How to sew like Stuart Kauffman
November 28, 2005
In his book, At Home in the Universe, my colleague Stuart Kauffman describes a simple model of random network formation.
Imagine dumping a box of buttons (as in shirt buttons) onto your floor. Now pick up two buttons at random, tie them together with thread, and put them back down. You now have one connected component of two buttons and N-2 singleton buttons. Repeat. Over time, when you pick up a random button it will become more likely that it will lift up a small group of others with it. The interesting thing is, as the number of threads approaches one-half the number of buttons, a single, massive connected component will suddenly emerge such that when you pick up a random button, it is very likely to lift the vast majority of the other buttons. In other words, there is a critical phase transition as the system suddenly shifts from a collection of many, small, isolated groups to a single monolithic group, plus a few outliers.
To demonstrate this phenomenon, I just created a NetLogo simulation of Random Button Networks (Java applet), after following an introductory tutorial at the Complexity Workshop. Enjoy.
Random Button Network Simulation (Java applet)
Posted by Daniel in : Networks, Research, SimulationAdd a comment
When do networks not matter?
November 18, 2005
The question may have never occured to network researchers and enthusiasts. When you’ve found a paradigm that you love, it’s hard to see the boundaries of its utility. It’s the old “when you have a hammer, everything looks like a nail” story. But actually, the question which titles this post is an important networks question — not just a caution against overzealous methodologizing — because knowing when the network doesn’t matter means knowing when it does.
Network analysts use random networks as the standard by which to measure order in the networks they study. That’s because a random network is the graph-theoretic way of saying structure doesn’t matter. If the network structure you’re studying is significantly different from the random net, most likely it can’t be explained by chance alone; it has order, pattern, maybe even complexity. In other words, for the purposes of studying whatever system produced that structure, the network matters, i.e. it’s worth paying attention to.
And in the games of life and science, what matters most is knowing what is worthy of your thought and attention, and what is not.
Posted by Daniel in : Networks, Research, Science CultureAdd a comment
Searchability and the Evolution of Structure in Socially Constructed Networks
November 15, 2005
That’s the tentative, super-scholarly title for the paper I’m working on this week.
Broadly speaking, I’m looking at the question of “why is society structured the way it is?” Of the vast structural possibilities, we see only a few general types in the world. Is there some evolutionary pressure that selects some social structures as being more successful than others? If so, what are possible criteria for structural fitness? This paper puts forth the hypothesis that searchability is a plausible fitness test: that a successful structure is one that facilitates the finding of important or knowledgeable actors with minimal effort. This theory of mine, and its accompanying simulation study, shed light on the search and structure of human networks and of socially-constructed networks like the world wide web.
Posted by Daniel in : Networks, ResearchAdd a comment
Science: the quest for knowledge or tenure?
November 14, 2005
Sometimes the social reality of science raises its ugly head: that we scientists aren’t all simply altruistic seekers of knowledge, always giving credit where credit is due, but that we have self-interest and motivations of attaining prestige, career advancement and the like.
For this reason and more, I have high hopes for the work Marko Rodriguez is doing around the scholarly communication process:
The general purpose of the scholarly communication process is to support the creation and dissemination of ideas within the scientific community. [...] This paper describes an associative network composed of multiple scholarly artifacts that can be used as a medium for supporting the scholarly communication process.
- A Multi-Graph to Support the Scholarly Communication Process
My hope is that emerging scientific research tools such as these will help support a shift to a more meritocratic social system in science, where the incentives to research and publish are more closely aligned to the “true aims of science”. Specifically, I would hope that the future network structure of the scientific communication process will represent more a knowledge structure than a social structure.
A paper’s authors, acknowledgements and references are all relational data which point to people or artifacts which gave some contribution to the finished product. It’s for inconsistent, socially-constructed reasons that we separate and prioritize these resources. Like when decisions about who gets the first author slot are based on seniority instead of contribution.
Authors, references and acknowledgements are rightly distinguished as different types of resources but they all share the value of being a relational input to a paper. It seems to me that the network created by, for example, a single ranked list of each paper’s human and artifactual inputs, would more faithfully represent the living knowledge network instead of merely the social structure of the authors (coauthorship net) or the relatedness of topics (citation net). For instance, a reference may have been more valuable to the paper’s creation than one of the five co-authors, or an acknowledged party could have been more informative than many of the thirty references.
If the true aim of science is the quest for knowledge, not the quest for tenure, then it makes sense to give credit where credit is due, and accurately weight the importance of all information inputs to a given paper.
Posted by Daniel in : Networks, Research, Science CultureAdd a comment
