Complete.Org: Mailing Lists: Archives: freeciv-ai: November 2002:
[freeciv-ai] Re: freeciv agent interaction

[freeciv-ai] Re: freeciv agent interaction

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: a99bf1a7@xxxxxxxxxxxxxx
Cc: Freeciv AI development <freeciv-ai@xxxxxxxxxxx>
Subject: [freeciv-ai] Re: freeciv agent interaction
From: Gregory Berkolaiko <Gregory.Berkolaiko@xxxxxxxxxxxx>
Date: Tue, 5 Nov 2002 19:51:35 +0000 (GMT)

Hi David,

Yes, I think your engine might be very useful for the diplomacy.  Hwoever 
this part of AI is under development.  So there is no easy place to plug 
it in and I doubt there will ever be.  But we can work towards it.

The first step is to develop a protocol of interaction with your engine.  
You need to describe exactly what sort of info it will expect to get and 
what sort of feedback it will give.  You should direct all your emails to 
freeciv AI development list -- this way more people will interact and also 
the mails will get archived.

Best wishes,

On Sat, 2 Nov 2002, Nick  Cosgrave wrote:

> Hello, this is David Cosgrave, I emailed you Thursday in relation to an 
> interaction engine which is to be incorporated into FreeCiv.
> Thank you for your reply. The purpose of the engine I am coding is to deal 
> with AI agent opinions and beliefs and to see how an interaction would affect 
> an agent's beliefs and opinion of other agents, computer or human. So for 
> example, if the human player was the Germans and he demanded a tribute from 
> the AI English, my engine would decide how this action would affect the 
> English agent's opinion of the Germans - it would not decide on the action to 
> be taken based on this updated opinion. 
>     As it stands, my engine takes as input an interaction and the two agents 
> involved in the interaction. At the moment, an interaction is broken up into 
> different aspects, such as how friendly this interaction is, how helpful is 
> it, etc, as a score of 0-100. Based on these scores, and the agent's current 
> opinion of the other agent, the engine returns an updates opinion of the 
> other agent. The current opinion is necessary to access because if an agent 
> has a high opinion of another agent, then a friendly interaction won't really 
> affect his opinion of the other agent very much, whereas if an agent has a 
> low opinion of the other agent, then his opinion of him will increase 
> dramatically if the other agent does a good deed. Another aspect i am 
> considering is a value for the agent's self image - i.e., a value of what he 
> thinks the other agent's opinion of him is.
>     This is the engine i am coding. Whether or not this is compatible at all 
> with FreeCiv is not quite clear, perhaps it is not compatible. Of course, the 
> technicalities of the engine could be adapted to fit in with the data 
> structures that exist in FreeCiv, but the purpose of the engine - e.g., to 
> update beliefs and opinions - is fixed. This engine could be used for any 
> type of interaction - that was the initial purpose - however, i found that 
> most extensible games involve strictly murder (Quake etc), which would not 
> show the true nature of interaction, hence i would prefer to use a game such 
> as FreeCiv. 
>     Hopefully that makes things a little clearer. I really appreciate your 
> help and hopefully you could advise me as to whether or not FreeCiv would be 
> useful for my project. If not, perhaps you could recommend another 
> open-source or extensible game to use?
>         By the way, if you reply, the address to use is 
> a99bf1a7@xxxxxxxxxxxxxx
>         Thank you very much
>         David Cosgrave

[Prev in Thread] Current Thread [Next in Thread]
  • [freeciv-ai] Re: freeciv agent interaction, Gregory Berkolaiko <=