the AI (if that is true), so let us see how "improved" the AI will be this weekend

 

[Grumpbeard]

This will likely be possible to mod. Simply increase the research costs to an absurd number then give Eurekas a 100% boost. Thanks for the idea!

Although the beakers are symbolic of knowledge accumulation. The beakers are realistic in one aspect by simulating tech diffusion through trade.

You would also need to give everyone the 3 eurekaless techs for free, I guess, and also what would you do about the end game ones that don't have Eurekas? But all in all an interesting variant.

 

[Frank327]

Like Marbozir said in his videos, eureka's are not like missions. They're pretty easy to get and just cause you to advance faster in directions you're already working on. If an AI builds builders and settlers early then it'll get lots of tech eureka's from the worker and the early empire tech without making a special effort for it. Same goes for military AI's and military techs/civics.

As long as the AI is doing something useful and has some kind of focus to its actions then it'll probably use eureka's just fine. The AI just needs proper weights to research the right things at the right time.

 

[REDfeller]

Have to wait and see how it pans out on higher difficulty. Human player may need the boosts just to keep up!

I agree that the criteria for boost seems too easy and that the % boosted seems way too much.

Are there any indications that they differ based on difficulty level?

Also worth noting that the human player will mostly beeline tech, so boosting one effectively boosts the next tech in line and so on. .

Does the AI beeline tech? Rome - Iron working , etc.

 

[mantis2007]

Thanks for the discussion, everyone.

I have no doubt that the AI will, eventually, get balanced in terms of its speed in progressing through the science and civic trees. At the same time, I am confident that in the release build we will play this weekend, there will be significant balance issues in this area.

The Eureka system is, in my opinion, a great addition to the Civ formula. It adds strategic depth and complex choices--both of which are what strategy games are all about. My concern is not that the whole Eureka system is a mistake or anything like that. But it is a new system that the AI will struggle with. The best way to get the AI to progress at the "correct" pace through the two (!) tech trees is, in my opinion (shared by many in this thread), observation of actual games, coupled with carefully measured bonuses or penalties for AI players. There is simply no way the devs have enough information at this stage to complete this process. Therefore, we can expect AI players to either fall too far behind human players in tech (and civic) progress at easier difficulty levels, and they will probably zoom too far ahead of human players at harder difficulty levels.

Consider the challenge that REDfeller notes: the human player will carefully execute tech "beelines." This is, of course, a longstanding strategy among Civ players. In Civ VI, for the first time in the series, such "beelines" can take advantage of Eureka moments, some of which require planning several turns in advance to maximize efficiency. I frankly think there is no way that the AI will ever match human players in balancing all of the complexities required to strategize around "beelines" in Civ VI. Balancing around this presents a unique challenge for the AI devs. It will be some time before they've gotten it right.

 

[demidyad]

In the AI Battle Royale live stream, the lead AI programmer (Andrew Derrick?) just explicitly stated that the AI do know how to seek out Eureka bonuses, but on "lower difficulties", that ability is removed.