Leader Personalities

One goal in Vanilla Enhanced is for AIs to use unique bonuses like a human would. A human playing Alexander will typically ally with citystates and attack major civs. There's several factors I used when determining a personality:

• What were the leader's vanilla AI stats?
• Does the leader have militaristic bonuses?
• Did the civilization historically involve citystates or confederations?
• Approximately the same number of leaders assigned to each personality type.

In addition to what's listed in the table below, there's other things involved. For example, the "conqueror" personality type values military strength above all else, so they might backstab you if they sense weakness, even if you're friends. Some leaders have additional tweaks encouraging the AI to better use their unique traits/units/buildings.

An AI's personality type determines its leader priorities.



.

Should there be a table below?

Sounds good. I will give this baby a whirl.

I've been tweaking with this and running autoplays for the past two days. I'm relatively happy with how things play out now in the autoplays, though I haven't playtested any "normal" games yet. Leaders appear to be forming factions (instead of a global lovefest or denounce-spree), the ones with a CS priority really go for those alliances, and warmongers build large armies.

The faction-forming is due to the ruthless/merciful split. Patterned off Civ IV style AI tendencies, ruthless characters have low warmonger hate and favor neutral attitude, while merciful characters hate warmongers but favor friendly attitude. The merciful ones tend to team up since they start out friendly and have high willingness to DoF. The ruthless characters band together simply because once they go to war, all the merciful ones hate their guts. Ruthless players should be very willing to go to war with anyone they sense weakness in, even friends, as I gave them a low loyalty rating. Merciful characters have a high loyalty rating and should rarely if ever backstab.

I think I reached the limit of what I can do in the personality/flavor files. For example, in one autoplay Rome had about 20 fantastic offense units on the border of their German neighbor, yet never went to war even though they disliked Germany, and Rome had high boldness and favor-war values. I think this is the area of the game that'll rely on your mod, Sneaks.

Currently I've been doing the autoplays with only these personality changes in effect, now I'll try some integration testing with WWGD active as well.

If you want to look at or tweak the files, they're Personalities.xml and Flavors.xml in the Balance - Leaders subfolder. They actually contain the entirety of the leader files, because it was a time-saver to just run a few regex replaces on the originals for generating update blocks (instead of selectively including only the changed data).

@SSgtDuke
Yep the attachment got somehow accidentally deleted!

Awesome. I am going to start on the next build of WWGD after my current run at dealing with custom notifications.

I played nearly 200 turns with this build with the English on Terra/Emperor.

The Aztec were pretty schizo: friendly from the start, offered DoF which I refused, declared out of nowhere, I took a city, they sued for peace, *immediately* back to friendly, another DoF offer following turn.

The Ottomans total Rexer/warmonger, score leader, seems to only be building units, eventually a dogpile of all but one civ on him, holding his own though.

Russia very standoffish, neither warring nor teching well.

Other civs (Inca, Babylon, Siam, India) all on one side if the continent, cuddly in a big love fest with me and each other. DoFs abound, will see if this continues...

(Writing on my phone = incomplete sentences!)

Sounds like the personalities are working out, then. Each leader's general attitude fits the personality type, but with one 'quirk' setting them apart from others in the category:

Like he was in Civ 4, Montezuma is the only character I left with a high 'deceptive' rating, so he probably hated your guts but pretended to be friendly, and attacks the moment he sees weakness. In vanilla almost all leaders have a moderate/high deceptive rating, which explains a lot about why they seem so random. The game was literally presenting false information to the player and we were assuming it was true.

Catherine is the one leader I gave a somewhat neutral attitude, neither particularly aggressive or peaceful, but favors rapid expansion. Alexander is the only 'ruthless' character who will protect citystates. Ramesses has an above-average Wonder flavor, and so on with the other leaders. Each quirk is designed to match their traits/uu/ub.

It makes sense the Ottomans were building tons of units, but the problem I've encountered is the one I described above: a warmongering AI with a big army, high war flavor, high boldness setting, low loyalty, and hatred towards a neighbor still won't necessarily attack that hated neighbor. All the variables seem to be in their favor so I'm not sure what holds them back. Sneaks might have more of an idea of what causes this behavior.

I just played a game with this mod, FR, WWGD and - to complicate things - the AI Equalizer mod for the first time as well.

I was the Inca on a continent with France, Germany, England and Russia. The response was exactly what I would expect on Immortal: Germany and France immediately targeted me due to my skimpy army, while England offered a DoF (which I took). Russia remained neutral for the entire game. Because I left Germany and France with two scattered cities each, there was no more warring after 200 AD or so.

Over on the other continent, Spain, America and the Mongols fought periodically, with Spain and the Mongols attacking CS, not exactly as intended. So far, it worked out basically as you planned. Do I like the feeling that "uh oh, Bismarck's next door" is much more reliable? Yes, at least for now. The game is more predictable (worse) but forces me to create strategy based on personality (better). Also, I felt it worked quite synergistically with WWGD.

There were no denunciations at all. Washington spoke to me of allying along with England, but he and I weren't Friends. I paid a trade price throughout for somehow breaking my word with the "armies on the border" query, although I moved from one border to another!

It's hard to comment on the CS alliance aspect, because this is where the AI Equalizer mod came in. I really enjoyed playing with it my first time. The AI handles the happiness hit without a hitch, mostly keeping its levels between -5 and +15. Naturally its pop was lower - in a game without a runaway civ, the highest city total was 14, highest pop 16.7M at 1800.

Where the difference shows is in gold. The AI has a lot less, so RA's are much less frequent with FR, and - in a game where those successful early wars left me with a larger early pop - the Inca's cost savings let me control all of the surviving Maritime and Cultural CS. I was challenged only once, doubled down, and moved on.

I would recommend trying a game with the AI Equalizer. It helped create the terrific sense of rational process that this game had.

The game was literally presenting false information to the player and we were assuming it was true.

Click to expand...

ARGH!
Dumbest design decision ever.

The game should *never* present you with false information.

Ahriman said:

ARGH!
Dumbest design decision ever.

The game should *never* present you with false information.

Click to expand...

I was laughing about this when Thal revealed it. Most people think the deceptiveness is a bug, but it's actually exactly what the devs said it was: frequently unreliable attitudes. Given the result, that was a dumb design decision indeed.

I don't know if the game should "never" present you with false information - take the case of Montezuma, for example. Duplicity could be part of the game, if properly integrated. The catch, of course, is that it's nowhere nearly well-integrated enough. The game would need to be exponentially better for the vanilla approach to ever make sense. A couple of liars in the apple basket might be okay, though.

I don't know if the game should "never" present you with false information - take the case of Montezuma, for example. Duplicity could be part of the game, if properly integrated.

Click to expand...

I'd go with never, actually.
Its one thing to have an aggressive leader who will attack you even when friendly.
Its another to have the game tell you that you have friendly relations when, in fact, you don't.

In Civ4 terms, its fine to have a leader who will attack you even at +6 relations. It wouldn't be fine to claim that you have +6 relations (+4 from religion, -2 from declaring war on friends, +1 for open borders, +1 for giving tribute, +2 from sharing techs) when in fact you are on -2 relations (because the AI doesn't actually share the same religion as you, and giving them tribute had no impact on them at all).

You should never feel like you're fighting the UI, or that the UI is giving you false data.

So, backstabbing leader? Sure.
Lying UI? No way.
I realize the gameplay effects are similar, but they aren't quite the same.

Ahriman said:

I'd go with never, actually.
Its one thing to have an aggressive leader who will attack you even when friendly.
Its another to have the game tell you that you have friendly relations when, in fact, you don't.

In Civ4 terms, its fine to have a leader who will attack you even at +6 relations. It wouldn't be fine to claim that you have +6 relations (+4 from religion, -2 from declaring war on friends, +1 for open borders, +1 for giving tribute, +2 from sharing techs) when in fact you are on -2 relations (because the AI doesn't actually share the same religion as you, and giving them tribute had no impact on them at all).

You should never feel like you're fighting the UI, or that the UI is giving you false data.

So, backstabbing leader? Sure.
Lying UI? No way.
I realize the gameplay effects are similar, but they aren't quite the same.

Click to expand...

I agree with your basic point in the first paragraph, but don't quite grasp the difference between the Civ 4 and 5 examples. In Civ 5 (no WWGD) the AI is frequently deceptive, which makes diplomatic efforts pretty meaningless. I doubt the game is purposely lying, rather than not worrying about consistency should circumstances favor them. That's why general deceptiveness doesn't make a lot of sense to me, as implemented in Civ 5.

But why would a Civ 4 leader declare war if he is legitimately +6? On some level, circumstances affect his attitude and cause him to act unpredictably. Is the difference between the two games that in Civ 4 these circumstances are much narrower, and to some degree understandable after the fact?

But why would a Civ 4 leader declare war if he is legitimately +6?

Click to expand...

Because they're a super-aggressive AI who will always attack if they sense weakness.

Its totally ok for Leader A to base their DoW decisions almost always on friendship, so they attack whoever they hate most, and never attack their friends, while Leader B bases their DoW decisions primarily on whoever they think the biggest victim is.
With a high diplomacy rating, B will still be happy to trade with you, and to join in with you on your wars, or denounce your mutual enemies, but will attack you if you seem weak.

Whereas a "duplicitous" leader that claims good relations but doesn't have them won't trade with you favorably or denounce mutual foes or get annoyed at people who attack you or join your wars or do any of the other things that good relations normally imply.

Thalassicus said:

Like he was in Civ 4, Montezuma is the only character I left with a high 'deceptive' rating, so he probably hated your guts but pretended to be friendly, and attacks the moment he sees weakness. In vanilla almost all leaders have a moderate/high deceptive rating, which explains a lot about why they seem so random. The game was literally presenting false information to the player and we were assuming it was true.

Click to expand...

Yup, I'm used to that, it's generally pretty easy to tell when there's a false friendly. What threw me off was that he proposed a DoF! Never seen that before when a "friendly" AI was plotting war; offering RAs, OBs, yes, but not that.

It makes sense the Ottomans were building tons of units, but the problem I've encountered is the one I described above: a warmongering AI with a big army, high war flavor, high boldness setting, low loyalty, and hatred towards a neighbor still won't necessarily attack that hated neighbor. All the variables seem to be in their favor so I'm not sure what holds them back. Sneaks might have more of an idea of what causes this behavior.

Click to expand...

The Ottomans are usually pretty easy to get along with, which is why I was so surprised when they were acting like the French.

What threw me off was that he proposed a DoF!

Click to expand...

Which is terrible. You should never have an AI offer you a DoF when they're intending to attack you soon, because they denounce before you do and then you get the "denounced by a friend" penalty.

The Ottomans seemed somehow deliberately designed to tend towards a peaceful early game and launch their big offensive with a huge army of Janissiaries.
But maybe I'm falsely assigning intelligent strategy to something that isn't really there in the AI parameters. [Or maybe they just built a big army of Janissiaries because they recognized them as a really good unit, due to some high flavor values, and then attacked in that era because they had a high power ranking military.]

Every leader but Monte now has their friendliness status match what's portrayed, and are relatively predictable (very unlikely to DoF+DoW, though I can't eliminate the probability entirely).

Monte is the one exception: he's interested only in strength. If we're stronger he'll want to be our friend... if he is stronger he will attack. He doesn't care about diplomacy or what was going on the last turn.

I do understand your point about the difference between:

• A leader is friendly
• Displays as friendly in UI
• Declares war

versus:

• A leader is hostile
• Displays as friendly in UI
• Declares war

Since Monte is the only remaining character who hides his true feelings, I think it works out. He's predictable in his unpredictability. Alternatively you can just think of him as crazy, which is how I looked at him in Civ 4.

Thalassicus said:

Every leader but Monte now has their friendliness status match what's portrayed, and are relatively predictable (very unlikely to DoF+DoW, though I can't eliminate the probability entirely).

Monte is the one exception: he's interested only in strength. If we're stronger he'll want to be our friend... if he is stronger he will attack. He doesn't care about diplomacy or what was going on the last turn.

I do understand your point about the difference between:

• A leader is friendly
• Displays as friendly in UI
• Declares war

versus:

• A leader is hostile
• Displays as friendly in UI
• Declares war

Since Monte is the only remaining character who hides his true feelings, I think it works out. He's predictable in his unpredictability. Alternatively you can just think of him as crazy, which is how I looked at him in Civ 4.

Click to expand...

I would add that even though you've revealed there was method to the diplomatic madness, there may still be something buggy about the "friendly" and "hostile" graphics. Often the specific message completely contradicts it.

One of the things I did in wwgd was reveal all the diplomatic opinions in the tooltip list for deceptive players rather than leave them hidden so the player at least has a heads up

Sneaks said:

One of the things I did in wwgd was reveal all the diplomatic opinions in the tooltip list for deceptive players rather than leave them hidden so the player at least has a heads up

Click to expand...

I didn't realize you'd done that, but I definitely make use of this by pretty much ignoring "friendly" or "hostile," and evaluating the diplomatic opinions.

@Txurce
I'm not really sure what we can do about their body language, to be honest... I also noticed it rarely matches their text-based friendliness level. Perhaps Sneaks knows more.


@Sneaks
I noticed that, and included it in the modified InfoTooltipInclude.lua for compatibility with autotips.