All Quiet on the Civ Front

[AmazonQueen]

In the context of a game like Civ, though, if the leader's personality isn't reduced to a trope, the leader ends up having no personality.

At some stage in it's history, pretty much every civ was successful militarily. Pretty much every civ was a leader in some aspect of science, generated a cultural influence on it's neighbours, etc. I think it's okay to pick a particular point in the civ's history and focus on one or two things that civ and its leader were particularly renowned for at that time, and build a leader personality around that.

Arent11 was talking about civ abilities, not leaders though.
And apart from Sparta the tropes were pretty arbitrary
Why diplomacy for Athens and not wonder building, science, naval warfare or trade all of which Greece was known for in that period
As for Persia for trade the Persians were about as much a nation of traders as the Romans were

 

[Arent11]

All of those civilisations were notable for military achievements, not just Sparta
Reducing civs to tropes isn't good design IMO

Well, yes of course it's tropes vs history. However, I strongly oppose the notion that tropes are "not good design". We are talking about a game here & the question is whether you or me will have fun.

 

[AmazonQueen]

Well, yes of course it's tropes vs history. However, I strongly oppose the notion that tropes are "not good design". We are talking about a game here & the question is whether you or me will have fun.

I'd rather go back to having blank civs that you can take in any direction like civ 2 or drop the pretence that it is at all based on historical models and just call them Militaria, Wonderland etc than oversimplify to that extent. Civ 6 has its problems. I agree that the AI Civs don't have enough variety in how they behave. However I think the more individualised civs of V and VI are an improvement over IV and that is one of the aspects of Civ VI I enjoy most, the different approaches to a science victory or a culture victory or (to a lesser extent) a domination victory they allow.

 

[kryat]

I often wonder how much personality actually existed in V compared to VI, or whether the personality was just the player projecting personality onto the AI. Conversely, how much personality in VI do we kiss because we aren’t looking for it.

There are some VI AI who seem to have definite personalities:
Gilgamesh: An easy friend who will always remain so
Gorgo: May want to conquer you if you are weak, but otherwise an easy ally
Catherine: Keeps to herself mostly
Tomerys: Slow to trust, but loyal if you avoid declaring war on her
Trajan: A forward-settling nightmare, who will definitely try to conquer you if he gets close

Mostly, the problem is the lack of actions in which they can express personality. Frequency of denouncements, war declarations, and forward settling are the prime indicators. Wonder frequency used to be a good indicator in V, but I’ve never observed anyone in VI to be as wonder-gluttonous like Ramses was in V.

If I could add more actions as ways to express personality, I would change:
-More neutrality towards city states in general (Fredrick’s agenda fails to stand out because everyone conquers city states)
-Add programmed city state liberators (Pericles, Tamar, Curtin, and Robert really should be doing this more often)
-Add something that makes certain AIs overpay for luxuries, and ask for trades more often.
-Increase the number of turns building up to an emergency to 3 turns, during which players can negotiate for deals and promises to participate, and make certain civs more likely to agree
-Add the “favors” system from Beyond Earth
-Add border patrols. Seeing armies on their borders shouldn’t be an indicator that they will definitely invade their neighbor. Realistically, they should be patrolling their borders, especially less trusting ones.
-Decrease hostility as culture dominance occurs.
-Increase declarations of their own accomplishments, not just remarking on you relative to their agenda. I want to know if they’ve been being environmentalists, or building walls, or nukes, or making alliances with everyone else.

 

[Arent11]

I'd rather go back to having blank civs that you can take in any direction like civ 2 or drop the pretence that it is at all based on historical models and just call them Militaria, Wonderland etc than oversimplify to that extent. Civ 6 has its problems. I agree that the AI Civs don't have enough variety in how they behave. However I think the more individualised civs of V and VI are an improvement over IV and that is one of the aspects of Civ VI I enjoy most, the different approaches to a science victory or a culture victory or (to a lesser extent) a domination victory they allow.

You would have to find a balance between player motivation:

(1) Players who enjoy history, want to learn something while they play
(2) RPGish players who want to create their own customized "avatar" civ/alter ego
(3) Players that don't care about traits and just want blank civs

Maybe you could do something like a few core civs that have traits according to history & then a custom civ option/expansion that has more "rpgish" traits. Then everyone should be happy.

 

[AmazonQueen]

You would have to find a balance between player motivation:

(1) Players who enjoy history, want to learn something while they play
(2) RPGish players who want to create their own customized "avatar" civ/alter ego
(3) Players that don't care about traits and just want blank civs

Maybe you could do something like a few core civs that have traits according to history & then a custom civ option/expansion that has more "rpgish" traits. Then everyone should be happy.

Yep, its certainly hard to keep all of us happy.
I think a custom civ option would be a good idea.

 

[Trav'ling Canuck]

There are some VI AI who seem to have definite personalities:
Gilgamesh: An easy friend who will always remain so
Gorgo: May want to conquer you if you are weak, but otherwise an easy ally
Catherine: Keeps to herself mostly
Tomerys: Slow to trust, but loyal if you avoid declaring war on her
Trajan: A forward-settling nightmare, who will definitely try to conquer you if he gets close

If any of those are true, it would be great, but I don't think they are.

I'd be happy to be corrected if/when the DLL code is released and people can dig in to see how it actually works, but I think that all AI leaders make decisions using the same decision tree. Other than a common set of factors used by all AI [relative military strength, etc.], I think the only other factor that comes into play is your current Relationship level. So your relationship with all Relationship +5 Leaders is going to be the same, impacted only by their internal factors. In other words, if you have the same Relationship score with the Red Civ and the Blue Civ, it doesn't matter whether Catherine or Gorgo is running the Red Civ, the Red Civ is going to act the way the Red Civ is going to act. If you swap the position of the Blue Civ for the Red Civ on the map, now the Blue Civ will act the way the Red Civ was going to act. Where Catherine versus Gorgo comes in is only whether you've satisfied or violated their agendas, which change your Relationship number.

Again, I'd be happy to be proved wrong.

​

If I could add more actions as ways to express personality, I would change:
-More neutrality towards city states in general (Fredrick’s agenda fails to stand out because everyone conquers city states)
-Add programmed city state liberators (Pericles, Tamar, Curtin, and Robert really should be doing this more often)
-Add something that makes certain AIs overpay for luxuries, and ask for trades more often.
-Increase the number of turns building up to an emergency to 3 turns, during which players can negotiate for deals and promises to participate, and make certain civs more likely to agree
-Add the “favors” system from Beyond Earth
-Add border patrols. Seeing armies on their borders shouldn’t be an indicator that they will definitely invade their neighbor. Realistically, they should be patrolling their borders, especially less trusting ones.
-Decrease hostility as culture dominance occurs.
-Increase declarations of their own accomplishments, not just remarking on you relative to their agenda. I want to know if they’ve been being environmentalists, or building walls, or nukes, or making alliances with everyone else.

I agree with this approach. I particularly like the idea of different approaches to City States, ideally consistent with their agendas and bonuses where relevant. I also think that cultural dominance should have an impact, and might go as far as to ban wars altogether at that stage.

 

[SammyKhalifa]

I think what we're seeing is fallout from the original "black box" approach to the Civ V personalities. Remember they originally didn't tell us why civs were doing what they were doing, and just had to guess from their actions. People decried that, but the mysterious missing information factor probably made players take more from the "individual personalities" than was really there.

 

[acluewithout]

If any of those are true, it would be great, but I don't think they are.

I'd be happy to be corrected if/when the DLL code is released and people can dig in to see how it actually works, but I think that all AI leaders make decisions using the same decision tree. Other than a common set of factors used by all AI [relative military strength, etc.], I think the only other factor that comes into play is your current Relationship level. So your relationship with all Relationship +5 Leaders is going to be the same, impacted only by their internal factors. In other words, if you have the same Relationship score with the Red Civ and the Blue Civ, it doesn't matter whether Catherine or Gorgo is running the Red Civ, the Red Civ is going to act the way the Red Civ is going to act. If you swap the position of the Blue Civ for the Red Civ on the map, now the Blue Civ will act the way the Red Civ was going to act. Where Catherine versus Gorgo comes in is only whether you've satisfied or violated their agendas, which change your Relationship number.

Again, I'd be happy to be proved wrong.

​

I agree with this approach. I particularly like the idea of different approaches to City States, ideally consistent with their agendas and bonuses where relevant. I also think that cultural dominance should have an impact, and might go as far as to ban wars altogether at that stage.

I may be wrong about this, but my understanding is they do use the same "tree" but their agendas etc. change the weighting. I think what happens at higher levels is initially your military is so low compared to theirs that the weighting just outweighs everything else.

 

[kryat]

I'd be happy to be corrected if/when the DLL code is released and people can dig in to see how it actually works, but I think that all AI leaders make decisions using the same decision tree. Other than a common set of factors used by all AI [relative military strength, etc.], I think the only other factor that comes into play is your current Relationship level. So your relationship with all Relationship +5 Leaders is going to be the same, impacted only by their internal factors. In other words, if you have the same Relationship score with the Red Civ and the Blue Civ, it doesn't matter whether Catherine or Gorgo is running the Red Civ, the Red Civ is going to act the way the Red Civ is going to act. If you swap the position of the Blue Civ for the Red Civ on the map, now the Blue Civ will act the way the Red Civ was going to act. Where Catherine versus Gorgo comes in is only whether you've satisfied or violated their agendas, which change your Relationship number.

I think you’re right that the relationship largely is the key to how they behave in regards to the decision tree, but I think you’re devaluing the relationship score and agendas with regards to the end behavior. Part of their agendas definitely directly affect their behavior to some extent (whether someone is building ships or holy sites, or declaring surprise wars, etc), which would affect the player swap hypothetical you describe.

The other aspect of agendas is how they perceive you satisfying their agenda, and incorporating that into the relationship score, which then affects behavior. It’s a little less direct than the first aspect, but I don’t think it’s correct to say that the AI leaders are perfectly interchangeable.

 

[Trav'ling Canuck]

I think you’re right that the relationship largely is the key to how they behave in regards to the decision tree, but I think you’re devaluing the relationship score and agendas with regards to the end behavior. Part of their agendas definitely directly affect their behavior to some extent (whether someone is building ships or holy sites, or declaring surprise wars, etc), which would affect the player swap hypothetical you describe.

The other aspect of agendas is how they perceive you satisfying their agenda, and incorporating that into the relationship score, which then affects behavior. It’s a little less direct than the first aspect, but I don’t think it’s correct to say that the AI leaders are perfectly interchangeable.

What you've described is, I expect, how the agenda system was envisioned. I agree their agendas impact their Relationship score with all other civs, and if you want to cultivate good relationships, you need to approach Harald differently than Teddy etc.

Whether, as @acluewithout asserts and you may also be suggesting with your first paragraph, it also affects their behaviour in other ways, that's the part I'm sceptical about, but would like to see implemented if it isn't currently a thing.

 

[PhilBowles]

If any of those are true, it would be great, but I don't think they are.

Really, I read that as 'different ways of describing the same character', and so very much in the eye of the beholder (very much not the case with Civ V - as I say, the threads on the forum describing individual characters were pretty in-depth and corresponded to most posters' impressions of them). Yes, in Civ VI if you make an ally they tend to be loyal - that's not a case of being Gilgamesh or Tomyris, it's just a case of backstabbing being rare in this game (actual backstabbing as in breaking a friendship, that is - I think people mean a range of different things by 'backstabbing' and the somewhat common 'he was green last turn and then declared war' is not actual backstabbing because no formal relationship is involved. One of the rare backstabs I recall involved Gilgamesh).

What difference is there between an easy ally, an easy friend, someone who "keeps to herself" (i.e. doesn't go to war with you either) or "loyal if you avoid declaring war on her"? Yes, sometimes it's slightly easier or harder to get an early friendship because of the agenda modifiers (and Gilgamesh is one of the easier ones), but those don't really qualify as behavioural differences.

"Definitely try to conquer you if he gets close" applies to everyone who starts fairly close, while declarations of war are rarer across the board once you're established - definitely not Trajan-specific. And I've observed no difference at all in AI settling behaviour - it's just dictated by location on the map.

And yes, as I noted before there are some differences in how civs play that simply follow from their civ bonuses - not a personality thing because it doesn't reflect any difference in coding. France, for instance, is simply a civ that consistently performs poorly in my experience, probably merely because it has no meaningful civ bonus for much of the game.

 

[kryat]

Whether, as @acluewithout asserts and you may also be suggesting with your first paragraph, it also affects their behaviour in other ways, that's the part I'm sceptical about, but would like to see implemented if it isn't currently a thing.

According to the wiki, only certain leaders have an agenda that would directly affect their behavior:
-Catherine likes to build spies
-Seondeok focuses on science output
-Amanitore maximizes district count per city
-Robert will abstain from war declarations except for broken promises (but presumably will be dragged into alliance wars)
-Shaka will maximize corps and armies
-Ghengis Khan will focus on cavalry
-Poundmaker tries to make as many alliances as his relationship score will allow (this may not actually have a discernible effect since the AI is bad at making friends with each other)
-Frederick conquers city states (is there anyone who doesn’t?)(I think Robert and Tamar might actually avoid CS conquering, but they don’t try to liberate that I’ve seen)
-Harold builds boats
-Tamar builds walls
-Cyrus declares surprise wars
-Ghandi doesn’t declare war
-Cleopatra tries to make alliances with militarily strong civs (unclear how)
-Jadwiga builds faith buildings
-Lautaro tries to keep high loyalty (unclear when this would be an issue that others wouldn’t also try to resolve)
-Victoria expands to every continent
-Montezuma collects luxuries (unclear how)
-Qi Shi Huang builds wonders
-Gorgo never gives anything for peace (I can confirm this one)
-Pedro tries to get great people
-Saladin builds worship buildings (I think that refers to the T3 buildings)

The other ~14 leaders have essentially no behavior that could be observed by the player in their set agenda. The hidden agendas could also affect direct actions though.

Regardless, these are all direct actions that we should be looking for if we’re tying to determine if there’s a difference in the behavior trees. Many of these are things I can at least anecdotally confirm. However, some of them only would be triggered in rare situations, or the resulting action is basically undiscriminably different than default behavior.

 

[Trav'ling Canuck]

According to the wiki, only certain leaders have an agenda that would directly affect their behavior:
-Catherine likes to build spies
-Seondeok focuses on science output
-Amanitore maximizes district count per city
-Robert will abstain from war declarations except for broken promises (but presumably will be dragged into alliance wars)
-Shaka will maximize corps and armies
-Ghengis Khan will focus on cavalry
-Poundmaker tries to make as many alliances as his relationship score will allow (this may not actually have a discernible effect since the AI is bad at making friends with each other)
-Frederick conquers city states (is there anyone who doesn’t?)(I think Robert and Tamar might actually avoid CS conquering, but they don’t try to liberate that I’ve seen)
-Harold builds boats
-Tamar builds walls
-Cyrus declares surprise wars
-Ghandi doesn’t declare war
-Cleopatra tries to make alliances with militarily strong civs (unclear how)
-Jadwiga builds faith buildings
-Lautaro tries to keep high loyalty (unclear when this would be an issue that others wouldn’t also try to resolve)
-Victoria expands to every continent
-Montezuma collects luxuries (unclear how)
-Qi Shi Huang builds wonders
-Gorgo never gives anything for peace (I can confirm this one)
-Pedro tries to get great people
-Saladin builds worship buildings (I think that refers to the T3 buildings)

The other ~14 leaders have essentially no behavior that could be observed by the player in their set agenda. The hidden agendas could also affect direct actions though.

Regardless, these are all direct actions that we should be looking for if we’re tying to determine if there’s a difference in the behavior trees. Many of these are things I can at least anecdotally confirm. However, some of them only would be triggered in rare situations, or the resulting action is basically undiscriminably different than default behavior.

Are these things someone has discovered in the code? Or "logical" behaviours that someone has extrapolated to the leaders? I don't recall Firaxis providing this information, but my memory's horrible and I may just be forgetting.

These may well be individual modifiers to the build priority decision tree and other game systems. Would be great if these exist and actually have a meaningful impact. Some of these I'm dubious about, Qin "come gaze at the wonders I never build" Huang in particular, but as you say, it would be difficult to distinguish some of these from normal behaviour, so they may well be there but not readily detectable.

As an example of that, in the 15 deity level test games I ran, the AI declared war on me 31 times, and every single one of those declarations, whether in the Ancient Era or the Information Era or any era in between, was a Surprise War. So distinguishing that Cyrus prefers Surprise Wars is a little tough in that environment.

 

[kryat]

Are these things someone has discovered in the code? Or "logical" behaviours that someone has extrapolated to the leaders?

These are all from Firaxis. Their technical result is unknown, but this what we’ve been told. I believe you can hover over agendas to see this or it should be in the civopedia.

They’re all there, but like I said, many of these are indistinguishable from normal behavior. This coupled with the fact that, as far as we know, agendas are the only path for behaviors that are different between leaders, and it's unsurprising that many of them seem to be interchangeable, even taking hidden agendas into account.

On top of this, the AI seems to be ignorant on how to use non-standard tools, which is why some leaders seem weak in AI hands, but strong in creative player hands (see @Victoria ’s excellent post on why Catherine is arguably the strongest leader). This includes:

-Not knowing how to effectively field or promote spies, based on your current victory path
-Not understanding how to get certain bonuses, like Greece or Tamar’s envoy bonus, or like the builder rushing abilities of the Aztec or China, or Spain’s trade route bonuses
-Not understanding civ-specific adjacency or resource bonuses before building cities

Firaxis could theoretically write AI that better used these tools, but my theory is that they are leaving these for the players to explore. That way, the player feels smarter than the AI. I’m not saying this is ideal, just that this seems to be the case.

 

[PhilBowles]

According to the wiki, only certain leaders have an agenda that would directly affect their behavior:

And we can easily test this hypothesis:

-Harold builds boats

Thinking back, I vaguely recall a game where this was true. I recall being impressed at the time - he had as many as two.

-Seondeok focuses on science output

Not a reliable gauge because Korea has a passive bonus that grants them extra science. In my games, if a civ with a more powerful general bonus like Australia is in play, the Australians are as or more likely to be ahead in science than Korea. Similarly Gilgamesh is usually a science leader. It's nothing to do with building priorities, it's just a consequence of the uniques.

-Ghengis Khan will focus on cavalry

Perhaps a better guide, but again not much. Every civ appears to have a bias towards building its own UUs (at least the civs with terrestrial UUs - I haven't observed the same for ships at all, which helps to explain Harald). Since Mongolia's UU is a cavalry unit it simply follows that they'll build those. Muddying the waters further, in general the AI is fond of producing cavalry if it has access to horses - and Mongolia apparently has a horse starting bias. The same can be observed with the Scythians. They just gave Genghis an agenda that works well with the way the AI's already coded, rather than coding Genghis in a way that promotes his agenda.

-Frederick conquers city states (is there anyone who doesn’t?)(I think Robert and Tamar might actually avoid CS conquering, but they don’t try to liberate that I’ve seen)

Indeed, there's no one who doesn't. Robert has been as active about it as anyone else in my games (I've rarely started near Tamar to judge).

-Tamar builds walls
-Cyrus declares surprise wars

The obvious problem here is that everyone does both.

-Ghandi doesn’t declare war

This is the only one I have registered (though I think he still eats city states).

-Jadwiga builds faith buildings

At least used to be the case for everyone. Any difference in behaviour should be more detectable now that not every civ spams holy sites, but building faith buildings is linked to planned victory condition, and any civ can apparently roll any victory condition in a given session with no clear bias towards one over another - I've seen Gorgo shoot for cultural or religious victory as often as she's aggressive, for instance.

-Victoria expands to every continent

I haven't observed any difference in expansion behaviour between civs. They're all better at settling other continents than they were in Civs IV or V, but still slow to do so.

-Qi Shi Huang builds wonders

I've had plenty of games where he moans at me endlessly because the one Wonder I've built is more than he has. Pretty sure this isn't the case.

-Gorgo never gives anything for peace (I can confirm this one)

I doubt this is generally the case, but can't recall any specifics of my deals.

-Pedro tries to get great people

I've noticed he often does, but once again this is probably no more than a consequence of the way his unique district works.

Are these things someone has discovered in the code? Or "logical" behaviours that someone has extrapolated to the leaders? I don't recall Firaxis providing this information, but my memory's horrible and I may just be forgetting.

I think they're just extrapolations from the leader's agenda (if they like X, they should do X themselves). Another example is that Roosevelt shouldn't declare war on his own continent - which is emphatically not the case.

I think we can rule basically all of them out, save as incidental consequences of leader/civ bonuses in a few cases. It seems fairly settled that the agendas just set what the AI likes/dislikes in other civs, they have no influence on their own behaviour.

 

[kryat]

I think we can rule basically all of them out, save as incidental consequences of leader/civ bonuses in a few cases. It seems fairly settled that the agendas just set what the AI likes/dislikes in other civs, they have no influence on their own behaviour.

I think this is far from ruled out. The anecdotes you provide don’t really demonstrate a difference in behavior, only that the behavior difference would be easily obscured in many instances (which is what I have been suggesting).

I would also suggest that when it says that a leader, “Likes” an action, that is more telling in the relationship score, and not of their own behavior tree. Your example of Teddy Roosevelt is bespoke: Teddy hates it when you declare war on his continent, but nothing precludes him from doing so (side note: his hidden agenda is always environmentalist; if we want to further test my suggested hypothesis on him, we could try to compare how often he removes terrain features compared to a standard leader.)

A more robust way to test these hypotheses would be to take actual statistics. Seondok may not be leading in science compared to Australia because of passive bonuses, but the real telling feature would be to compare the rate of campuses and campus buildings built by Seondok to Australia (science improvements/city), and if possible, also check policy cards, and whether research grants are being pursued. These are the only direct actions the AI can affect.

You’re absolutely right that passive traits (starting biases and uniques) would affect their overall performance, but that isn’t really indicative of their behavior. The thought experiment provided earlier about interchangeability is on the right track. If you could switch the AI for two leaders mid game, would you get different actions compared to if you had left them as is? Even if they got to keep the uniques of the leader they were switching into? This is hard to test, but more directly addresses the heart of the issue: Are the AI players differentiable?

 

[Trav'ling Canuck]

-Catherine likes to build spies
-Seondeok focuses on science output
-Amanitore maximizes district count per city
-Robert will abstain from war declarations except for broken promises (but presumably will be dragged into alliance wars)
-Shaka will maximize corps and armies
-Ghengis Khan will focus on cavalry
-Poundmaker tries to make as many alliances as his relationship score will allow (this may not actually have a discernible effect since the AI is bad at making friends with each other)
-Frederick conquers city states (is there anyone who doesn’t?)(I think Robert and Tamar might actually avoid CS conquering, but they don’t try to liberate that I’ve seen)
-Harold builds boats
-Tamar builds walls
-Cyrus declares surprise wars
-Ghandi doesn’t declare war
-Cleopatra tries to make alliances with militarily strong civs (unclear how)
-Jadwiga builds faith buildings
-Lautaro tries to keep high loyalty (unclear when this would be an issue that others wouldn’t also try to resolve)
-Victoria expands to every continent
-Montezuma collects luxuries (unclear how)
-Qi Shi Huang builds wonders
-Gorgo never gives anything for peace (I can confirm this one)
-Pedro tries to get great people
-Saladin builds worship buildings (I think that refers to the T3 buildings)

The easiest of these to test was Gorgo. As you state, she won't give up gold or luxuries no matter how badly she's losing the war. She will Cede cities, but maybe that's because she's been let in on the secret that doing so only harms the civ you cede the city to.

That makes it more likely that there's some coding to the other preferences as well. So that's positive.

 

[PhilBowles]

I think this is far from ruled out. The anecdotes you provide don’t really demonstrate a difference in behavior, only that the behavior difference would be easily obscured in many instances (which is what I have been suggesting).

It's certainly possible that there are intended personality differences but issues with the AI priority setting obscure them - e.g. if building non-campus buildings is set too high, it will override any lesser tendency Seondok has to build more campus buildings. Similarly no AI places a high priority on ships, so even the civs with naval uniques don't build them, Harald included. The fact that Gandhi doesn't tend to declare war suggests something is intended - other than meme value there's no obvious reason they'd have coded Gandhi, uniquely out of all the leaders, with a personality trait.

A more robust way to test these hypotheses would be to take actual statistics. Seondok may not be leading in science compared to Australia because of passive bonuses, but the real telling feature would be to compare the rate of campuses and campus buildings built by Seondok to Australia (science improvements/city), and if possible, also check policy cards, and whether research grants are being pursued. These are the only direct actions the AI can affect.

The game provides a way to approximate this: Great Scientist points. If Korea consistently gains more GS points it's evidence that they have more campuses and science buildings, the major way to produce them. From what I've seen there's little correlation between Korea's high science output and their GP performance - in my current game, before they were wiped out they were well behind Australia in GS points (though, as they were wiped out - apparently by the Aztecs - that could at least partly reflect that they had a poor start or had lost too many cities).

Then again, this isn't entirely conclusive. Seondok is claimed to like "high science", which can come from cards like Rationalism that don't influence GS points - although it's hard to get high science just from this. And lots of civs like taking that card.

You’re absolutely right that passive traits (starting biases and uniques) would affect their overall performance, but that isn’t really indicative of their behavior.

Precisely my point. There is still, so far as I can tell, no to almost no detectable difference in behaviour between any AIs - every apparent difference (other than Gandhi's war avoidance) cited, suggested by the agendas, or that I've observed can be explained wholly in terms of a preference for their own uniques, their passive stat bonuses, and their relationship modifiers resulting from agendas.

The thought experiment provided earlier about interchangeability is on the right track. If you could switch the AI for two leaders mid game, would you get different actions compared to if you had left them as is? Even if they got to keep the uniques of the leader they were switching into? This is hard to test, but more directly addresses the heart of the issue: Are the AI players differentiable?

Unfortunately a thought experiment can't tell you anything practical about the way the coding actually works, only the way it ought to work. The closest we can come in this case is comparing civs with different leaders, and the best example is India. This is partly because Gandhi has a detectable character trait, and partly because neither he nor Chandragupta has a particularly relevant passive stat bonus as a leader ability.

Chandragupta most definitely does not avoid war - he behaves identically to every non-Gandhi leader in the game so far as I can tell.

 

[Alexander's Hetaroi]

-Ghandi doesn’t declare war

This is the only one I have registered (though I think he still eats city states).

I definitely have seen him declare war before, granted it was a joint war early in the game. From the wiki I can only see that he will declare a war only when he won't be branded a warmonger.