In this paper, me and Julia Kasmire from Delft University of Technology introduce an improved, simplified computer simulation model for studying technological evolution. The paper was presented by yours truly at the 26th European Conference on Modeling and Simulation in Koblenz, Germany in May 2012. I recommend the place and especially the wines there – jolly good all round.
The background for the model is in the realization that simulation models for technological evolution are somewhat limited. For example, the NK model made famous by Kauffman (1993) of Santa Fe Institute fame and used, often with modifications, by many researchers (e.g. Auerswald et al 2000; Frenken 2001, 2006; Murmann and Frenken 2006; Almirall and Casadesus-Masanell 2010, etc etc.) has provided very nice insights into complex technological problems. Nevertheless, it suffers from several problems.
The first problem is exogeneity. Exogeneity means that the search landscape is fixed at the beginning of the simulation; it does not change or evolve as a result of discoveries that the agents in the simulation may make. In philosophical terms, this is roughly equivalent to the Platonic idealism: that there exist somewhere an abstract form of perfect, ideal Thing-ness for each thing, waiting for us to find it. While interesting, this reduces the technological evolution to a search problem: how to search most efficiently through a complex landscape for the most ideal Thing that can possibly be found?
In addition, the NK model does not really model the way technologies are related to each other, or how some technologies depend from each other. If a lucky guess in the search process lands an agent on the perfect blueprint for perfect fusion reactor, it doesn’t matter whether the unobtainium for that reactor has been discovered or not.
An alternative, the percolation model, has been used by e.g. Silverberg and Verspagen (2005). In percolation model, technological development is modeled as (surprise surprise) percolation through 2-dimensional lattice. The idea is to find an uninterrupted route from the bottom to the top, when there are various obstacles strewn around. (It’s bit more complicated than that, but read the paper.) One can rather easily alter the parameters of search, and most importantly, this model does model the “keystone” technologies that are required for finding other technologies.
But the lack of internal structure presents a problem for both models. Technologies are hierarchies of components, just as e.g. Murmann and Frenken (2006) so nicely illustrate. But NK models do not model internal complexity at all – they do not really have components, only alternatives – and the percolation models stoop to this level only as far as the uninterrupted route can be found. Therein lies a problem: what if we want to model not just how technologies are found, but how they improve over time?
Thankfully, the brainiacs with the Santa Fe Institute – namely, the famed economist W. Brian Arthur and computer scientist Wolfgang Polak – have thought about this, too. Arthur and Polak gave the world a very intriguing model of technological evolution in 2004. In the model, they show how technology “bootstraps” and builds itself from simpler components, and how the components themselves improve over time. Their model was based on Boolean logic gates (NAND to be exact); random combinations of these gates eventually satisfied a host of needs speficically defined in the simulation (e.g. adder circuits, other logic gates, etc.).
Very nice, but try to code the model with the experience of a social scientist! Furthermore, extending the model seemed to be a bit difficult. So one night, lying on the carpet, I came up with an idea: for the computer, it’s all ones and zeros anyway when you get down to it, so why not just use ones?
The end result was the ADDER. The very simple idea in the model is that technologies are composed of fundamental building blocks or “primitives.” Different combinations of these primitives produce new products, which can then be used as components in further technologies. And so forth!
This model is easy to code, it’s fast, and it’s understandable. Furthermore, it’s extendable in a way Arthur and Polak’s model really isn’t. We have high hopes for it, and wish to use it in many, many experiments later on. For the details, go forth, to the paper already, shoo!
Korhonen JM and Kasmire J (2012) ADDER: A Proposal For An Improved Model For Studying Technological Evolution. In: Troitszch KG, Möhring M and Lotzmann U (eds) Proceedings of the 26th European Conference on Modelling and Simulation ECMS 2012. University of Koblenz-Landau, 108-114. DOI: 10.7148/2012-0108-0114
Arthur WB and Polak W (2004) The evolution of technology within a simple computer model. Complexity, 23-31. Available at:http://doi.wiley.com/10.1002/cplx.20130.
Almirall E and Casadesus-Masanell R (2010) Open versus closed innovation: A model of discovery and divergence. Academy of Management Review 35(1): 27-47.
Auerswald P, Kauffman SA, Lobo J and Shell K (2000) The production recipes approach to modeling technological innovation: An application to learning by doing. Journal of Economic Dynamics and Control 24(3): 389-450. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0165188998000918.
Frenken K (2001) Understanding Product Innovation using Complex Systems Theory. University of Amsterdam.
Frenken K (2006) Innovation, Evolution and Complexity Theory. Cheltenham and Northampton: Edward Elgar.
Kauffman SA (1993) The Origins of Order. Self-Organization and Selection in Evolution. New York and Oxford: Oxford University Press.
Murmann J and Frenken K (2006) Toward a systematic framework for research on dominant designs, technological innovations, and industrial change. Research Policy 35(7): 925-952. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0048733306000862.
Silverberg G and Verspagen B (2005) A percolation model of innovation in complex technology spaces. Journal of Economic Dynamics and Control 29(1-2): 225-244. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0165188904000132.