http://www.gamasutra.com/blogs/JonShafer/20130218/186843/Revisiting_the_Design_of_Civ_5.php
QuoteMy original goal was for the AI leaders to act human. But humans are ambiguous, moody and sometimes just plain crazy. This can be interesting when you're dealing with actual, real humans, but I learned the important lesson that when you're simulating one with a computer there's no way to make this fun. Any attempt to do so just turns into random, unproductive noise.
I came to realize that while diplomacy is a unique challenge, it's ultimately still just a gameplay system just like any other. Regardless of whether your enjoyment is derived from roleplaying or simply a game's core mechanics, if your opponents' goals and behavior aren't clear then you'll have absolutely no idea what's going on or what to do.
In Civ 5, you might have been lifelong allies with a leader, but once you enter the late-game he has no qualms backstabbing you in order to win. With this being the case, what's the point of investing in relationships at all?
By no means should AI leaders be completely predictable. However, they do need a clear rhyme and reason behind their actions. The computer opponents in Civ 5 were completely enslaved to their gameplay situation, and as a result they appeared random and very little of their personalities shone through.
They were all crazy, and in the exact same way. In the months after the game was released I modified their behavior to be more predictable, but it was too late to completely change course. The biggest takeaway from this is that the only thing which matters in a game is the experience inside the player's head. It doesn't matter what your intentions are or what's going on under the hood if the end result just isn't fun.
much more at the link
Thanks for posting that quote. I just wished he realized that BEFORE Civ 5 was released. Hopefully Civ 6 will be different.
No way! Then people will think diplomacy is broken.
The EU-type diplomacy with transparent numerical ratings seems to be the way most games (incliding Civ) do it these days. For a game, players need clear feedback that their diplomatic actions are working. If not, there's not point in doing diplomacy in the first place.
That's the way Civ has always been, though. I don't see how he could have not seen the AI as anything but a slave to the game's core victory conditions, just like the player. You play to win. If that means your ally is in your way, you drop some nukes, send in the Giant Death Robots, pave over their territory, and turn it into a satellite parking lot for your shopping centers.
If the game's victory condition was to be bestest buddies and tra-la-la-la-la, you're talking about an entirely different game.
Quote from: Banzai_Cat on June 07, 2015, 08:07:42 PM
That's the way Civ has always been, though. I don't see how he could have not seen the AI as anything but a slave to the game's core victory conditions, just like the player. You play to win. If that means your ally is in your way, you drop some nukes, send in the Giant Death Robots, pave over their territory, and turn it into a satellite parking lot for your shopping centers.
If the game's victory condition was to be bestest buddies and tra-la-la-la-la, you're talking about an entirely different game.
Nah. There are other ways to win other than crushing everyone. The AI should be smart enough to win those other ways as well.
They do focus on non-military victory (especially in Beyond Earth).
They do, indeed. But if your Civ is about to win it all, I've seen plenty of AI Civs declare war, even if they were your bestest buds.
Quote from: Banzai_Cat on June 07, 2015, 08:57:46 PM
But if your Civ is about to win it all, I've seen plenty of AI Civs declare war, even if they were your bestest buds.
^QFT
Quote from: jomni on June 07, 2015, 07:59:56 PM
No way! Then people will think diplomacy is broken.
The EU-type diplomacy with transparent numerical ratings seems to be the way most games (incliding Civ) do it these days. For a game, players need clear feedback that their diplomatic actions are working. If not, there's not point in doing diplomacy in the first place.
Arcen Games seem to have solved this problem in The Last Federation. You can, at any time, see a detailed breakdown of races' attitudes toward each other; general disposition along with past actions that have influenced said disposition and by how much. For instance, I can select Acutians and Thoraxians, and see that the Acutians hate the Thoraxians because of recent attacks, for helping the Acutians' sworn enemies, for doing environmental harm etc. Each of these has weight and feeds into the general disposition. Or there might be a variety of recent actions, in addition, that the Acutians appreciate so much that the disposition is net positive.
I'm not sure what else you can do as a game designer to make it any more realistic.
Quote from: Banzai_Cat on June 07, 2015, 08:57:46 PM
They do, indeed. But if your Civ is about to win it all, I've seen plenty of AI Civs declare war, even if they were your bestest buds.
Yes, and I hate that. That's the "AI acting like a human" part he's talking about. If I wanted to play a human, well, that's what multiplayer is for. If there are AI's out there that don't like you and you are close to victory I can understand why they would attack. I would expect them to! But an ally shouldn't. In fact, if it's a strong enough ally it should jump in on your side.
if y'all recall the Civ 5-Africa AAR from the Mythical Front Page™ that's exactly what happened on the last turn of the game. The Portuguese turned on me and broke up the Southern Hemisphere Integration Treaty to attack me when it was obvious I was going to win it.
Quote from: bayonetbrant on June 07, 2015, 04:13:18 PM
http://www.gamasutra.com/blogs/JonShafer/20130218/186843/Revisiting_the_Design_of_Civ_5.php
QuoteMy original goal was for the AI leaders to act human. But humans are ambiguous, moody and sometimes just plain crazy. This can be interesting when you're dealing with actual, real humans, but I learned the important lesson that when you're simulating one with a computer there's no way to make this fun. Any attempt to do so just turns into random, unproductive noise.
I came to realize that while diplomacy is a unique challenge, it's ultimately still just a gameplay system just like any other. Regardless of whether your enjoyment is derived from roleplaying or simply a game's core mechanics, if your opponents' goals and behavior aren't clear then you'll have absolutely no idea what's going on or what to do.
In Civ 5, you might have been lifelong allies with a leader, but once you enter the late-game he has no qualms backstabbing you in order to win. With this being the case, what's the point of investing in relationships at all?
By no means should AI leaders be completely predictable. However, they do need a clear rhyme and reason behind their actions. The computer opponents in Civ 5 were completely enslaved to their gameplay situation, and as a result they appeared random and very little of their personalities shone through.
They were all crazy, and in the exact same way. In the months after the game was released I modified their behavior to be more predictable, but it was too late to completely change course. The biggest takeaway from this is that the only thing which matters in a game is the experience inside the player's head. It doesn't matter what your intentions are or what's going on under the hood if the end result just isn't fun.
much more at the link
When I got to the end of the snippet, I was somewhat shocked that a key guy on such a large and important project was reaching such a basic conclusion. Civ5 is essentially just a glorified board game and there are decades of design lessons to fall back on. The gaming industry really needs to diversify its hiring of design people if such basic conclusions are reached at the end of a hugely expensive project.
This goes along with my personal opinion that trying to design and implement an AI that acts, plays, and is challenging like a human being is a blind alley. They key is to make it seem like your opponent is human and everything between start and finish should be irrelevant from the players perspective. Some players might demand that the AI plays by the same rules as the players, but most people who say that have no idea what they are asking for. To me, its a question of fairness and illusion. You want the system to seem fair and as long as you maintain the illusion, how you got there is unimportant. A lot of bad AIs have been created and will continue to be created that ignores these principles.
Quote from: Huw the Poo on June 08, 2015, 01:52:15 AM
Quote from: jomni on June 07, 2015, 07:59:56 PM
No way! Then people will think diplomacy is broken.
The EU-type diplomacy with transparent numerical ratings seems to be the way most games (incliding Civ) do it these days. For a game, players need clear feedback that their diplomatic actions are working. If not, there's not point in doing diplomacy in the first place.
Arcen Games seem to have solved this problem in The Last Federation. You can, at any time, see a detailed breakdown of races' attitudes toward each other; general disposition along with past actions that have influenced said disposition and by how much. For instance, I can select Acutians and Thoraxians, and see that the Acutians hate the Thoraxians because of recent attacks, for helping the Acutians' sworn enemies, for doing environmental harm etc. Each of these has weight and feeds into the general disposition. Or there might be a variety of recent actions, in addition, that the Acutians appreciate so much that the disposition is net positive.
I'm not sure what else you can do as a game designer to make it any more realistic.
I don't think this is something pioneered by Arcen at all. This kind of an attitude break down was in distant worlds long before the last federation was released. In distant worlds, you can see exactly why a race likes or dislikes another race, as well as a numerical representation showing how significant each factor is.
Civ IV BTS with the Realism Invictus mod gives you an exact numerical breakdown of an opponent's opinion of you as well, though whether they're willing to fight you or not seems to depend a lot on you relative military strength.
Quote from: Jarhead0331 on June 08, 2015, 07:11:31 AM
Quote from: Huw the Poo on June 08, 2015, 01:52:15 AM
Quote from: jomni on June 07, 2015, 07:59:56 PM
No way! Then people will think diplomacy is broken.
The EU-type diplomacy with transparent numerical ratings seems to be the way most games (incliding Civ) do it these days. For a game, players need clear feedback that their diplomatic actions are working. If not, there's not point in doing diplomacy in the first place.
Arcen Games seem to have solved this problem in The Last Federation. You can, at any time, see a detailed breakdown of races' attitudes toward each other; general disposition along with past actions that have influenced said disposition and by how much. For instance, I can select Acutians and Thoraxians, and see that the Acutians hate the Thoraxians because of recent attacks, for helping the Acutians' sworn enemies, for doing environmental harm etc. Each of these has weight and feeds into the general disposition. Or there might be a variety of recent actions, in addition, that the Acutians appreciate so much that the disposition is net positive.
I'm not sure what else you can do as a game designer to make it any more realistic.
I don't think this is something pioneered by Arcen at all. This kind of an attitude break down was in distant worlds long before the last federation was released. In distant worlds, you can see exactly why a race likes or dislikes another race, as well as a numerical representation showing how significant each factor is.
This transparency is actually what makes it gamey. It's a game of numbers and all the player has to do is play the numbers trusting that the AI will follow it. These stats should be hidden from the player to add suspense and unpredictability as the article states. But as I said, then players won't bother working on diplomacy because they're not sure it's working anyway.
Quote from: Jarhead0331 on June 08, 2015, 07:11:31 AM
I don't think this is something pioneered by Arcen at all. This kind of an attitude break down was in distant worlds long before the last federation was released. In distant worlds, you can see exactly why a race likes or dislikes another race, as well as a numerical representation showing how significant each factor is.
It really doesn't matter. The point is, the problem is solved.
Quote from: Huw the Poo on June 08, 2015, 08:26:09 AM
Quote from: Jarhead0331 on June 08, 2015, 07:11:31 AM
I don't think this is something pioneered by Arcen at all. This kind of an attitude break down was in distant worlds long before the last federation was released. In distant worlds, you can see exactly why a race likes or dislikes another race, as well as a numerical representation showing how significant each factor is.
It really doesn't matter. The point is, the problem is solved.
Sorry. Not trying to be a nudge, but some people may still be interested in accuracy. O:-)
I think its important, anyway, because through these examples, there is always room for improvement and innovation.
Quote from: jomni on June 08, 2015, 07:45:40 AM
Quote from: Jarhead0331 on June 08, 2015, 07:11:31 AM
Quote from: Huw the Poo on June 08, 2015, 01:52:15 AM
Quote from: jomni on June 07, 2015, 07:59:56 PM
No way! Then people will think diplomacy is broken.
The EU-type diplomacy with transparent numerical ratings seems to be the way most games (incliding Civ) do it these days. For a game, players need clear feedback that their diplomatic actions are working. If not, there's not point in doing diplomacy in the first place.
Arcen Games seem to have solved this problem in The Last Federation. You can, at any time, see a detailed breakdown of races' attitudes toward each other; general disposition along with past actions that have influenced said disposition and by how much. For instance, I can select Acutians and Thoraxians, and see that the Acutians hate the Thoraxians because of recent attacks, for helping the Acutians' sworn enemies, for doing environmental harm etc. Each of these has weight and feeds into the general disposition. Or there might be a variety of recent actions, in addition, that the Acutians appreciate so much that the disposition is net positive.
I'm not sure what else you can do as a game designer to make it any more realistic.
I don't think this is something pioneered by Arcen at all. This kind of an attitude break down was in distant worlds long before the last federation was released. In distant worlds, you can see exactly why a race likes or dislikes another race, as well as a numerical representation showing how significant each factor is.
This transparency is actually what makes it gamey. It's a game of numbers and all the player has to do is play the numbers trusting that the AI will follow it. These stats should be hidden from the player to add suspense and unpredictability as the article states. But as I said, then players won't bother working on diplomacy because they're not sure it's working anyway.
I see your point, but I'm not entirely sure I agree 100%. Its important to know, for instance, that a player's dealings with a particular race is irritating relations with a third race. Its also important to know how significant of a factor this is to the third-party race. That way the player can make a decision to modify its dealings so as to either improve relations or risk worsening them further. In real diplomacy, nations would have this feedback.
Bring on Civ 6 im ready already.
Quote from: Jarhead0331 on June 08, 2015, 08:59:57 AM
Quote from: jomni on June 08, 2015, 07:45:40 AM
Quote from: Jarhead0331 on June 08, 2015, 07:11:31 AM
Quote from: Huw the Poo on June 08, 2015, 01:52:15 AM
Quote from: jomni on June 07, 2015, 07:59:56 PM
No way! Then people will think diplomacy is broken.
The EU-type diplomacy with transparent numerical ratings seems to be the way most games (incliding Civ) do it these days. For a game, players need clear feedback that their diplomatic actions are working. If not, there's not point in doing diplomacy in the first place.
Arcen Games seem to have solved this problem in The Last Federation. You can, at any time, see a detailed breakdown of races' attitudes toward each other; general disposition along with past actions that have influenced said disposition and by how much. For instance, I can select Acutians and Thoraxians, and see that the Acutians hate the Thoraxians because of recent attacks, for helping the Acutians' sworn enemies, for doing environmental harm etc. Each of these has weight and feeds into the general disposition. Or there might be a variety of recent actions, in addition, that the Acutians appreciate so much that the disposition is net positive.
I'm not sure what else you can do as a game designer to make it any more realistic.
I don't think this is something pioneered by Arcen at all. This kind of an attitude break down was in distant worlds long before the last federation was released. In distant worlds, you can see exactly why a race likes or dislikes another race, as well as a numerical representation showing how significant each factor is.
This transparency is actually what makes it gamey. It's a game of numbers and all the player has to do is play the numbers trusting that the AI will follow it. These stats should be hidden from the player to add suspense and unpredictability as the article states. But as I said, then players won't bother working on diplomacy because they're not sure it's working anyway.
I see your point, but I'm not entirely sure I agree 100%. Its important to know, for instance, that a player's dealings with a particular race is irritating relations with a third race. Its also important to know how significant of a factor this is to the third-party race. That way the player can make a decision to modify its dealings so as to either improve relations or risk worsening them further. In real diplomacy, nations would have this feedback.
Yup. The key is balance.
Quote from: Jarhead0331 on June 08, 2015, 08:59:57 AM
I see your point, but I'm not entirely sure I agree 100%. Its important to know, for instance, that a player's dealings with a particular race is irritating relations with a third race. Its also important to know how significant of a factor this is to the third-party race. That way the player can make a decision to modify its dealings so as to either improve relations or risk worsening them further. In real diplomacy, nations would have this feedback.
Precisely this. It's the lack of this kind of feedback that leads to players asking "why did that stupid empire do this when I gave them a pile of gold a few turns ago" etc. It's reassuring to know that an AI player is weighting its decisions based on actual actions from Human and AI players alike, and not just acting more or less randomly. It's also good to know that an AI player will still be pissed off with you for conquering part of its territory just because you gave it a palty sum of cash a turn later to avoid hostilities. I'm all for it.
I'm still waiting for the love child of Call of Power and EU, with some influences of Victoria II and (even though I've never played it) CKII
I wonder if this guys new game will have larger maps. The whole lack of stacking issue seemed that the designers didn't really understand the importance of combined arms warfare. I'll have to read up on it and see where he plans on going.
Quote from: bayonetbrant on June 07, 2015, 04:13:18 PM
http://www.gamasutra.com/blogs/JonShafer/20130218/186843/Revisiting_the_Design_of_Civ_5.php
much more at the link
It's odd how people who make games about the development of entire "civilizations" (whatever that fictional entity might be) seem to
have less than no idea how things like urbanization, long-distance trade, ideologies (for example with temples and gods or not), weapons
and militarization actually happened. Or rather the theories and evidence of how such things happened since in most cases the exact
mechanisms are not entirely clear. Anyway, apparently the game-builders theory is that whatever they think about markets (not particularly evident in early long-distance trade) and economic rationality (not necessarily what people were doing in the Bronze Age) is just the way it was and they have to sort of simulate this lameness in their own minds without recognizing it or analyzing it.
I would think that an attempt to emulate human behavior at the simplest level would require some preset personality AI packages. Starting with typical Expansionist/Defensive/Diplomatic/Peaceful/Etc basics that are weighted more towards one side of the behavioral "pie". Add in some specific racial or behavioral bias plug-ins when interacting with another specific racial type or gov't/diplo type for a little variety, and it should seem much less random.
It's been done plenty before but creating multiple AI types with extra bias plug-ins, fixing all the extra bugs that may cause, and tweaking them so that they're less random than some generic human-like AI probably takes a lot more time and testing to get right.
Quote from: Nefaro on June 08, 2015, 07:00:06 PM
I would think that an attempt to emulate human behavior at the simplest level would require some preset personality AI packages. Starting with typical Expansionist/Defensive/Diplomatic/Peaceful/Etc basics that are weighted more towards one side of the behavioral "pie". Add in some specific racial or behavioral bias plug-ins when interacting with another specific racial type or gov't/diplo type for a little variety, and it should seem much less random.
It's been done plenty before but creating multiple AI types with extra bias plug-ins, fixing all the extra bugs that may cause, and tweaking them so that they're less random than some generic human-like AI probably takes a lot more time and testing to get right.
It's odd, but the kind of information one gets about the early development of things like trade networks in say the late Bronze Age in the Mediterranean, is very local and precise. For example, lists of goods stored and work expected in Linear B. Could any existing game use the actual information that we have on such things? It appears not. You would have to start on a very local scale and very short time frames. Would this sort of game be interesting? Well, people are happy to play survival games and think how much happier they would be to survive as say a Bronze Age War Lord? It's true, most days you'd have plenty of wine, women and song, but what about your prestige? If you can't mobilize a few hundred men and a dozen ships you are going to fall off the trade network (which is mostly about prestige and redistribution, not accumulating profits) and they your followers will leave you and you will end up a sad old man in an poem no one remembers in a few decades. No fame after 1200 BC for the average slacker-warlord, I always say.
In many ways, that sort of prestige-based networking and mobilization for more prestige would be a lot easier for AI routines to manage than grinding out whole civilizations for thousands of years.