I routinely point out in my outdoor naturalist explorations that one of the great innovations in evolutionary history is indirect communication - communication via the substrate - rather one-to-one communication. What this does is to make the information more permanent and less vulnerable to the death of individual organisms. It is the reason why you cannot destroy an ant colony by stomping on the workers walking on a trail.
You might see that to some extent this is what internet forums do, or what books do, they pass on information even after the death of the originator. But books are not location specific, I cannot find out who has walked at a specific spot, the way a dog or tiger might find out by sniffing a tree. Books are not sensitive to temporality - the dog or tiger might find out by the scents left on a tree how recent the last passer by was.
Social insects like ants and termites have evolved indirect communication to coordinate the activities of individual organisms without the need for centralized command and control. The terms stigmergy and stigmergic collaboration have been used for this and here is an explanation I found online (slightly edited):
Stigmergy is a word used to describe a particular type of control: the control of the actions of a group of agents via a shared environment. Crucially, the agents do not directly communicate amongst themselves. Instead, each agent is caused (by its environment) to act upon and change the environment. These changes in turn alter the later actions of the agents.
The word stigmergy comes from the Greek stigma, meaning sign/mark, and ergon, meaning work, capturing the notion that the environment is a stimulus that causes particular work (behavior) to occur. It was originally coined by zoologist Pierre-Paul Grassé,who explained the mound-building behavior of termites by appealing to the stigmergic control of the mound itself.
So if a termite mound is breached - the workers passing by might use a chemical marker saying - there is a breach here - as more and more workers pass the point, the chemical scent becomes stronger and it recruits workers who specialize in fixing breaches to the specific breach location. Workers might also mark trails towards the breach for others to follow. Once the breach is sealed, the trail scents and breach indicators fade away, leaving workers to follow their other activities. Notice that there is no central control and that chemical markers of different kinds may be produced by agents who may not know how to deal with the specific situation. Agents that do know how to deal with the situation are guided to a specific location.
Insect societies have task specialization - some workers specialize in foraging, some in nest care, some in defense and so on. Task specialization is sometimes based on the age of the insects, with older ones taking up risky activities.
I have tried to explain how this might or should guide construction of software such as the MediaWiki system - but evidently with little success - among some in the Wikipedia community. Agents need to be able to indicate centrally about areas of Wikipedia that are undergoing disturbance. Other agents need to be able to find, act at the areas of disturbance. Currently Wikipedia does this through central bulletin boards where agents explicitly post their notices. Unfortunately this is too taxing for a naive agent. WikiRage was a third party system that could detect increased editing activity and show articles that were currently "hot". There is no real-time system that shows currently highly visited articles. There is no system for currently highly sought after article - although this might be something for a search engine company like Google/Bing to think of. Now look at this also from the point of view of an agent with specialization - I as an editor might only act if I know that I can help, so overwhelming me with too many stimuli might only push an agent like me into confusion and inaction. If I were a specialist editor working in a particular cluster of articles, I should be able to filter out and indicate only if there is a rise in activity within articles in my cluster of interest. Ideally I shouldn't have to declare my own interest explicitly but article clusters should be determined from linkages or past editing history and so on. For a while now I have sought a rather simple means to detect traffic spikes in articles that I have on my watchlist. Now some software designers will immediately object that such as system could impinge on user privacy - although much of this information (other than mere reading) is already public in the MediaWiki system. I think many of these security concerns can be reduced by "aging" - the deletion of data over time - to simulate the dispersal of scents in social insects. Further such a system could perhaps be designed as a browser plugin, keeping data entirely local and off from the center. For instance if I wanted to look at what is hot on my watchlist - I could easily do with some kind of coloring and sorting of entries on my watchlist with a factor = yesterday's (or the last available) traffic / (average of the previous N days of traffic) [dealing of course with division by zero etc.] - that might help me narrow down my responsiveness to improve articles that I have an interest in. It would also make the system more responsive to user needs.
A super-organism - the term used for colonies of social insects - needs to have mechanisms for how its agents act as sensors, how those sensations are quantitatively expressed, how those quantitative expressions tip thresholds that drive actions or reaction.
Note: I have been bumbling with these ideas for a while and my knowledge of software development for implementing this particular idea has been rather limiting. I hope some talented software developer feels inspired to create something along these lines. I for one would be grateful for it!
PS: WikiRage went defunct and there is now a site called WikiShark which gives trending pages globally (for the English Wikipedia) but there is still a role as mentioned above for what is trending in what one can contribute to - ie based on task specialization.
No comments:
Post a Comment