Darwinbots Forum

General => Biology => Topic started by: shvarz on January 15, 2007, 05:03:55 PM

Title: Five Rules for the Evolutionof Cooperation
Post by: shvarz on January 15, 2007, 05:03:55 PM

uploaded for your enjoyment: http://www.darwinbots.com/shvarz/1560.pdf (http://www.darwinbots.com/shvarz/1560.pdf)

Title: Five Rules for the Evolutionof Cooperation
Post by: Jez on January 15, 2007, 10:14:38 PM

Oh wow! (/me kicks all preceeding judgements of altruism out the car window)

Thanks for posting that, it was an interesting read.

Now I just need to consider how that applies to DB and press for changes!!  

Title: Five Rules for the Evolutionof Cooperation
Post by: Endy on January 19, 2007, 01:47:47 AM

Great paper thanks for posting.

Has everything I've been seeing in DB there(and some I haven't yet).

Behavior Based Altruism and Altruistic Networks are what we're still missing.

I like how he allows multiple mechanisms to explain altruism, which seems a more realistic explaination to me.

How mcuh did that subscription cost by the way? I've looked at subscribing to other journals but the price is insane.

Title: Five Rules for the Evolutionof Cooperation
Post by: shvarz on January 19, 2007, 12:02:55 PM

I have no idea how much it is - we get an institution-wide subscription.  I know there are discounts for students though...

The system is actually quite ridiculous. The wide majority of the research that is published in these journals have been done on tax money collected from US population.  And the population that paid for the research has to pay again to find out what they paid for...

The system is changing, with new some journals (like PLoS) providing free access to all published papers, but it's changing quite slowly.

Title: Five Rules for the Evolutionof Cooperation
Post by: EricL on January 19, 2007, 04:08:35 PM

Quote from: Endy

Great paper thanks for posting.

Has everything I've been seeing in DB there(and some I haven't yet).

I would be interested in understanding exactly what you have seen in DB that demonstrates altruism.  I for one, would be very hesitant to label the fact that hand coded bots programmed by their human authors not to prey on each other actually follow their programming as 'altruism' or that mutations to that programming which break this behaviour as 'canabalism'.

It's a fine paper, covering very sophisticated and complicated behaviour in biological organisms.  But as we have yet to evolve even the simpliest of behaviour much less the sophisticated mechanisms necessary for such things as kin selection or reciprocal altruism (like sex, species recognition, etc.) I fail to see what in DB comes anywhere close to demonstrating altruism.

Title: Five Rules for the Evolutionof Cooperation
Post by: Numsgil on January 20, 2007, 12:34:01 AM

I dunno, not losing altruistic behavior is in some ways similar to developing it from scratch.  Evolution likes to break things, and I'd say having a behavior not break would be almost as important as seeing it evolve from scratch.

Title: Five Rules for the Evolutionof Cooperation
Post by: Endy on January 21, 2007, 01:47:55 AM

Haven't seeen it truly self evolve, is easy however to design it into the sim.

The main method I've found is with epigenetics for identifying relativly close family. It makes sense that the standard "be nice to all member of my own species" doesn't work well over a long run. Eventually the bots are only competing against their own distant/close relatives, giving cannibots a decisive advantage. If the bots are altruistic towards only closely related bots the cannies loose their advantage.