Stigmergy (Swarm Behavior) and Information Cascades
Stigmergy, also known as a universal coordination mechanism, is when a larger group coordinates through the indirect actions of its individuals. Stigmergy relies on signals left in the environment by an entity that signals other individuals to follow the same course of action. There are no instructions and each individual is not aware of the broader impact of their actions. Most of the evidence and also the applications of stigmergy involve swarm intelligence. The most widely known examples relate to insects that form social networks. Insects such as ants, wasps, bees, termites show evident signs of a universal coordination mechanism. We can take ants for example. Ants by themselves are not intelligent creatures, they are very disposable creatures by themselves. Yet in a group ants can do amazing things. They can build complex nests and trace paths in thick vegetation to food sources. For example, ants that return from a food source secrete a pheromone that signals to the other ants that there is food in the path the ant just came from. This source could be an isolated crumb or a large food source. The more ants that go to the food source and return, the stronger the pheromones, and thus stronger the signal. Ants without any form of central organization can easily weave a path to nearby food sources that can support the entire colony. Another example is how ants build complex three-dimensional nests without being intelligent themselves nor having a central authority. They modify the environment so that other ants, specifically worker ants, can act accordingly. However, this method is not a hundred percent foolproof. Distractions and other elements can confuse ants into secreting their hormones at the wrong time leading other ants to go off foraging in places without very much food or in places with danger. There is always a chance that the pheromone path may not lead to a food source.
This probability could be quantified using Bayes Theorem. It is also very similar to how information cascades work. After a group of people signal the same thing, a person receiving these signals would be more compelled to follow their path. For example, let’s say multiple people buy a product. They leave positive reviews on Amazon and tell their friends. A person noticing the fact that these people are content with their purchase would be more likely to buy that product. This can lead large groups of people following paths that end up at a certain product. The product has a large chance of being good but there is a smaller chance it might be bad. Again, Bayes Theorem can calculate the probabilities.
But there are also several differences between stigmergy and information cascades. In stigmergy, the agent is modifying the environment, not explicitly giving a signal to another agent. In information cascades an agent can get information directly from another agent. Stigmergy is more passive while information cascades are more active. Individuals in an information cascade are self-aware and can mislead the group if they choose to. Stigmergy has lots of applications for the future especially in the field of robotics and the networks these robots create. Stigmergy has potential in drone swarms and artificial intelligence. While stigmergy is a relatively new concept, it has a strong future.
https://www.cio.com/article/3193073/why-is-stigmergy-a-good-platform-for-swarm-intelligence.html
https://www.pnas.org/content/113/5/1303
http://pespmc1.vub.ac.be/Papers/Stigmergy-varieties.pdf