wu :: forums (http://www.ocf.berkeley.edu/~wwu/cgi-bin/yabb/YaBB.cgi)
riddles >> general problem-solving / chatting / whatever >> human-based genetic algorithms
(Message started by: amichail on Oct 10th, 2006, 3:50am)

Title: human-based genetic algorithms
Post by amichail on Oct 10th, 2006, 3:50am
It seems to me that human-based genetic algorithms are really the future of social software.

See for example my proposal for a collaborative game invention system in such an evolutionary playground:

http://targetyournews.com/?cmd=surf&urlid=656165

Could someone tell me why this sort of thing has not caught on in a big way yet? In fact, there seems to be little research on the topic.

Title: Re: human-based genetic algorithms
Post by Grimbal on Oct 10th, 2006, 4:17am
This reminds me of a game I read about a long time ago. I think it was called Democracy. It is a game with simple rules: you throw dice and get points, the one who reaches a given target first wins. But the idea of the game is that rules can be amended. Everybody can propose a new rule (for instance doubles give extra points, or the target can not be exceeded). When a proposal is made everybody votes and the rules applies if the majority agrees. Of course, you also can propose amendments on the rules about how rules get voted (for instance that the one with the most points cannot vote).

I never tried, though.

Title: Re: human-based genetic algorithms
Post by towr on Oct 10th, 2006, 4:23am
Outside of games, it seems somewhat like open source devellopment, where everyone can contribute to and change the code of a program. So I would say the approach has caught on.

I'm not sure how well it would apply to develloping game rules though. For applications the criteria are much simpler. The concept of a "fun game" is much fuzzier than for a "good program".
I'll have to give it more thought.

Title: Re: human-based genetic algorithms
Post by amichail on Oct 10th, 2006, 4:26am

on 10/10/06 at 04:23:51, towr wrote:
Outside of games, it seems somewhat like open source devellopment, where everyone can contribute to and change the code of a program. So I would say the approach has caught on.

I'm not sure how well it would apply to develloping game rules though. For applications the criteria are much simpler. The concept of a "fun game" is much fuzzier than for a "good program".
I'll have to give it more thought.

Unlike open source development, users of this system don't need to be programmers.

As for determining what is fun, note that playing the games and evaluating them is part of the process.

Title: Re: human-based genetic algorithms
Post by towr on Oct 10th, 2006, 4:40am

on 10/10/06 at 04:17:11, Grimbal wrote:
But the idea of the game is that rules can be amended.
There's a number of games like that. And some keep the rules hidden. For example there's the card game Mao. You have to learn the rules in play, but whenever someone wins, he can add an additional rule (without sharing what it is, in fact talking is generally forbidden, unless mandatory due to a rule).
Quite hilarious, especially when rules start to conflict, and you're reduced to picking the lesser of two evils.

Title: Re: human-based genetic algorithms
Post by towr on Oct 10th, 2006, 4:42am

on 10/10/06 at 04:26:45, amichail wrote:
As for determining what is fun, note that playing the games and evaluating them is part of the process.
You need a lot more contributors, and explore a much larger phase space though.

And of course, most kids these days don't want to play games that don't have stunning graphics. And least not  much.

Title: Re: human-based genetic algorithms
Post by amichail on Oct 10th, 2006, 4:57am

on 10/10/06 at 04:42:31, towr wrote:
You need a lot more contributors, and explore a much larger phase space though.

And of course, most kids these days don't want to play games that don't have stunning graphics. And least not  much.

I think this is different from wikipedia say. Wikipedia with only a few users who generate little content would not be interesting at all.

But if this system only has a few users initially who invent/evolve games, then that might still be ok provided that those games are sufficiently interesting to attract lots of players.

Title: Re: human-based genetic algorithms
Post by towr on Oct 10th, 2006, 5:39am

on 10/10/06 at 04:57:09, amichail wrote:
But if this system only has a few users initially who invent/evolve games, then that might still be ok provided that those games are sufficiently interesting to attract lots of players.
Seems to me you lose any advantage from an evolutionary approach if you have too few people. Because that's already how games are designed, a couple of people working at it together.
Without a proper amount of variety and competition you won't get speciation. You'll be tweaking a game, rather than creating a multitude of variants that may end up completely different from each other.

Title: Re: human-based genetic algorithms
Post by Icarus on Oct 10th, 2006, 3:31pm

on 10/10/06 at 04:17:11, Grimbal wrote:
This reminds me of a game I read about a long time ago. I think it was called Democracy.


I own a game called "Democrazy" which might be the one you are thinking of, though the description differs significantly. It works by a set of cards containing laws that are voted on. The winner is determined by whatever rules are in force at the end of play.

Title: Re: human-based genetic algorithms
Post by Grimbal on Oct 11th, 2006, 7:28am
It seems more than one game was designed on the voting principle, and "Democracy" or a variation is probably the most obvious name for it.

In my "Democracy" the rules were made freely, not drawn from a deck, so it is probably more in line with amichail's idea of collaborative game invention.  And what is still missing is the evolutionary part, the set of rules at the end should spread and be taken as a start for new games by other people, who could either play it through and make it evolve further or discard it as "not fun".

Title: Re: human-based genetic algorithms
Post by rmsgrey on Oct 11th, 2006, 12:21pm

on 10/10/06 at 04:17:11, Grimbal wrote:
This reminds me of a game I read about a long time ago. I think it was called Democracy. It is a game with simple rules: you throw dice and get points, the one who reaches a given target first wins. But the idea of the game is that rules can be amended. Everybody can propose a new rule (for instance doubles give extra points, or the target can not be exceeded). When a proposal is made everybody votes and the rules applies if the majority agrees. Of course, you also can propose amendments on the rules about how rules get voted (for instance that the one with the most points cannot vote).

I never tried, though.

Nomic (http://en.wikipedia.org/wiki/Nomic)

Mao is fun, but I've failed in my attempts to introduce it locally.

Title: Re: human-based genetic algorithms
Post by Grimbal on Oct 11th, 2006, 2:44pm

on 10/11/06 at 12:21:20, rmsgrey wrote:
Nomic (http://en.wikipedia.org/wiki/Nomic)

It is probably from the Scientific American article that I read about it.  Maybe it wasn't called democracy after all...

Title: Re: human-based genetic algorithms
Post by rmsgrey on Oct 18th, 2006, 4:15am
As an update, I've recently re-introduced Mao with what appeared to be more success (time will tell)



Powered by YaBB 1 Gold - SP 1.4!
Forum software copyright © 2000-2004 Yet another Bulletin Board