Overuse of Nukes

[MartinHarper]

I think AIs should value gifts of military units more highly if they are at war. Maybe 2x?

 

[Afforess]

This is partly a thought on the (iPower ^ 1.3) formula, and partly on nukes. Nukes have an iPower in vanilla BTS of 40. Modern Armors have the same iPower value. This leads the AI to believe if I have one nuke, it should build a modern armor to compensate. Needless to say, this logic is, well, illogical. If I have a nuke, I'm like 2-3x as powerful as my neighbor who has none, automatically. I would want the AI to realize this, but just changing the iPower tag to 400 for ICBM's would just cause AI's without access to nukes to spam units like mad to keep up. So, I think there should be, if there isn't already, some kind of weight separate for nukes. That way, If I have nukes, the enemy doesn't focus on ground troops anymore (why bother, I can just 'nuke em?) and focus on getting the resources/techs for Nukes, or building them if they already can. That way their is more of a MAD build up of nukes.

 

[Roland Johansen]

Jdog already talked about splitting the power value in land, sea and air power. Maybe nuke power should have a separate value so that the AI can also focus on that part of its power.

 

[MartinHarper]

In my iPower mod I've set one nuke = five modern armour. I'm not worried about nations without many nukes spending more resources on their military - militarilly inferior nations only do this up to a point. The big difference should be that they become "scared" of nuclear nations, less willing to declare war, and more willing to vassalise.

 

[CivFanCCS]

After reading this forum, I though wow, this is finally being addressed. I recently finished a game where this was a huge issue.

Myself, the Incans and Egypt are all basically super-powers. Shaka had a small nation of 4 cities on the Incan continent and was constantly under attack. I kept sending stacks of Armor and other troops to help them. It got to the point where Shaka was winning the war, taking a total of 2 cities from them. The war ceased, but I had a huge Aircraft carrier fleet of their shores. (8 Carriers, all upgraded in ROM for additional units, so 7 aircraft per carrier), plus like 20 destroyers and battleships.
Getting to the point, the Inca's declare on Shaka again and Shaka asks me for help, I agree and the Incans the next turn nuke my carrier fleet and send a nuke at my capitol. I have over 60 ICBMS and because I was UN leader I had made non-nuke agreement in the past and it succeeded, so the Incans only had 3 nukes. They still nuked me????? just crazy.They had about 25 cities and each one got either 2 or 3 nukes, effectively destroying them, while I suffered a loss of carriers and a partially handicapped city.

The AI def. needs to take this into account. If they have 3 nukes and another civ has 10,50,100 maybe it would not be such a good idea to be the first to launch. The AI should be coded to expect a counter attack, just like a human would. This was basically a suicide, because even though I agreed to join the war, it would have been a limited war. I had no intention of invading.

AI should def. consider diplo penalties, and even if they need to use them.

As for global warming, I don't think any nation that comes to the point that they are deploying nukes would consider this.

 

[Roland Johansen]

@CivFanCCs:

Why would you only use your nukes when the AI would use theirs? That's a real world logic that you're using on this game but it doesn't actually make any sense. In this game, no-one cares about the global effects of a nuclear war as it hurts everyone equal. And after a war has already started, nuclear deterrence shouldn't have any effect any more. It's like saying that a weak nation shouldn't use its weapons against a strong nation that declared war against it because the strong nation might use its own weapons. Since the strong nation declared war, it's a pretty reasonable assumption that it will use its weapons. So the weak nation can better use its weapons as effective as possible. The carrier fleet sounds like a very good target as it is a serious threat and contains a large number of units. Of course the weak nation will lose but reading your description that was a foregone conclusion.

 

[r_rolo1]

Now the real question... would you not use the nukes if HC really had not used some against you? Not using them would be unrational

EDIT: Roland beated me as usual

 

[Roland Johansen]

A huge coincidence that we posted within 1 minute, more than one and a half hour after CivFanCCs posted his game experience.

 

[Dom Pedro II]

@CivFanCCs:

Why would you only use your nukes when the AI would use theirs? That's a real world logic that you're using on this game but it doesn't actually make any sense. In this game, no-one cares about the global effects of a nuclear war as it hurts everyone equal. And after a war has already started, nuclear deterrence shouldn't have any effect any more. It's like saying that a weak nation shouldn't use its weapons against a strong nation that declared war against it because the strong nation might use its own weapons. Since the strong nation declared war, it's a pretty reasonable assumption that it will use its weapons. So the weak nation can better use its weapons as effective as possible. The carrier fleet sounds like a very good target as it is a serious threat and contains a large number of units. Of course the weak nation will lose but reading your description that was a foregone conclusion.

Yeah, but on the other hand, if the use of nuclear weapons makes other players more likely to use them, then each player has a motivation not to be the first to use them. The AI should expect, as a real world country should, that a nuclear strike should be answered by a retaliatory strike. This is true for both an AI and a human target. If the AI hits me with the only 3 nukes it has and I have enough to leave their whole country a charred wasteland... I'm going to do it! Precisely because I'm not the leader of a real country and those aren't really millions of real people being vaporized by my quest for revenge.

The question then becomes, is it better for the AI to launch the nukes if they have them and basically consign themselves to nuclear annihilation? Or is it better for them to seek first to fight a conventional war and then if nukes are used against them to respond in kind? And don't you also take a diplomatic hit for using nukes from everybody else? So using them needlessly has no real upside except to maybe delay what might have been only a half-hearted invasion attempt and replace it with a hellish firestorm from above to be followed shortly thereafter by a committed process of conquest and subjugation.

And actually, he didn't say the Inca was a weak civ. He said they only had 3 nukes compared to his 60. He actually said that they, like him, were a global super power. In that case, a conventional war might not be favorable, but encouraging the human player to let loose with the full fury of a considerably larger nuclear arsenal is even less favorable.

 

[r_rolo1]

The issue is: why would the AI be coded to think that the war will be a gentlesman quarrel, with unwritten rules about what weapons to use, and not a gloves out street fight, where everything is allowed? A lot of the humans will play the former and saying to the AI to not use the nukes ASAP is pretty much inviting the human enemy to destroy their enemy nukes before the AI finally thinks it is a good idea to use them. And how would the AI discern about a human that will not use nukes before nuked and one that will use them at t0 of war ( or worse, one that will nuke them to submission without using land troops, like I did once) ? Lacking a good way of making the AI aware of the human real intentions ( and roleplay inclination/lack of it ) IMHO it is better to say to the AI that all the fights are rabid kale borroka, where they might overuse nukes sometimes , then saying to the AI to try to guess if it is a good idea to not use the nukes just because the human has not used them yet, that would be pretty much the same to say to the AI to wait for it's demise with nukes stored in the backyard some of the times ....

P.S DP II, if you read it right, the poster you are refering to says that the Inca already had lost 2 cities vs a 4 city Shaka acting as proxy of the human in that game. That clearly shows that the Inca were not up to the chalenge in conventional terms IMHO

 

[MartinHarper]

In the real world, this is achieved by having a No First Strike policy. In a dedicated mod, such as that for civ editor11, this could be coded as a new type of civic, for example. It might have options such as Non-Nuclear, National Interest, No First Strike, Territorial Integrity, and Unilateral Disarmament. You would need to add a rule that you cannot launch nuclear weapons during Anarchy.

 

[civ editor11]

Is there any way to do that it is basically everything I wanted. What do I need to code or is it already somewhere around here for 3.19?

 

[Dom Pedro II]

The issue is: why would the AI be coded to think that the war will be a gentlesman quarrel, with unwritten rules about what weapons to use, and not a gloves out street fight, where everything is allowed? A lot of the humans will play the former and saying to the AI to not use the nukes ASAP is pretty much inviting the human enemy to destroy their enemy nukes before the AI finally thinks it is a good idea to use them. And how would the AI discern about a human that will not use nukes before nuked and one that will use them at t0 of war ( or worse, one that will nuke them to submission without using land troops, like I did once) ? Lacking a good way of making the AI aware of the human real intentions ( and roleplay inclination/lack of it ) IMHO it is better to say to the AI that all the fights are rabid kale borroka, where they might overuse nukes sometimes , then saying to the AI to try to guess if it is a good idea to not use the nukes just because the human has not used them yet, that would be pretty much the same to say to the AI to wait for it's demise with nukes stored in the backyard some of the times ....

I think you might be missing the point though. You're talking about battlefield etiquette. I'm talking about risk vs. reward. You might have a point if these two countries have similar numbers of nukes. Or at the very least, if the AI player has enough to nuke all the other player's cities where a first strike policy could potentially cripple the opposing enemy and destroy many or all of the enemy's ICBMs before they ever launch. But we're not talking about that. We're talking about an AI player that was hopelessly out-gunned in nuclear arms without any possibility to build more (since they'd agreed to the ban).

The question for the AI is not: Should I use them first if the human player hasn't? The question is: Should I trade possible destruction for certain destruction? And the answer there is absolutely not. And it's precisely because the human player is not going to fight a gentleman's war, that the AI shouldn't launch the first nuclear strike when it has so few nukes to use. If an AI nuked one of my cities, I'm going to nuke ALL of theirs. Period. Because I'm a bastard. Right? Fine. So why should the AI invite that kind of devastation upon itself when its the only logical consequence of initiating a nuclear attack?

In that case, it is necessarily in the AI's best interest NOT to start the nuclear war. But if we just want the AI to be a spoiler for the human player, then sure... fine... that makes sense.

P.S DP II, if you read it right, the poster you are refering to says that the Inca already had lost 2 cities vs a 4 city Shaka acting as proxy of the human in that game. That clearly shows that the Inca were not up to the chalenge in conventional terms IMHO

Fair enough. I missed that part.

 

[PieceOfMind]

Note ICBMs cannot be destroyed by other nukes. To my knowledge, tactical nukes can be destroyed by nukes but ICBMs can only "indirectly" be killed by ICBMs if the city is captured or razed as a result.

Realistically, an AI does not often take a city on the first turn of war, making it unlikely to take out many if any of the opponent's ICBM stockpile.

 

[Jawa'sRevenge]

This is a very difficult decision to call because unlike most idea's in civ, a nuclear war has never taken place and we can only guess at country would do if in one.


On country (few nukes) vs country (lots o nukes) consider Iran, they want nuclear weapons (obviously) because there leader/dictator is crazy: he would build one nuke send it at Israel and waited for the US and Russia to turn his country into radioactive slop. Maybe civs fascist governments could do that.

 

[MartinHarper]

I think it's more likely that Iran wants a nuke for its deterrent effect against Israel. A pre-emptive nuclear strike on Israel would achieve none of Iran's objectives, stated or unstated, and would invite destruction from Israel's superior conventional and nuclear arsenal. However, if Israel declared war on a hypothetical nuclear-armed Iran, I would expect Iran to use those nukes.

If the AI's response to a declaration of war is to nuke the person who declared war on it, then this means that players who don't want to invite that kind of devastation won't declare war on it. This maximises the deterrence effect of its limited nuclear arsenal. Now, some players will declare war on a nuclear power and then complain that they got nuked. However, this is a player failure rather than an AI failure.

 

[r_rolo1]

@ DP II

I guess I haven't expressed myself as clearly as I should ( a common ocurrence when discussing how to program a machine, i guess ). In fact I was talking not only of battleship ettiquette, but also of what you called risk vs reward. In fact you can't separate both in this area, because the risk is directly proportional to what the enemy is willing to put in the field. My point was that, because there is no way in the current game of the AI knowing what are the intentions regarding nuke usage ( in mods where that is possible/mandatory, things would be diferent OFC ), when the AI is at disavantage as in the example above, what is better: a) the AI assuming that the enemy will use their vast stockpile of nukes b) assuming the enemy will not use the nukes unless nuked? The issue is , if the human decides to use the nukes, certain destruction is assured no matter what and only if the human decides not using nukes unless provoked ( a thing that the AI has no way to know, remember ) not nuking first might look like a competing option, in spite of being as bad as a option as all the others if the non-nuke side is skewed against them. In both cases ( nukes regardless and nukes if nuked ) , nuking first definitely brings short term advantages, that might be enough to win the game ( see this for a good example .... ) or to stem certain death for a while.

My point in resume: the discussion about possible vs certain discussion you brought might be posed, but it comes after the AI decision ( or lack of it ) about what the enemy ( human or AI ) will do regarding nukes. Due to the ignorance of the AI regarding to how the enemy will act in that regard, in atleast half of the possibilities nuking first is always the better option ( when the enemy will nuke no matter what ) and in the other half ( when the enemy will not nuke unless nuked, the classical prisioners dillemma solution ) it is not clear if not nuking first will be better than nuking first in all cases. Seeing things in this perspective , IMHO, it is objectivelly better for the AI to assume that striking first is a better option than clinging on the hope that their enemy will be a nice guy and refuse to use a weapon they have just because.

(loosely based on the work of the paranoid strategists that planned MAD in RL )

 

[Dom Pedro II]

@ DP II

I guess I haven't expressed myself as clearly as I should ( a common ocurrence when discussing how to program a machine, i guess ). In fact I was talking not only of battleship ettiquette, but also of what you called risk vs reward. In fact you can't separate both in this area, because the risk is directly proportional to what the enemy is willing to put in the field. My point was that, because there is no way in the current game of the AI knowing what are the intentions regarding nuke usage ( in mods where that is possible/mandatory, things would be diferent OFC ), when the AI is at disavantage as in the example above, what is better: a) the AI assuming that the enemy will use their vast stockpile of nukes b) assuming the enemy will not use the nukes unless nuked? The issue is , if the human decides to use the nukes, certain destruction is assured no matter what and only if the human decides not using nukes unless provoked ( a thing that the AI has no way to know, remember ) not nuking first might look like a competing option, in spite of being as bad as a option as all the others if the non-nuke side is skewed against them. In both cases ( nukes regardless and nukes if nuked ) , nuking first definitely brings short term advantages, that might be enough to win the game ( see this for a good example .... ) or to stem certain death for a while.

My point in resume: the discussion about possible vs certain discussion you brought might be posed, but it comes after the AI decision ( or lack of it ) about what the enemy ( human or AI ) will do regarding nukes. Due to the ignorance of the AI regarding to how the enemy will act in that regard, in atleast half of the possibilities nuking first is always the better option ( when the enemy will nuke no matter what ) and in the other half ( when the enemy will not nuke unless nuked, the classical prisioners dillemma solution ) it is not clear if not nuking first will be better than nuking first in all cases. Seeing things in this perspective , IMHO, it is objectivelly better for the AI to assume that striking first is a better option than clinging on the hope that their enemy will be a nice guy and refuse to use a weapon they have just because.

(loosely based on the work of the paranoid strategists that planned MAD in RL )

The question of whether or not the human player will initiate a nuclear attack is largely irrelevant. In fact, the example you posted actually undermines your case because the player clearly states that he initiated the nuclear war. At this point, the AI already has nuked cities and a crippled army and only needs to buy a couple turns to win the space race victory.

The question is: Can the AI buy more time with intact cities capable of building new units and a standing army? Or would they be better off launching two nukes, and accomplishing little more than enraging an enemy with a vastly larger nuclear arsenal? If the human player is willing to use nukes, then it doesn't matter when the AI launches the nukes. Since ICBMs aren't destroyed when the cities they are in are nuked, the prisoner's dilemma is not appropriate. In the prisoner's dilemma, the second prisoner acting first prevents the first prisoner from taking the deal. In this scenario, it would be as if the second prisoner's decision had no effect on the first prisoner's options. The AI can still nuke even if the human player launches first.

We have two possible circumstances.. a moral human player (one who will not nuke first) or an immoral human player (one who WILL nuke first). If the AI does not nuke a moral human player, no one will be nuked. If the AI does not nukes an immoral player first, they'll both be nuked when the human attacks and the AI retaliates. If the AI nukes the moral player first, both will be nuked. If the AI nukes the immoral player first, both will be nuked. So the AI player has a 50% chance of being nuked by not attacking first vs. a 100% chance of attacking first.

What you're asking the AI to do is only to assume the immoral player situation and commit itself to a mutually assured destruction scenario in total disregard to the consequences of its actions. You want the AI to assume that there will be no retaliation for launching first. And let's not forget that the human player has options as well. There are penalties to the human player for nuking... 1) Retaliation, 2) War weariness spikes, 3) Diplomatic penalties with civs that might potentially enter the war, and 4) Global Warming. If the human player knows that the AI will probably not use its nukes first, then he has less of an incentive to strike first. Personally, after playing civ for 12 years and across three games, I have used nukes exactly once. It's not because I feel bad about using them. It's because I find I'm more than able to handle the AI without having to take the risk of nuking.

Don't get me wrong though. I'm not saying the AI should never nuke first. I'm saying that in the particular scenario presented earlier where the AI has far fewer nukes than the enemy and where the number of nukes the AI does have cannot possibly stop a future invasion and no ability to build more, there's nothing to be gained. Sure, he got that carrier force, but at the expense of his entire national infrastructure. Using nukes freely, the human player should be able to just sweep through the shattered remains of their empire.

Now... if the AI player has built the SDI and has Bomb Shelters in all its major cities, and if the AI player has more nukes, then the risk vs. reward balance shifts completely. Then it might benefit the AI to nuke first. Or if the AI player is a few turns away from winning the game.

 

[Dom Pedro II]

On country (few nukes) vs country (lots o nukes) consider Iran, they want nuclear weapons (obviously) because there leader/dictator is crazy: he would build one nuke send it at Israel and waited for the US and Russia to turn his country into radioactive slop. Maybe civs fascist governments could do that.

While I don't think he really wants to do that (I think he's actually a much slicker customer than people in the West generally give him credit for), its not really relevant to the current topic. If we assume he's a completely irrational character willing to engage in self-destructive behavior, it does us no good whether this should determine AI behavior. This mod is all about making the AI smarter. Humans do stupid things, but we don't need to code stupid behavior into the AI. The AI does enough stupid things already

I think it's more likely that Iran wants a nuke for its deterrent effect against Israel. A pre-emptive nuclear strike on Israel would achieve none of Iran's objectives, stated or unstated, and would invite destruction from Israel's superior conventional and nuclear arsenal. However, if Israel declared war on a hypothetical nuclear-armed Iran, I would expect Iran to use those nukes.

If the AI's response to a declaration of war is to nuke the person who declared war on it, then this means that players who don't want to invite that kind of devastation won't declare war on it. This maximises the deterrence effect of its limited nuclear arsenal. Now, some players will declare war on a nuclear power and then complain that they got nuked. However, this is a player failure rather than an AI failure.

I disagree. The most powerful nuke is the one not used. If Israel declares war on Iran and Iran responds by nuking Israel, Iran will then be out of bargaining chips. On the other hand, But this is not the same as if Iran says to Israel, "Stop this war or we will nuke you." and then carries it out if Israel refuses. Again, I'm arguing this from the perspective of AI behavior. Iran might very well nuke as soon as Israel declares war, but in terms of what an rational actor would do, delivering an ultimatum would be more effective than simply nuking the enemy.

Which might actually be interesting in game... Imagine we make it so that the human player declares war on the AI player, the AI player comes to them and says "Cease your agression against us, or we will use our NUCLEAR WEAPONS!" And if the human tells them to get lost, the AI will fire off the nukes.

 

[Afforess]

The question of whether or not the human player will initiate a nuclear attack is largely irrelevant...

I agree. Really, the comparison should be between the size of the arsenals, not distinguishing between human and AI.

Also, I think other scenarios must be considered for the AI. Consider the scenario where an AI has enough land for a domination victory, but not enough population. Then, if the AI has enough nukes, or is rather close to the right population, it might be wise to nuke other AI's largest cities, since their population will decrease, and possibly putting the AI over the threshold for victory.

I have used nukes exactly once...

You haven't played many games with Cultural victory enabled then, I'm guessing. I uses nukes as a last ditch attempt at stopping a cultural victory. 2 nukes, some well-placed paratroopers, and one city ruins later, the AI is much farther away from a cultural victory.

The real problem with Nukes for the AI is when to use them. There are many times when nukes can forestall an enemy victory.

For instance, I can imagine a similar scenario as the one above, except, instead of a cultural victory, a diplomatic victory. Let's say Player 1 has the UN, and is basically a friend of everyone. Player 2 has a larger army and bigger country, and doesn't want to lose. Player 2 nukes the city with the UN and razes it to the ground preventing a diplomatic victory.

Strategic strikes like that are what I would really like to see the AI do.