Paranoïa

Naokaukodem

Millenary King
Joined
Aug 8, 2003
Messages
4,303
I heard Sid talk about paranoïa in Civ. I don't really remember how oriented was his speach (sorry Sid ;) ), but I think there is no mystery really about it... if the Civ Player becomes paronoïd, this is because the AI is made so.

Example: when you set in Custom or Multiplayer games "Always War", the player will be at war with every AI as soon as he encouter them. The fact is, that the AIs between themselves won't be at war at all!!!!!! Do you know what i mean?

That, is a perfect example of human-centred AI that makes the game so annoying. I won't list all the examples here, as they didn't hit me in the same time and I lost track of them usually. But this example reflects very well how is programmed the AI.

That's why i voted for an "AI that plays to win" in the appropriated thread. Because, an AI that do not play to win will be, invariably, centered on the human. Hence the paronoïd feeling. And that's not only a feeling! It's a shape that the whole game takes, in its deep gameplay mechanics.

If Sid and/or his acolytes could take the time to program an AI that does not discriminate human player and AIs, that would be a great step in order to improve Civ. And if you ask me, that would even be an essential step in order to improve Civ.
 
I heard Sid talk about paranoïa in Civ. I don't really remember how oriented was his speach (sorry Sid ;) ), but I think there is no mystery really about it... if the Civ Player becomes paronoïd, this is because the AI is made so.

Example: when you set in Custom or Multiplayer games "Always War", the player will be at war with every AI as soon as he encouter them. The fact is, that the AIs between themselves won't be at war at all!!!!!! Do you know what i mean?

That, is a perfect example of human-centred AI that makes the game so annoying. I won't list all the examples here, as they didn't hit me in the same time and I lost track of them usually. But this example reflects very well how is programmed the AI.

That's why i voted for an "AI that plays to win" in the appropriated thread. Because, an AI that do not play to win will be, invariably, centered on the human. Hence the paronoïd feeling. And that's not only a feeling! It's a shape that the whole game takes, in its deep gameplay mechanics.

If Sid and/or his acolytes could take the time to program an AI that does not discriminate human player and AIs, that would be a great step in order to improve Civ. And if you ask me, that would even be an essential step in order to improve Civ.
The AI plays a role. Maybe it has a more clear set path to victory than before, but if the AI really played to win then you would get invaded by a huge army each and every time you were close to culture, or no AI would ever vote for you in the UN. Seems not that fun to me.

Also the AI may have some biases, but that is not because the AI does not play to win. It is absurd to assume that, and it is based on feelings and not facts.
 
The AI plays a role.

It doesn't have to. ;)

if the AI really played to win then you would get invaded by a huge army each and every time you were close to culture

Culture win is out of point anyway. But yes, with culture win enabled, you would face a huge army when on the verge to win... providing the AI can "lurk" into your cities or in a Winning Conditions Panel in order to see where you're at. That's to say, if the player had the same capacities. (who would want an AI capable of seeing you're on the verge of a culture win if the player can't see it himself?)

AI would ever vote for you in the UN. Seems not that fun to me.

Same remark than before. Diplomatic Victory is out of point. I would NEVER exchange my barel of fair AI against... a barel of culture victory!

the AI may have some biases, but that is not because the AI does not play to win. It is absurd to assume that, and it is based on feelings and not facts.

I didn't say such a thing. THE MAIN POINT IS THAT THE AI IS HUMAN CENTERED. That's the core of the reasonning. (prooved by a fact: the AI don't automatically delcare wars between themselves when "Always War" activated) Now, the fact that it wouldn't play to win would make it human centered. There's a fair shade there.
 
AI is playing to win this time around according to devs. The AI is layered and thinks at distinct levels. 4 levels IIRC. Can't remember exact names of them all, but there is the Global level, Strategic, Tactical, Combat (something like that). It's like a hierarchy in that upper level decisions are followed by the lower level AI's.

If Global level decides war against 2 fronts, the level down decides how best to fight against the 2 fronts, next level decides how to coordinate units most effectively, bottom level decides individually combat between what units, etc.

This is the basis of it, I probably have the specifics wrong, because I didn't go back and look it all up. Should be exciting solely watching the AI work it's new moves.
 
AI is playing to win this time around according to devs. The AI is layered and thinks at distinct levels. 4 levels IIRC. Can't remember exact names of them all, but there is the Global level, Strategic, Tactical, Combat (something like that). It's like a hierarchy in that upper level decisions are followed by the lower level AI's.

If Global level decides war against 2 fronts, the level down decides how best to fight against the 2 fronts, next level decides how to coordinate units most effectively, bottom level decides individually combat between what units, etc.

This is the basis of it, I probably have the specifics wrong, because I didn't go back and look it all up. Should be exciting solely watching the AI work it's new moves.

Having different levels of reasonning does not mean the AI would not be human centered. I want to catch the attention of the developpers on this, this is a very sensitive subject to me, and potentially to others, and, as i said in the OP, to Sid himself.
 
It doesn't have to. ;)



Culture win is out of point anyway. But yes, with culture win enabled, you would face a huge army when on the verge to win... providing the AI can "lurk" into your cities or in a Winning Conditions Panel in order to see where you're at. That's to say, if the player had the same capacities. (who would want an AI capable of seeing you're on the verge of a culture win if the player can't see it himself?)



Same remark than before. Diplomatic Victory is out of point. I would NEVER exchange my barel of fair AI against... a barel of culture victory!



I didn't say such a thing. THE MAIN POINT IS THAT THE AI IS HUMAN CENTERED. That's the core of the reasonning. (prooved by a fact: the AI don't automatically delcare wars between themselves when "Always War" activated) Now, the fact that it wouldn't play to win would make it human centered. There's a fair shade there.

Unsure what you mean when you say that you shouldn't want an AI seeying if you are close to victory when you cannot se this yourself. You do realise that you can indeed see which AI is close to a culture win, right? And your method might make sense for the people who play 'bloodlust' civ, but I like my peaceful options.
 
Unsure what you mean when you say that you shouldn't want an AI seeying if you are close to victory when you cannot se this yourself.

For example if the AI is not based on a matrix. Then, the AI "would be the game", so it would potentially know everything. And for example, nothing garrant to have a Victory Condition panel accessible to the player in the game (like it is the case in Civ4. But it might be the case also for Civ5. But that's only an example). In that case, the AI could precisely know when you are one the verge to win a Culture Victory, and could be programmed to take part of this, while you might not.

You do realise that you can indeed see which AI is close to a culture win, right?

See the brackets above.

And your method might make sense for the people who play 'bloodlust' civ, but I like my peaceful options.

There have been so many wars in History... if you're not happy with this you could always go play Tropico or Sim City. :p
 
So you mean, AI that:
Does not even take the human player into account in any way shape or form, but sees all players equally.
Every AI plays with absolute desire and will to destroy every other player with no regards to how it does it, only that it succeeds.
Every AI will take every single advantage, no matter how cruel, distasteful, or disgusting it might be; in order to come out victorious.

In that case, I'm with you... nothing like an AI that has no holds barred to get the win.
 
So you mean, AI that:
Does not even take the human player into account in any way shape or form, but sees all players equally.
Every AI plays with absolute desire and will to destroy every other player with no regards to how it does it, only that it succeeds.
Every AI will take every single advantage, no matter how cruel, distasteful, or disgusting it might be; in order to come out victorious.

In that case, I'm with you... nothing like an AI that has no holds barred to get the win.
It is indeed nice to have a real challenge. That is what deity is for. If you do not like deity, you can go multiplayer, you are guaranteed to get into a fight there.

For the single player experience I do not like the idea at all, since I am sure the AI will go to war over anything. I would prefer it if the AI tried to beat my space date by getting a culture win slightly faster, rather than the AI declaring war all the time.

Sure wars are a part of the game, but it also should be just that, a part. It needs not to be the main focus, there are plenty of games that focus on that.
 
So you mean, AI that:
Does not even take the human player into account in any way shape or form, but sees all players equally.

Exactly! :)

Every AI plays with absolute desire and will to destroy every other player with no regards to how it does it, only that it succeeds.

Yes. For example, in going into war very early... just to grab a defenseless worker working on the edge of a city, or to take an undefended capital. (on the base of Civ4, Civ5 will be different and not allow such exploits)

Every AI will take every single advantage, no matter how cruel, distasteful, or disgusting it might be; in order to come out victorious.

Providing the player has the same potential.
 
It doesn't have to. ;)
[...]
I didn't say such a thing. THE MAIN POINT IS THAT THE AI IS HUMAN CENTERED. That's the core of the reasonning. (prooved by a fact: the AI don't automatically delcare wars between themselves when "Always War" activated) Now, the fact that it wouldn't play to win would make it human centered. There's a fair shade there.

I'd like to see some actual proof of the AI being human centred in civ4. Your always at war example does not really count in that respect, since it is not the AI that is discriminating, but the game option. (The game option is made to make the game harder for the human player. Having all the AIs at war with each other, all the time, would make the game a LOT easier for the human player.)

Actually having snooped around in the AI code, I know that it only very rarely calls the isHuman() property for players. When it does, it is usually because of interface differences. (For example, the cannot take into account the human players, attitude towards it, because no such attitude is registered. However, when dealing with an AI player it obviously takes into account its attitude, just like a human would.)

Most perceived cases of the AI being centred on the human player, like AIs ganging up on the human, are actually related to difficulty level induced asymmetry caused by the handicaps. For example, the AIs have a preference for dogpiling weak opponents. Due to the handicap bonusses given to all AIs at higher difficulty, the human player will often appear as the weakest opponent to the AI. Especially, because (almost by definition) the human is gaining an advantage over the AI through something the AI is not taking into account.

So, I'd like to see some actual proof that the AI is human centered. Preferably, backed up by some reference to the actual AI code.
 
So, I'd like to see some actual proof that the AI is human centered. Preferably, backed up by some reference to the actual AI code.
Not very likely, this thread is named paranoia for a reason...
 
About paranoia, I've the feeling that AI's scouts go directly (or how lucky they are!!)to goody huts without searching randomly like us in the fog of war. Is it wrong ?

Yes. I've seen many examples of the AI missing goody huts. It is mainly that the AIs, at higher difficulties always start, with multiple scouts, and thus scout much faster.
 
That is wrong. Civ4 was the very first in the Civ series which removed AI omniscience from the game. Also, with the advent of proper diplomatic bonuses & penalties, the AI civs also became less centered on beating the human player. I've played loads of games where I've forged strong friendships with a neighbour, then encouraged them to join me in a war against another AI civ. Civ4 was the very first Civ game where I actually *didn't* feel totally paranoid ;)!

Aussie.
 
Thought I had is that AI scouts never take huts that pop barbarian warriors for human players.
I've had a lot of games where I scout a continent with my beginning warrior, meet some AI scouts, walk into a part where they've certainly scouted already, and any huts left all spawn barbarian warriors.

Far as AI's declaring war go, I have to agree with others here that Civ4 is the first game where I'm not paranoid. Obvious exceptions like Monty and other highly agressive civs not counting of course, though they're just as likely to attack AI's than players.
 
Yes. I've seen many examples of the AI missing goody huts.

They left only those with barbarians inside!That's why I'm feeling paranoïd with the AI'S Scouts.


Try to let some warriors/scout in observation in several points in order to see all the landscape, you will see the AI's scouts go straight to the huts.
 
Good point Trias. I may have misunderstood the function of this option as I play multiplayer and it's the way it is used in multiplayer. Nonetheless, this option reflects how the game mechanics or difficulty levels can be human centered. (with arbitrary bonuses for the AIs)

As to examples of the AI being human centered, here are some i can remember:

* the AI doesn't play to win. Therefore it allows, for example, the human player to exploit its behavior like not declaring war during the first X turns. It is not the case of other AIs that develop in the eventuality of an early war, to prevent the player, and only the human player, for too easy early conquests.

* The AI leaders propositions. The Ais with bad relations to us will always harrass us with propositions we do not like, demanding us to choose a side camp. "You are with us or against us". I strongly doubt that AIs demand things to other AIs like that. (on which basis would it answer other than "I already have bad relations with you so no", in which case the proposition itself is useless)

* The AIs seem a lot more enclined to declare war to the player than they are to the other AIs. In my last game, i was going fine and the only war declaration was an AI one against me. I had not too many defense power but I think i had one not very far from other AIs.
 
They left only those with barbarians inside!That's why I'm feeling paranoïd with the AI'S Scouts.
This simply cannot be true, since the game only determines the contents of a hut when they are popped. They simply cannot know whether a hut will pop barbarians or not. (What pops from a hut also depends on the player doing the popping. Huts close to a players (in the so called safe zone) will never pop barbarians for that player. The result of popping a hut depend on the players handicap level.

Late game, huts are just more likely to pop barbarians, because some of the other options are no longer available.
 
Good point Trias. I may have misunderstood the function of this option as I play multiplayer and it's the way it is used in multiplayer. Nonetheless, this option reflects how the game mechanics or difficulty levels can be human centered. (with arbitrary bonuses for the AIs)

As to examples of the AI being human centered, here are some i can remember:

* the AI doesn't play to win. Therefore it allows, for example, the human player to exploit its behavior like not declaring war during the first X turns. It is not the case of other AIs that develop in the eventuality of an early war, to prevent the player, and only the human player, for too easy early conquests.
I'm not sure what you are trying to say here.

As far as I know the algorithms contain no explicit prohibition against early declarations of war. It is just that the conditions for the AI to consider going to war are unlikely to be met in the early game.

The general problem is that the AI in civ4, just does not have plan, period. It plays on basis of a turn by turn heuristic. (With some small exceptions, for so general overall modes it can enter.)

This is supposed to be fixed in civ5, with long term strategy layers for the AI.



* The AI leaders propositions. The Ais with bad relations to us will always harrass us with propositions we do not like, demanding us to choose a side camp. "You are with us or against us". I strongly doubt that AIs demand things to other AIs like that. (on which basis would it answer other than "I already have bad relations with you so no", in which case the proposition itself is useless)
This is actually true. (You will never see two AI players have a "You refused to help!" penalty against each other.) This however is the result of an information mismatch. Since all the AIs respond deterministically to diplomatic request, an AI player always knows how another AI will respond to a request, and will never place a request that will be denied by the other AI.

The AI can however not know how a human player will respond to a request, leading it to make some very silly (and annoying) requests.

* The AIs seem a lot more enclined to declare war to the player than they are to the other AIs. In my last game, i was going fine and the only war declaration was an AI one against me. I had not too many defense power but I think i had one not very far from other AIs.
My experience, is that AIs have absolutely no qualm of going to war with each other. I'm also pretty certain, that the AI algorithms that decide whether an AI goes to war, do not contain an isHuman() check.
The game from your example sounds like it had a religious lovefest going on. (With all player having the same religion.) There must have been some other factor (or combination of factors) that lead the AI to decide that you were the most attractive war target. For example, how were your relationships with the AI players?
 
Back
Top Bottom