Complexity

Mitchell Waldrop’s Complexity is an interesting book. Unfortunately, I haven’t had time to blog about it as much as I wanted, but here are some brief things I thought about when reading it.

Innovation

Waldrop writes about innovation in the context of the economy, which is something that Brian Arthur and the others at the Santa Fe Institute were thinking about. For instance, Arthur talks with Stuart Kauffman about viewing innovation within an economy as a network, where the nodes represent innovations – existing as well as possible future ones. A new innovation would then be “switched on”, sending a signal across the network and activating other new innovations along the way. This resonated with Kauffman’s work with DNA, which reconfigures itself by propagating signals, prior to a cell splitting in two, determining the “kind” of cell to “create”. (I’m not sure I have understood this correctly, I’m afraid.)

To me, this way of seeing it speaks in favor of keeping an open mind when creating something. Often – especially in software – there’s a tendency to want to nail things down at the beginning, as a result from a desire to achieve control and predictability. (If we know exactly what we should do, we will surely know how long it will take and what it will cost, right?) Within an economy, innovations always are the result of combining existing technologies or ideas. What’s to say that the technologies we create won’t lead to new innovations within the same project – things we never could have anticipated up front?

Economics and physics

Waldrop also writes about a meeting at the Institute between a group of physics and a group of economists. The idea was to see what they could learn from each other. During the meeting, Brian Arthur describes the similarities and differences between the disciplines. First, both deal with particles, or agents, that interact with each other. In an economy, however, the agents are smart; they have expectations and are capable of creating strategies – while in physics they are essentially dumb. This means that in physics, the agents (particles) are predictable, while in economics they are unpredictable. He also said that economists seldom paid much attention to the “spatial dimension” – which I interpreted as regarding the nature of the interactions between agents.

In this discussion, Nobel laureate Phil Anderson, as Brian Arthur talked about his ideas about increasing returns, associated economies with “spin glass”, which I understood is glass where the molecules are magnetic and spinning. The molecules try to adapt to its neighboring molecules, but since these try to adapt to other molecules that spin in different directions, every molecule constantly alter its spin. In other words, it can’t function as a regular magnet, for which the molecules align into a fixed configuration. So a piece of spin glass is a system where positive and negative feedback are constantly at work. Local equilibria occurs, but as a whole it’s constantly changing. This is how Brian Arthur sees economies, as I understand it, and also the reason why he reacted towards the common view of economists, that an economic system always returns to an equilibrium, and where there’s only diminishing returns.

Autocatalytic set

Waldrop writes also about Stanley Kauffman’s work on “autocatalytic sets”. At the beginning, he thought about how life began, and he found the idea way too unlikely that cells should have emerged from amino acids and proteins by random. He thought about isolated ponds containing such simple compounds, which could perform two basic operations: splitting, and joining. Natural selection wouldn’t occur within such a pond, unless it would “spill over” into a nearby pond. Then, only, the compounds would start competing for resources. (I don’t understand this fully, so please bear with me. I guess I’ll have to go back and read this again.)

Anyway, at a specific level complexity of the interactions between a set of amino acids – that is, the probability of individual amino acids either splitting or joining with other others – it would give rise to spontaneous order – that is, that it would start forming cells, which would form organisms, and so on. Later, he verified this using computer models.

As I read about this (and it seemed clearer to me at the time) I thought that this was somehow different from the theories presented in Linked. Here, the “nodes” are joining each other or splitting into two new nodes. As I understand it, Kauffman regarded these things as “abstract networks”, but did he see joining as forming a link between two nodes, and splitting as breaking such a link?

The relevance to software development

I’m not sure what conclusions one can make from this, in trying to regard programs as networks. I’m reading this book because the founder of Information Laboratory (the company behind the Small Worlds analysis tool; see my previous posts: here – and here), Alex Iskold, recommended this book in John K. Water’s article Make complexity work for you. (Searching for the article I found his Amazon.com book lists, with lots of interesting books.) I want to understand this better.

If you regard the elements in the source code – whether they are methods, classes, or categories of classes – as the agents in Brian Arthur’s comparison of economics and physics, they seem both smart and dumb. They surely aren’t as complex and unpredictable as humans are, but then again they are more complex than single particles in physics. What I do know is that the interactions between “software agents” can get very complex, which is why hunting down bugs can sometimes take very long. ––– Got to go.