[Feature request]AI understanding the 'Prisoner's dilemma' metagame

See this link for wiki on what Prisoner's dilemma is about: http://en.wikipedia.org/wiki/Prisoner's_dilemma

Basically, I'd really like the AI to play as if it is 'wanting to win'. So, if one player is growing powerful beyond what the AI will be able to deal with that will lower its diplomatic standing to said opponent. It will then seek to form alliances with other, lesser players in order to be able to defeat the superior player.

Clearly, this fits the whole Diplomacy expansion pack concept well. It could be featured in the game in the way that just like the AI will like you more if you offer it gifts of ressources, it may dislike you simply because 'you have become too powerful' and it is afraid of you. This might cause it to take radical actions such as forming a temporary(?) alliance with even its former bitter enemies. It will try to do whatever it can to survive - just like a real human player would in a FFA game. Later on, it might then try to backstab its former allies.

5,199 views 5 replies
Reply #1 Top

Actually believe it or not there are sometimes when that does happen in the game. Im always the player that sits back and let my power grow in economic and army wise while the AI fights each other. And i discover that when i dont form any packs or alliences with any of the AI and start taking over just one of the AIs planet in thier galaxy, that AI will start ingoring the other AI's and form alliences with them. Which is what i love because it starts to make the game more interesting|-) .Im not sure if is still the case in diplomacy cause i havent played it that much but in enternchment it's exactly like that most of the time.

Reply #2 Top

Yeah, before entrenchment the AI tends to team up against the most powerful player. I'm not sure how it works in Diplomacy.

Reply #3 Top

I'll get on my "Allied Victory" soap box and say that the AI would be best to stay allied with you.

If the AI thinks "this guy is to powerful, I should attach him now to ensure my own victory," to me this implies that there is a limit to the number of people who can win, which is what I've been pushing for.  See my "Allied Victory" post.

Currently, every single player can win (not a single loser), so there is no logical reason for the AI to "think" about betraying you.  Afterall, you can both win.

 

This is a good idea, and would make most logical sense if there were a limit on the number of possible winners in a game.

Reply #4 Top

I think, with regard to allied victory, it should generally be every ally must be allied with all the others in the group (i.e. A is allied with B & C, B is allied with both A & C, C is allied with both A & B) thus making very large alliances tricky to gain group supremacy.

It may even come to a point where a 'core' of 4 or 5 factions are at the centre of a web of alliances from which they ultimatly hope to benefit as a group.

The problem is getting an AI to be able to track n-person metagames with n-iterations efficiently.