AnalyticBridge
Data Intelligence, Business Analytics
Similarities between bird flocks and the economy
If you live in a place with a 100,000 small birds population, you've certainly see strong patterns in the way they flight, in organized clusters and sub-clusters of 500 or 5,000. They go up or down, turn left of right, or accelerate or brake, all at the same time in a perfectly synchronized ballet.
You might ask yourself how can they accomplish this. Surely, they don't have a shared, well known flight path, as they turn every 5 seconds. So there must be a leader, and each bird follows someone just above her in the hierarchy. Sometimes though, something goes wrong, and the synchronization is lost: birds in a very large compact flock seem to suddenly be flying erratically, bumping against each other, as if they've lost control. This type of flight is very destructive and can result in a massive collapse.
Is there any analogy with the way the economy operates?
You might ask yourself how can they accomplish this. Surely, they don't have a shared, well known flight path, as they turn every 5 seconds. So there must be a leader, and each bird follows someone just above her in the hierarchy. Sometimes though, something goes wrong, and the synchronization is lost: birds in a very large compact flock seem to suddenly be flying erratically, bumping against each other, as if they've lost control. This type of flight is very destructive and can result in a massive collapse.
Is there any analogy with the way the economy operates?
Views: 6
Replies to This Discussion
-
Permalink Reply by Theodore Omtzigt on -
I would postulate it to be different. Flocking behavior can be modeled by a straight forward n^2 algorithm where each individual checks against all other individuals and decides which flock they want to be in. To its simplest degree, a binary decision. Economic behavior is governed by more complex rules and two distinct additional constraints:
1- incomplete knowledge
2- large hysteresis
Incomplete knowledge creates the opportunity for 'bad' optimization behavior. Say a big company making a bet on the wrong macro-economic trend and destroying itself in the process.
Hysteresis creates instability, and large hysteresis creates catastrophic failure modes. For example, the credit crisis.
So, elements of flocking appear to be present in economic dynamics, but to me they feel like a non-causal description of the resulting dynamics that leaves too much uncertainty to be useful for predictive analytics. -
-
Permalink Reply by Theodore Omtzigt on
-
Ever since you posted this question, I have been finding popular article after article about traders and their flocking behavior. Turns out, there are a billion articles coming from psychology that capture the human behavior part of why 'flocking' makes us feel good. Here is a quick portal into that space:
http://www.spring.org.uk/2010/02/conformity-ten-timeless-influencer...
My contention still is that from a dynamics point of view flocking is too simple. One new thought I had is that there are multiple feedback systems at work. This is a refinement of the hypothesis of incomplete knowledge. If you think of the system to have many 'invisible' feedback systems that can drive group opinion, and these invisible feedback systems are driven outside of the decision makers then you have a description of a system where flocking is not the cause but just a symptom.
I got this idea when thinking about recommender systems. The concept of collective intelligence, and the philosophical ideas describing it (Emile Durkheim) gives us an example whereby you can predict likes and dislikes by association and without the need to understand the taste or specifics of the object for which you predict likability. That is the ultimate example I can think of of incomplete knowledge about the object.
So from a mathematical modeling point of view, the 'flocking' is described by correlation to social neighborhoods, but the system is the collective of many such correlations across many participants AND objects. The influence of each of these systems is likely to be a time varying weighting function.
From a usability point of view, such descriptions would need to be constrained to the decisions that can be affected. When trying to apply it to the market as a whole, the universe of decision makers appears to me to be much smaller than a typical population for a consumer product and thus seems entirely possible to capture. In business intelligence, the group of decision makers is even smaller, and there is tremendous value in aiding the decision process with at least some feedback when 'group think' is taking place. -
Top Content
Follow us
© 2013 AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC
