The concept of Path Dependence is used to capture the way in which small, historical contingent events can set off self-reinforcing mechanisms and processes that “lock-in” particular pathways of development.1 The result of this being that the set of decisions one faces currently are limited by the decisions one has made in the past, even though past circumstances may no longer be relevant. Even though previous choices were made on chance or limited information with better options now being available, it is still easier to simply continue upon a pre-existing sub-optimal path than to create an entirely new one. This is to say that the present is never a clean slate where we are free to make any decision. It is in fact contingent on how we got to this point. In a very broad sense, it means that history matters.2
When we look around us we can see that many of our systems of organization are a product of a path dependent process. Why do we still have the QWERTY keyboard that was designed for typewriters when it is not the most efficient for today’s keyboards? Why do we still used the standard gauge train track designed two centuries ago for horse drawn coal carts to run today’s powerful trains when it is far from optimal? Why is it so difficult for us to switch to renewable energy sources? Why do businesses all cluster in a particular area like Silicon Valley when there is nothing special about that particular location? All of these examples are because the choices we made in the past as to what technology we adopted influence the choices we make today.
Path dependency is particularly acute in complex systems because of their high degree of interconnectivity and more importantly interdependency. Things don’t happen in isolation. During the system’s development parts to the system interact with others and they co-evolve to become interdependent. Likewise, it is also particularly acute in complex systems because of their hierarchical structure. They are multi-tiered with different levels dependent upon each other, higher levels are built out of components on lower levels. When a new platform technology is adopted like Microsoft’s Windows operating system in the 90’s, over time many new technologies are built on top of this and become dependent upon it, a whole ecosystem of new application, new programming languages, new firmware, hardware, vendors, instructors, technicians and so on, meaning that small changes in the platform technology may result in a large effect across the ecosystem that has been specifically designed for it. And this is often the case for infrastructure systems, like transport networks and electrical power grids. They are deeply embedded within the socio-economic and technological fabric of a society with many deep dependencies.3
The basic theory to path dependency is that it is a product of a self-organizing process where some small initial event, that is, often somewhat arbitrary in nature, comes through positive feedback to create a lock-in effect. This lock-in effect leads to negative externalities, inertia and drives a particular course of events that are difficult to change in the future. So let’s analyze this process a bit further to try and understand it better.
Path dependency maintains that the starting point as well as feedback loops along the way affect and shape the end outcome to the technologies of today. In the language of chaos theory, this is called sensitivity to initial conditions, more popularly known as the butterfly effect. Because of feedback loops some small, possibly random event in the past can, in fact, turn out to have very significant consequences in the present or future, and that we cannot predict this process a priori. We have to run or simulate the running of the system in order to understand its future state. An example of this might be the initiation of the First World War through a relatively small event in Bosnia. There was no way of knowing that this small event would lead to a world war and the reshaping of Europe’s borders because this phenomenon really emerged out of the nonlinear interactions during the system development.
Next, positive feedback and negative externalities take hold to drive the system’s development. Economics of scale is a good illustration of this. The more users there are of a particular technology the more we can leverage economies of scale to reduce its price, which will in turn feedback to attract more users. This is a positive feedback and this is how some companies can get exponential growth as they ride this wave of positive feedback during the early state of a new technologies life cycle. Added to this, we have the network effect. The network effect is really due to the fact that the value of many technologies is in their capacity to interoperate with other users. Urban mass transit systems have the network effect. Every time we build a new station it adds value, not just to users of that particular station, but to the entire network, as everyone now has more possibilities in their destination.4
Both positive feedback and the network effect are powerful forces that once they take hold of, a particular technology will amplify it. But we can also add to this negative externalities, meaning that when someone chooses a particular option, that choice may be detrimental to other options. For example, once a particular industry or company adopts a standard, this will crowd out others, because the more a particular technology or standard grows, the greater the cost to other people if they choose not to use it. This makes it very difficult to change some technology or standard once it has taken hold, even if alternatives may be more efficient, and thus, this particular preexisting technology is essentially being subsidized by the network effect and the negative extremities of not being able to interoperate with others if you change.
This positive feedback and negative externality combine to create an attractor, meaning once they have taken hold around a system, they work to subsidize that state and make it an easier solution to any other of a number of different possible solutions, as it becomes the default. Because the other options are now more costly or difficult, this particular state now has an attractor built around it. In non-mathematical terms, an attractor is a set of states towards which a system will naturally gravitate from any given initial state and will remain within these set of states unless significantly perturbed. This is essentially the same thing as a default, where default means a value or state to a system that is automatically selected if no other option is specified. New entrants to the industry or new adopters of the technology without specific reason to do otherwise will adopt this default technology because of the attraction around it. This subsidizing of a particular option that comes with the network effect and negative externalities and the attractor space that it creates, results in inertia, the resistance to change.5
An example of this might be what is called Carbon Lock-In referring to the self-perpetuating inertia created by large fossil fuel-based energy systems that inhibit the adoption of alternative energy technologies. Now that we have built up sophisticated machinery for extracting and processing petroleum and the combustion engine has become a default technology, the industry is being subsidized by economies of scale and the network effect, meaning because of historical events we can produce a barrel of oil very cheaply. And if you have a barrel of oil, you can use it to do almost anything from making raincoats to greasing you car’s wheels, to trading it on the futures market. It is interoperable across a wide set of technologies giving it the network effect, an attractor and creating inertia.6
All of these, positive feedback, the network effect and negative externalities mean that once you decided to go down a particular path, it is self-reinforcing and excludes other possibilities in the future creating the inertia of the lock-in effect. Breaking out of this will require either a greatly more efficient technology coming along or a very effective organization for people to cooperate on changing to better available solutions. And cooperation is a very important aspect to this. It would take widespread cooperation for us to globally standardize the electrical plug or train track gages, and this is an example of how the sociopolitical domain influences the development of the technical domain.