Per,
One of the best diplomacy models I've encountered was the one in Master of Orion. Each AI had a "relationship modifier" number, which ranged from -100 to +100. When you made an offer to the AI, it would add in the modifier before evaluating the offer. So you could make a lower offer to an AI with which you had good relations than an AI with which relations were poor. Some requests required that relations be above a certain threshold (>25 for non-aggression pact, >75 for alliance). Diplomatic actions would impact the relationship modifier either positively (trading, tribute, non-agression pact, hurting AIs hostile to this AI) or negatively (espionage, massing units on borders, killing the AI's units, being "too far ahead" in the game).
Each race had an agenda for expansion (expansionist, perfectionist), research (military, production, science, trade), and diplomacy (pacifisitic, xenophobic, erratic). They had a 60% chance of using their default agendas; otherwise, they would randomly use another. This meant you could usually (but not always!) count on a given race to act a certain way.
The in-game feel was remarkably human-like. Good AI relationships could be built up over time, betrayals were remembered, the AI seemed to have a genuine personality. I agree that the AI should make the game more fun; I think it can do that best by simulating as closely as possibly the way good human players play the game. I also agree it would be boring to have all the AIs aim for spaceship victories; the best games have a mix of warmonger and builder AIs.
One of the best diplomacy models I've encountered was the one in Master of Orion. Each AI had a "relationship modifier" number, which ranged from -100 to +100. When you made an offer to the AI, it would add in the modifier before evaluating the offer. So you could make a lower offer to an AI with which you had good relations than an AI with which relations were poor. Some requests required that relations be above a certain threshold (>25 for non-aggression pact, >75 for alliance). Diplomatic actions would impact the relationship modifier either positively (trading, tribute, non-agression pact, hurting AIs hostile to this AI) or negatively (espionage, massing units on borders, killing the AI's units, being "too far ahead" in the game).
Each race had an agenda for expansion (expansionist, perfectionist), research (military, production, science, trade), and diplomacy (pacifisitic, xenophobic, erratic). They had a 60% chance of using their default agendas; otherwise, they would randomly use another. This meant you could usually (but not always!) count on a given race to act a certain way.
The in-game feel was remarkably human-like. Good AI relationships could be built up over time, betrayals were remembered, the AI seemed to have a genuine personality. I agree that the AI should make the game more fun; I think it can do that best by simulating as closely as possibly the way good human players play the game. I also agree it would be boring to have all the AIs aim for spaceship victories; the best games have a mix of warmonger and builder AIs.
Comment