The Psychogeographical Trade-off
I've been thinking about the main purpose for which AlgoMantra was setup in 2004 on a street in Bandra while talking to Matti. The initial ideas were indeed inspired by Wilfried Hou Je Bek, but I get the feeling that both of us are missing some crucial elements to really make it work for a large crowd. I've zeroed down to a few basic properties which must be inherent in the system to fulfill my own vision of the thing:
1. Each person or agent who is doing the walk must have a simple instruction card, which should be fun to play with, easy to mutate. The agent must have a personal goal to achieve, a winning prize in mind. This would mean that his experience of being a part of the psychogeographical computer must be game-like, until the day this game becomes too commercial for its own good. Most multi-agent modelling systems or swarm simulation experiments are useless for our purposes because none of them take into account the fact that the bits that do the running around for them, do not seem be having any fun themselves. Humans are better at making aesthetic judgements rather than executing mechanical tasks, so is this game asking you to make aesthetic judgements in a mechanical manner? That seems to be a good conceptual start to the problem.
2. The various agents and their instruction sets make up distributed and parallel parts of a larger mechanism, a computation. The intent of this computation cannot be trivial, and must be very clear to the designer. This level can be as complicated as the designer's capacity for detailed abstraction and depth. It cannot be as trivial as crowdsourcing, or the Amazonian Mechanical Turk program. Also, you shouldn't be able to simulate the walk on a piece of paper - the interactions between agents effect the course of each player. so you can't just program it on a computer or execute it on a paper napkin to advance-guess the resulting path.
3. Algorithmic walking in left-right binary algorithms assumes its elegance based on a Cartesian grid, which is mostly absent from the city of my concern - Bombay. What you have is a random 2-D graph that has to be traversed. Are we going to have each walk tailor-made for a section of the city? Is this simply some kind of urban neo-tourism? are we using cellphones which are using Twitter? I'm flooded with these questions. It's going to rain soon in the city. What happens to the game during rain?
4. Another key aspect will be the card/pass, the real navigational instrument of our autonauts. It cannot be a fixed thing, it must have some degrees of change. Each move made by the player changes the state of the instrument, so that - the available choices in the next move depend on the last. Now we're getting somewhere to generative modeling of the path - so that the final path taken can never be predicted. We're probably looming close to a maze-gen algorithm.
5. A stray thought - could the data on the card/pass be represented by phonemes? This would truly be algomantric! Ringtones? Cellphones? Compose mode? WHAT?!!
Some random browsing about spy fiction landed me at 
Hm, I'm just going to coredump some of the useful webpages I saw today from the cybercafe. Notes to myself, as usual: