EXPLANATION
Explaining this
experiment meant facing a panic struck crowd of art-lovers who had anticipated
a nice Sunday afternoon stroll, but suddenly saw themselves diminished to
the role of a hard drive mounted on the back of an ant: .walking chunks of
peripatetic-computer code. Another example of (.walk) codephobia. Like always
this fear was really not necessary.
After being handed the code & a piece of a jigsaw puzzle, every agent determines
his/her sequence of turns (first street right, second left, third right) that
are to be repeated for the next 90 minutes. This is classic generative psychogeography,
the only difference lays in what's coded in the last half of the script. Here
the agents are instructed what to do when they meet. These actions are;
1) Write down the identification code of the other agent
2) Compare the pieces of the puzzle to see if they match.
2a) If they do then all pieces travel along with the agent carrying the lowest
code, both mark the code of the other to distinguish this encounter from the
non-productive ones, then go to 3.
2b) If they don't match go to 3.
3)Walk on: execute algorithm.
When after 90 minutes all pieces are in the possession of one agent the experiment
can be said to be a 100% success.
QUE PASA?
It's a thing
called emergence. Individual agents are not aware of what happens elsewhere
& all are performing the same simple routines. Over time however, these small
acts of localised behaviour, by means of the interactions, turn into a network
that is utilised by the pieces of the puzzle to find their match. This network
is a swarmbrain; a problem solver that finds the solution to a problem without
'thinking' but that creates a certain amount of 'intelligence' from the bottom-up.
Taking in account the odds that 2 random pieces of a 32 piece jigsaw puzzle
match, it's hardly strange the puzzle was not even near complete after the
given time. But it was time well spend as the agents were collecting some
hard-needed statistical data about all interactions during the walk. The number
of interactions is the backbone of the Universal Pedestrian Computer as is
controls the exchange of information within it, so how bigger our knowledge
is of its probability, the better walkware can be designed in the future.
This data was fed into Krackplot, a social network analysis program developed
by David Krackhardt, one of the pioneers in the field, using this we generated
this psychogeogram that maps all connections between agents(nodes). Please
note that the lay out of the nodes is only meant for clarity & doesn't in
it self signify anything.
Note that agent KK carried all the pieces that were not distributed, therefore
operating with an increased probability of finding a match, as indeed turned
out to be the case.
What the psychogeogram shows, is that the number of interactions was, taking
in account earlier experiments with much smaller groups, fairly high. This
indicates that the interactions go up exponentially with the number of agents
rising. Another bit of worthwhile information is that after 90 minutes the
swarmbrain has a 'Reachability' of 100%; meaning that every node is (indirectly)
connected to all others guaranteeing the steady flow of information. (Assuming
these statistics would be roughly the same after another 90 minutes).
In this particular experiment information is only changing hands between nodes
when there is a match. But suppose information being copied & exchanged even
if that particular bit of information can't be used at that moment. If it
was possible for agents to make a copy of their piece of the puzzle & exchanged
that during any encounter, the localised information would be distributed
through the network, instead of remaining in the hands of one person, the
chance of a match would then (over time) grow tremendously as every copy gets
copied again, etc.
You can analyse & manipulate the data yourself by downloading the krackplot
file (stroom.kp). For every node is given the algorithm
& if there was a match.
(june 2003)
Homepage Krackplot
http://www.contrib.andrew.cmu.edu/~krack/
Download Krackplot:
http://www.analytictech.com/downloadkp.htm
Walking Apart Together was made possible by Maureen Timmermans / Stroom Travel
http://www.stroom.nl
A more detailed report on this experiment can be found here (in dutch)
http://www.socialfiction.org/dotwalk/stroom.html