Let me field that particular issue that Luckystrike raised. As I understand it, Civ5's diplomacy has the goal of modeling a more "human" system of interactions between players, by attempting to create an AI that is less trustworthy and more unpredictable. In this it is somewhat successful, at least in the sense that the AI players are indeed untrustworthy and prone to backstabbing. The argument that's advanced to defend this system usually goes something like this: "If you were playing a game with other people, and you were getting close to winning, then they would always backstab you to stop you from winning. There's nothing wrong with programming an AI to play the same way."
But there's a problem with this: actual humans *DON'T* act like this, not all the time anyway. Think back on your experience playing games with other people, whether it's board games or online games or whatever. Some people are hyper-competitive, and they will indeed be willing to betray anyone, backstab, whatever in order to win. Some people are prone to taking sides, loyal to their friends and determined to see their enemies lose. Some people are "just happy to be there", not really interested in winning so much as enjoying the game and having fun. There's a wide spectrum of different behavior types, because different people react to events in different ways. It's what you would expect because everyone has their own distinct personality, and we don't all see the the world the same way.
This is why Civ5's diplomacy breaks down so badly. It attempts to mimic human behavior, but it has nowhere near the complexity or nuance to pull it off. It's an
uncanny valley representation of human behavior, which comes off as creepy and robotic (because that's what it is after all, just a bunch of programming code!) Individual aspects of the AI design make sense - tension from settling nearby, tension from declaring war, etc. - but they combine together in completely nonsensical ways, like the AI who hates you as a warmonger for joining him in battle. By attempting to make the AI play like a human, the designers have created a profoundly inhuman system of behavior. Because no, in fact, humans do not ALWAYS backstab their friends to stop them from winning, and no, countries are not ALWAYS looking to betray their allies for their own self-advantage. You can write all the essays you want on realpolitk and game theory, but the real world doesn't work that way, and it's not an accurate model of human interactions. Civ5 would greatly benefit from a wider range of AI responses in the different leaders, like having some leaders who were extremely pacifistic and looking to make allies, other leaders who would be loyal until betrayed, and so on. But that would require a complete redesign of the system, because as currently constituted, there's literally no benefit to making allies in Civ5, which surely seems like a failure of design goals. All you do is sell stuff to AI civs for gold or research agreements, and that does not a friendship make.
Thanks again to Dan for having me on the show.
Embrace the wordplay. The one-hundred-and-twenty-ninth episode of PolyCast, "The Letters T and Eh?", features regular co-hosts Daniel "DanQ" Quick and Philip "TheMeInTeam" Bellew with returning guest co-host Kelly "KMadCandy" and first-time guest co-host Michael "Sullla" Soracoe; it has a runtime of 59m59s.
