History simulator

Started by
19 comments, last by StrategicAlliance 19 years, 6 months ago
Quote:Original post by vanevery0
You are giving Black & White way too much credit for actually doing anything. It provides a nice interface for configuring canned beast behaviors, that is all.

I should have specified that I was referring to the behaviors of the people rather than the creature. They do their own little thing of building their town, etc.

Dave Mark - President and Lead Designer of Intrinsic Algorithm LLC
Professional consultant on game AI, mathematical modeling, simulation modeling
Co-founder and 10 year advisor of the GDC AI Summit
Author of the book, Behavioral Mathematics for Game AI
Blogs I write:
IA News - What's happening at IA | IA on AI - AI news and notes | Post-Play'em - Observations on AI of games I play

"Reducing the world to mathematical equations!"

Advertisement
Quote:Original post by fup
Sugarscape has been used to model historical events (such as the migration of native American Indians). It might be worth a look.

http://www.brook.edu/es/dynamics/sugarscape/default.htm


Excellent! Rating++!
Quote:Original post by InnocuousFox
I should have specified that I was referring to the behaviors of the people rather than the creature. They do their own little thing of building their town, etc.


I see no intelligence here. How is it anything other than standard RTS fare? I mean, 'they're a simulation of sorts', but there's nothing special about them, no 'pinnacle' that B&W has achieved regarding population simulation that other games haven't done a zillion times.

Anyone remember SimEarth BTW? It's a sim, it's not particularly AI.
Cheers, Brandon J. Van Every(cruise (director (of SeaFunc) '(Seattle Functional Programmers)))
Quote:Original post by happylrac
Quote:Original post by fup
Sugarscape has been used to model historical events (such as the migration of native American Indians). It might be worth a look.

http://www.brook.edu/es/dynamics/sugarscape/default.htm


Excellent! Rating++!


I agree. It was very useful and helped me find other papers on Artificial Societies.
Many thanks to everybody
Quote:Original post by tonyg
Has anybody got any reference material for History simulators?
What I mean is create a virtual world (rivers, mountains etc) introduce 'humans', foodstuffs etc etc and rules for building civilizations, creating wars etc etc and leave it to run for 5000 years with each 'civilization' writing it's own history.
The idea is to 'uncover' the history of the planet using virtual archaeology and documents.
I've had a look at some A-Life sites but can't find anything similar.




The problem is the complexity.
Like most AI projects, the engine is less than 10% of the project and the behavior logic is quite involved.

You would have to define the patterns ahead of time, and what you have specified is not as simple as existing game rules/pseudo AI. Even if you try some kind of Genetic Algorythm to evolve civilization, its still based on precanned behavior primitives.

Try taking a sample A-life ruleset and project all the influences and factors that you think will make a rich enough simulation for your 'history' and see how quickly it grows huge.

Simulating 'tribes' might be reasonable but think of the dynamics of an empire (let alone a world spanning civilization as exists today).


Quote:Original post by vanevery0
Quote:Original post by InnocuousFox
This would require a massive agent-based model that has all the abilities of individual survival and awareness of/ability to take advantage of social connectivity. This is not small at all. The only things that have approached this are games like Black & White.


You are giving Black & White way too much credit for actually doing anything. It provides a nice interface for configuring canned beast behaviors, that is all.



All the events were scripted.
And the beast's adaptive/feedback behaviors were quite shallow (yet it was a big innovation in game behavior programming)
Quote:Original post by Anonymous Poster
Quote:Original post by tonyg
Has anybody got any reference material for History simulators?
What I mean is create a virtual world (rivers, mountains etc) introduce 'humans', foodstuffs etc etc and rules for building civilizations, creating wars etc etc and leave it to run for 5000 years with each 'civilization' writing it's own history.
The idea is to 'uncover' the history of the planet using virtual archaeology and documents.
I've had a look at some A-Life sites but can't find anything similar.




The problem is the complexity.
Like most AI projects, the engine is less than 10% of the project and the behavior logic is quite involved.

You would have to define the patterns ahead of time, and what you have specified is not as simple as existing game rules/pseudo AI. Even if you try some kind of Genetic Algorythm to evolve civilization, its still based on precanned behavior primitives.

Try taking a sample A-life ruleset and project all the influences and factors that you think will make a rich enough simulation for your 'history' and see how quickly it grows huge.

Simulating 'tribes' might be reasonable but think of the dynamics of an empire (let alone a world spanning civilization as exists today).




Oh and I forgot that the detail level you mentioned requires some task/planning mechanisms which can simplify your behaviors
by using multi-tiered task/solutions (which still have to be canned...)

BTW the hardest part of a plan/task/goal/solution system is the evaluation to pick the best available solution to the problem situation -- think of a fitness test that evaluates a combination vector of cost/risk/time/resourse/certainty recursively on a planning tree search. Note that this evaluation has to take into account partial completion/interruptions and constant reevaluation as an agents situation changes.


You should look into event calculus. It seems appropriate for what you are doing.
Thanks again everybody.
I'll look into Event Calculus (new term for me).
AP... I think you're right. Looking at it a bit more adding something like 'farming' might be easy when a 'tribe' has lived in a fertile area for an amount of time. The problem, as you point out, is how many of these 'events' would there be. The plough? The ploughshare? Horse pulled? etc etc. Hmmm. Might need to think about it a bit more. Maybe a simple asbtract of the tribe getting more efficient at farming over years without any specific advancements.
Back to the drawing board then.
A good place to start as far as thinking about history in a way that lends itself to computation, is with the writings of Georg Henrik von Wright. In particular his book "Explanation and Understanding". (See also: Georg Henrik von Wright (1916-2003)). Here's the relevant paragraph from this second source:

Quote:
In 1963 von Wright published three books. THE VARIETIES OF GOODNESS he considered his best and most personal work. Other studies were NORM AND ACTION and THE LOGIC OF PREFERENCE. During this period he began to take an interest in cybernetics and mathematical behavioral science. EXPLANATION AND UNDERSTANDING (1971) showed the influence of Wittgenstein's last writings. It is perhaps von Wright's best-known work. FREEDOM AND DETERMINATION (1980) was von Wright's last major book on logic. In it he continued to elaborate his ideas of the relation between actions and their reasons and also the differences between the human and the natural sciences. Von Wright also wrote in his later years of ethics, cultural philosophy, and ecological questions. His analytic hermeneutics is considered a bridge between two rival approaches to philosophy.


I read "Explanation and Understanding" in college as part of a course on the philosophy of history. I found the book quite useful. It's not a huge tome of ponderous thoughts either.
"I thought what I'd do was, I'd pretend I was one of those deaf-mutes." - the Laughing Man

This topic is closed to new replies.

Advertisement