Orbital Teapot

Heaping in count data

2010-07-29T09:30:00.000-06:00

Survey questions are often open for interpretation for the respondent. One type of question may ask the respondent for the 'number of times' something has happened.

At my job we use the GPRA to gather data on substance use. The substance use questions ask "How many days in the past 30 days did you use X" (where X is alcohol, cocaine, cannabis, etc).

If someone does something 7 times, they'll say seven, but if they really did it like 26 times, they'd say 25 or 30 instead. When an exact amount is hard to estimate, the respondents tend to round to nearest 5. This bias increases as the estimate increases.

I've only found one real article on the heaping phenomena. In order to model it you need a funky model... a "zero-inflated heaped negative binomial".

Birthtime

2010-05-31T13:11:00.003-06:00

This year (AD 2010) there have two interesting dates relative to my birth.

The first occurred on February 25th, 2010. I had aged 10,000 days. Consequently on July 13th, 2037 - I'll be 54 years and 20,000 days old. If you want to do it yourself in Excel, it's very easy. Just type your your birth date into cell A1, then in A2 type "=A1+10000".

This year is auspicious for my actual birthday though. My birthday is on October 10th, or 10/10. This year marks the date 10/10/10. Also, if you take the binary number 1010 and translate it into decimal it's 10... yeah.

This birthday is actually taking place a week after a friend's wedding and a week before my own wedding. It's going to be a crazy October.

Radiolab: Carl Sagan And Ann Druyan's Ultimate Mix Tape

2010-02-12T11:00:00.003-07:00

Radiolab is by far my favorite NPR program. It's not produced frequently enough though.

I heard this bit on the way to work this morning - Radiolab: Valentine's Day edition.

People particles

2009-07-29T13:47:00.002-06:00

John Hawks responds to the new issue of Science with its focus on complex systems, networks, and social science. It seems that, overall, Hawks sees the value in agent-based modeling and complexity. Individual differences are not the only determining factor of behavior and network effects have to be considered.

I do share his criticism though. That often models assume homogeneity among all the agents. There are models that embrace heterogeneity and investigations of the effects of heterogeneous populations on collective outcomes is actually a specific topic for many complexity researchers now. The other criticism, that any modeling approach that treats the units as particles be called 'physics', is incredibly valid. I despise the term 'socio-physics' or 'econo-physics' - I think they're ridiculous names. They are the result of the group of physicists who crossed disciplines and decided to come up with a new name to distinguish themselves from social science. Well, Hawks also makes a remark about that as well (regarding disciplinary boundaries).

Anywho, I would highly advise against calling agent-based models as "particle models". Agent implies something very different than particle does. The term particle, to me, makes the agents in the system seem very passive, while in many models its key that the agents show, well, agency and have intentions, opinions, and tastes.

I'm glad to see complex systems making their way into the mainstream, although I had hoped I would be embedded more fully in them before that happened. Like catching a wave at the right time.

Gallery of 2-d classification

2009-07-28T09:17:00.002-06:00

Through the Machine Learning Newsgroup I was pointed to a gallery of 2D classifiers.

There are a number of strange or simple classifcation problems such as this one:

And Fawcett (the author) has trained a variety of different classifcation algorithms to try and detect the pattern. Most social scientists will be familiar with the logistic regression. You can see on the above example that logistic regression will fail horribly:

Then, one of my recent machine learning interests, support vector machines, does a bit better:

In any of linear cases logistic regression performs fantastic. While some of the other algorithms will get messy or incorrect results using a small N, the logistic will get an almost perfect classification. This page is a great resource for exploring the power and limitations of several classification methods.

Special Issue of Science: Complexity

2009-07-23T21:43:00.003-06:00

The July 24th Science is devoted to complexity and network science. At every turn in complexity and network science we see physicists, computer scientists, and social scientists in cahoots. Science also has a piece on how network science on its own is growing as its own discipline replete with career paths and funding. Barabási looks friendlier than I thought he would.

Do you enjoy looking at graphs?

2009-07-14T11:13:00.004-06:00

I saw a graph made from age distribution data from the Census Bureau at Economist's View. It's a pretty cool chart. Except for the fact that the default Excel coloring and labellings are uglier than sin.

Excel has some incredible visualization capabilities if you know how to use them. I thought that such interesting presentation of interesting data should not go to waste so I produced my own version also in Excel:

It's like geological strata on the side of the highway.

Free downloadable book on computational modeling

2009-06-15T14:12:00.002-06:00

It's mighty scant at this point, but it includes plenty of problems for people who might want to challenge themselves and improve their understanding of the topics. And it's at Version 0.0.10... it has a ways to go. It's all done in Python too, for anyone who might be interested in Python.

I'd probably do everything in R though.

Its the simple things

2009-06-08T09:24:00.007-06:00

vs.

I found it really difficult to tell when the trend was negative or positive without a good 0-line. Also, you probably only have to label every few month labels, like Jan., Apr., July., Oct. - the current axis makes everything all crowded.

Via Traffic

On Amazon, other people can't see your shopping cart

2009-05-29T11:49:00.003-06:00

This reminds me of the parable of beer and diapers. I guess moms need more than just a rest break.

Reading list on complex networks

2009-05-20T12:00:00.000-06:00

Apparently there's a reading group in Santa Fe on complex networks. The list provided is excellent.

This one looked interesting:

Aguirre, B. E., Quarantelli, E. L. and Mendoza, J. L. The collective behavior of fads: the characteristics, effects, and career of streaking. American Sociological Review 53, 569-584 (1988).

Evolution of the bio-tech industry

2009-05-19T14:29:00.000-06:00

Herein are slides on the development of the bio-tech industry. I honestly can't recall where I found the link to this, but I like topic. The best part is it uses bipartite networks and doesn't collapse the different classes of nodes and studies them in the raw.

The interesting part about the dynamics of this network is that it grows and grows until it hits a capacity some time in 1992. The authors find a power-law distribution in the number of links, which is pretty typical for such networks. And then the vogue thing to do with dynamic networks with power-law distributions is to look at the change in 'alpha' over time. There's not a lot to gain from the slides though, I'm interested in the paper this might go along with.

So I found Powell's website and found this paper on Network Dynamics and Field Evolution. It seems similar, but not exactly. Anywho - it's a good example of anlysis of a dynamic two-mode network.

Trends in cognitive science: sociology

2009-05-19T09:23:00.003-06:00

Computational Models of Collective Behavior

From the conclusion:

Cognitive scientists often act as though individuals are the sole loci of organized thought, but ABMs remind us that organized behavior can be described at multiple levels, and that our thoughts both depend upon and determine the social structures that contain us as elements within those structures.

A nexus of political network lit.

2009-05-18T14:27:00.002-06:00

I found a really big list of political network literature sorted by topic. For anyone working on networks of political networks (qualitative or quantitatively) this could be a good resource.

Taste tripping parties

2009-05-17T16:26:00.001-06:00

Wild.

Berries that contain a substance that binds to taste buds. The result is sour taste is attenuated and sweetness is amplified.

“You kept hearing ‘oh, oh, oh,’ ” he said, and then the guests became “literally like wild animals, tearing apart everything on the table.”

“It was like no holds barred in terms of what people would try to eat, so they opened my fridge and started downing Tabasco and maple syrup,” he said.

The Necessity of Complexity: Taking environmental-conflict research beyond mechanism

2009-05-17T09:50:00.000-06:00

The Necessity of Complexity: Taking environmental-conflict research beyond mechanism
Tom Deligiannis, Thomas Homer-Dixon, and Dirk Druet

The authors argue that there is a deep divide between qualitative and quantitative (as I took it) researchers regarding violence and environmental scarcity. The quants argue that there is little correlation between scarcity and violence, while the quals think case studies show clear causal relationships. Enter complexity science: "A mechanistic ontology is entirely inappropriate for investigation of causal processes within socio-ecological systems, because these systems are fundamentally "complex." Recent advances in complexity theory show that such systems are characterized by causal openness, emergent properties, disproportionality of cause and effect (i.e., nonlinear behavior), fractal scaling, and causal interaction (synergy)." They argue that a complexity approach reveals a number of the missing nonlinear causal connections between violence and scarcity.

The paper makes it incredibly obvious that there is a lot of controversy over the topic. Much of it is laid at the creation of variables for abstract concepts (How the heck do you quantify environmental scarcity? asks one research. Like this! says another. That's stupid. says the first researcher. So it goes...). Of course the debate is far more nuanced than this paper could even elaborate, but that's not the point. What we want to know is what a complexity perspective can give to this debate.

The most powerful element of complexity theory, according to the authors, is its ability to reframe causality. The authors take issue with the idea of 'if X occurs then Y occurs' and suggest that multiple causes can act together to produce an effect. The idea is that environmental scarcity can cause conflict, but it is not necessary and sufficient to cause conflict - I think that is obvious, but it appear the authors are criticizing other researchers who pose a strawman of cause-and-effect ideas about environmental scarcity.

They use a case study of community unrest in Peru to illustrate this idea. I feel that I couldn't quite summarize the list of causal relationships that lead to the violence in the region. There was a mixture of migration, lack of government legitimacy, environmental scarcity, over-population, and heterogeneity. One thing causes another which amplifies something else, which aggravates the first thing, and *pow* avalanche! For instance migration from one area to another might mean that the new area is not equipped to handle the new population resulting in lack of water, land, and food. Scarcity is a problem, but overpopulation is a problem in and of itself - and both together is a bigger problem than either in isolation.

I feel this paper dressed up pre-existing ideas about causality with a complexity veneer. While complexity science really does bring a lot of explanatory power to complex social situations, the author's use of it in this context seemed superficial. Well, at least in places. I think the application to a review of causality, using complexity science for legitimation, was effective. But there probably was little need of injecting unnecessary complexity jargon words into the discussion. Plus there was not very much illustration of what exactly complexity science and it's corresponding theories have done and truly can do. I agree, complexity is a necessary consideration, but I don't think someone who wasn't already convinced of that would buy into it from this paper.

Financial Contagion on the International Trade Network

2009-05-16T12:00:00.000-06:00

Financial Contagion on the International Trade Network
Raja Kali & Javier A. Reyes

I've posted about this paper before when the financial collapse was on fire. As before I quoted that the international trade system is affected more if a well-connected country is in crisis than if a less-connected country is. Conversely more well connected countries can absorb the shock better, probably because they transfer all that 'crisis' to others.

I found this paper incredibly interesting because it is one of the few papers that integrates network statistics, like centrality, into the statistical models of an outcome. I'm surprised that this is not done more often. The downside is that they don't do their analysis on a weighted network as they should. Instead they generate the network using various thresholds for link weight. This makes the analysis easier, but it also means they have like 28 statistical models testing for the significance of various thresholds. I found it interesting that trade with the 'epicenter country' (the country where a particular crisis occurred) wasn't terribly significant in predicting outcomes.

Overall network measures are very significant in predicting outcomes, even in the presence of other well-known control variables. It happens to explain even if a country's economy is small a crisis within can have large and lasting effects if it is well-connected. The end by suggesting that a network approach to financial shocks has important policy implications. They suggest further 'what-if' simulation using network models in the future since they are more important than previously considered.

Micromotives and Macrobehavior

2009-05-16T09:54:00.000-06:00

Micromotives and Macrobehavior
Thomas Schelling

How can we theorize the origins of segregation? How sensitive is segregation to the whims of actors? Thomas Schelling conceived a model of segregation that helps illuminate a number of tricky questions and produces some stunning insights. The model was first run on a checker board. The idea is that agents may desire at least some degree of similarity of people they move in next door to. If they do not find a desired level of similarity they move randomly to an open spot. The process continues.

Checker boards are slow though, so if you want to play with this idea with the NetLogo model and thousands of actors, please do so. You will find that even a desire for 30% similarity among the agents will globally produce 70% similarity. I played with the model and found a nonlinear relationship - at about 76% desired similarity you get about 99.9% global similarity. Over 76% desired (with 2000 agents) you get chaos, the system never crystallizes.

What the heck is the invisible hand of the economy? I think that's what this book seeks to explain. And it's a fantastic book. Really, I can't play it up enough. It's going to be assigned in classes I teach in the future, I spent half the time reading it thinking 'ooohhhh, right!'

Schelling is really looking for true principles of organization. For instance the Inescapable Mathematics of Musical chairs - someone is always left without a chair when a stable state is achieved. There are some inescapable facts of some social systems. For instance one person's raise is everyone else's inflation.

One thing this book really counters, like Sawyer did (but more concisely and effectively, I think), is methodological individualism. Individual experience only marginally explains global effects. I think case-studies and interviews are useful for understanding, but they tend to ignore how individual behavior interacts with the rest of the system and creates structure that can in turn affect behavior. Like how Schelling opens his book about a lecture he gave where everyone sat in the back of the auditorium and the first 15 rows were completely empty. Agent decision was important in this, but not as important as path-dependency. When the first few people sat down it set a norm that newcomers conformed to. Assuming humans are predictable behavior systems, initial seating conditions are more responsible for the seating arrangement than the individuals who sat down.

The book is rife with examples. Schelling often starts listing off cases where his principles are displayed. "I walk across the lawn if that seems to be what others are doing; I sometimes double-park if it looks as though everybody is double-parked. I stay in line if everybody is standing politely yin line, but if people begin to surge toward the ticket window I am alert to be - though never among the first - not among the last" (93).

Schelling comments on a number of phenomena common to the social science and complexity science. He mentions 'tipping points' for instance. This is a result of feedback. If we invite a certain number of people to a party there is an expectation among those invited that not everyone will attend. Since no one wants to be the few awkward people at a party people will hesitate. If there is knowledge that plenty of people are invited, so much so that the liklihood of being an awkward few is small, then there is a critical change and a higher proportion of people invited will actually attend. That's the idea at least, it's because of positive feedback, but the term isn't mentioned here.

The emphasis of the book is on modeling phenomena as a way of understanding it. Schelling is particularly interested in modeling sorting and aggregation of income, race, and cultural preference. His modeling assumes that the read has the mathematical prowess of an average economist (which, I think he is an economist, actually, which would explain the chapter on intersecting curves). Like Miller and Page's book, Schelling emphasizes the science over the technical bits of modeling. Good science always comes first and should provide insights and understanding to complex social situations. This book predates much of the vernacular of complexity science, but it belongs right smack in the middle of it. *recommended*

Coping with Complexity: The adaptive value of changing utility

2009-05-15T15:00:00.003-06:00

Cohen, Michael D and Axelrod, Robert (1984). Coping with Complexity: The Adaptive Value of Changing Utility American Economic Review, 74 (1), 30-42

I took this paper on particularly because of the authors and because I thought it might be a short version of Harnessing Complexity, their book on using complexity science ideas in organization management. The paper is about management and managing complexity, actually, as per the title. It moves from econometric modeling and implies that even the best models are incomplete because complex organizations always have elements that screw up your model.

They create a model based on an AI model of playing checkers- yes, checkers, I know, it's pointless because we've solved checkers. The AI model for checkers based learning on 'surprise.' If ever a move on the other player's part was unexpected and the algorithm is surprised then there is a change in preferences. The example they start with is how a manager is trying to maximize productivity by choosing the right amount of labor. The manager is assumed to use an inverted parabola as the utility function, after each time period the basic algorithm would adjust a coefficient as it searches for the optimal distribution of labor for productivity.

But there could be things like thievery! The manager did not account for this. So Cohen and Axelrod quantify a measure of surprise, which to me, just feels like a residual from a regression model. They then suggest a dynamic model of adjusting preferences to account for the hidden complexity. The model adjusts preferences, but not too much since it dampens the changes.

I didn't see a terribly large amount of application here. This paper is from 1984 and it is reasonable to expect that improved econometric models have come into vogue since then. The one thing I did like was that "the simple principles underlying the success of the Samuel Checker Player can be transfered with powerful effect to other task domains." The other importance here is that hidden complexity can be something that can be 'coped' with. Perhaps their book, mentioned above, goes much further than that.

Emergence: The Dynamics of Network Formation

2009-05-15T12:00:00.001-06:00

Emergence: The Dynamics of Network Formation
Brian Uzzi, Roger Guimera, Jarrett Spiro, Luis Amaral

This paper could almost be considered the culmination of everything I've read on my list. It has a lot of elements, network analysis, network modeling, empirical data, and empirical verification of a model. And it's all wrapped up by a veneer of Broadway musicals. I'm a fan of Brian Uzzi and Roger Guimera, they both do some very interesting work. In this case, they and their coauthors hope to explain the mechanisms of growth for bipartite networks. They examine empirical data of the growth of co-workers on broadway musical playbills and suggest a model where the type of link is important. They even analytically solve the model and make predictions which they successfully test on their data.

When researchers construct networks from existing data we have to ask questions such as how they define actors and how they define links. In this case the network is originally bipartite. The actors are directors, costume designers, actors (the acting kind) who appear on a playbill for a musical. They form links by being on the same playbill together. The network when it is expressed and modeled is not bipartite though. Bipartite networks are two-colorable, meaning you can assign every node one of two colors and no adjacent nodes will have the same color. A true representation of the network would have actors and playbills as two kinds of nodes in the network. Instead they treat the playbill as creating 'cliques' where every node in the playbill is connected to every other node in the playbill. As the network evolves there aren't so much as links being created so much as cliques.

When the network starts out there are only newcomers. Newcomers have collaborated on their first work and appear on a playbill for the first time. Eventually we see incumbents who have worked before and they can form links with other incumbents. There are several levels of links N-N (newcomer, newcomer), N-I (newcomer, incumbent), I-I (incumbent, incumbent, but the actors have not collaborated before, they're just incumbent), and I-R (same as incumbent incumbent except that both actors had collaborated on something before. These types of links, it is hypothesized, drive the emergence of the fully connected network.

There are a number of parameters: p, the probability of linking to an incumbent; q, the probability of linking to a past collaborator; m, the "team-size" which means the number of people on the playbill; and f, the amount of time you keep an inactive actor on the field before removing him.

The model displays some interesting dynamics. The image above is from a NetLogo simulation of the model you can play with. I liked setting all the variables constant but varying, up and down, the q variable. This made the network very modular, while otherwise it would have just been a single densely connected community.

The best part, which is something terribly lacking from most models, is the empirical validation. They analytically solve for the predicted fraction of of N-N, N-I, and I-I and find that all the parameters but p drop out (which is interesting). Then they show that the curves match the empirical data very well.

I'm left wondering what else could be done with this model. If it is a valid model then what sort of perturbations can we make that produce interesting predictions that we can go out and test for? Like varying the probability of working with a past collaborator, does that really produce more modular networks, like I was finding? Are there social systems where that value actually does oscillate where we can test the modularity? Conference papers can be pretty cool sometimes.

A systems approach to unravel complex water management institutions

2009-05-15T12:00:00.000-06:00

A systems approach to unravel complex water management institutions

This paper was not what I was looking for at all. The methods were interesting as I'd never seen an implementation of Bayesian network analysis before. This paper is a good reminder that 'systems' approaches and 'complexity' approaches are different in some subtle but important ways. The goal is to understand the way water management rights are decided by analyzing how much surveyed individuals felt one thing led to a feeling about something else. The idea was that there were 'rules' governing the behavior of agents in an institutional water management environment in a hamlet in the Indian Himalayas and that decisions made did not account for the adaptive behavior of agent beliefs in their arena.

There were a number of rules the author defined in the system. The first, and which I thought most important, were the boundary rules. Control of boundary rules were critical since they defined who had some control and who did not. For instance, India's Land Reform Act defined who had the right to act in the water management based on land ownership. On top of that court rulings had an effect on the definition of land ownership as well, so two different acts defined actors in the system.

There were a number of things I didn't understand. For instance this sentence: "About 67% of the households perceived the probability of inadequate leadership affecting water distribution." What does that mean? They perceived it to be inadequate? They perceived the probability to be what? Did they just perceive it to be probable? There were something things I understood, for instance the periphery of the belief network were the boundary rules, which fed into position rules. In essence we could map out how boundary rules (defined who the actors were) interacted with position rules (the political position the actors take) with lead to aggregation rules (perception of the outcomes).

The goal the author had was to do away of the linear approach to outcomes and consider a nonlinear approach where relationships between actors and rules were important. I think the author accomplished this. But I don't feel any more enlightened by the approach. I think there was better information in the narrative than there was in the Bayesian network - the Bayesian network seems completely superfluous. I think this is a clear case where the methodology has outstripped the ability to fruitfully apply it.

Illegal and saves lives?

2009-05-14T18:13:00.004-06:00

Per a CNN story:

Cheney has said he wants the documents released so there can be a more "honest debate" on the Bush administration's approval of "alternative" interrogation techniques against suspected terrorists. He argued that those techniques provided valuable intelligence that saved American lives, but critics say they amounted to the illegal torture of prisoners in U.S. custody.

Emphasis is mine. Why are these two things exclusive? Maybe there could be some honest dialogue if people dumped their tribal idealism and confronted the issue in the raw.

Anywho - I'm wholly against torture. People should experience consequences for breaking the law. I listened to a really interesting story on NPR about contracted interrogators who were mostly responsible. The best quote, I thought (and I'm probably butchering it), was when a senator said that 'torture persists because it works.' And Ali Soufan replied, 'No, torture persists because it's easier to hit someone than outsmart them.' So true.

Science stands mute

2009-05-14T12:35:00.001-06:00

It is possible to search twitter then set up a feed for the search to put in your reader. So I have one set up for "complexity science". A single quote has been hitting that feed rampantly over the past few days:

There is a complexity to human affairs before which science and analysis simply stands mute.

It refers to an op-ed article by David Brooks in the NYT. The article refers to a longitudinal study of 268 men called the Grant Study. It seems the purpose was the see what makes people happy and how to get there.

The quote bothers me since it implies social science is a fruitless endeavor. I think the quote has spread through Twitter because it says something about the ethereal qualities of human relationships. People revel when they have access to something that science does not, like a soul maybe. When an exhaustive longitudinal study reveals "Happiness is love" there is a common relief that their intuitions are correct and the science is useless. Their connection to an realm outside the bounds of this physical dreary world is secure.

Brooks though is commenting a single study and a single essay. There has been gobs of research*, from psychology, sociology, economics, etc, recently studying happinness from a number of angles. Science doesn't stand mute on the topic... just that one ill-designed study of a single convenience sampling of white males.

* - ask me later if you really want me to start citing stuff. I need a collection of happiness paper and cites for my endnote anyhow.

P.S. Let's extend that 'happiness is love' into science, no?

Inequality and network structure

2009-05-14T12:00:00.002-06:00

Inequality and network structure
Garud Iyengar, Willemien Kets, Rajiv Sethi, and Samuel Bowles

One of the more powerful implications of network analysis is that structure is an explanation. How things are connected to each other can effect the the distribution of resources or cooperative behavior. Here we have a paper that explores this idea in strictly mathematical terms - wow, it takes me back to advanced analysis. I haven't had to consider the behavior convex functions* in quite awhile.

The authors build off work by Myerson (1977) on cooperation on graphs. So, the way their model works is that each player in the network generates "value" if they are connected to other players. In addition the players can break off and define their own network, in which case their value is distributed only with their own network and they may or may not cut off some players' connections to other players. A player will only break off if they gain from it. The gains are relative to some function that determines value based on the number of individuals in the network 'clique'**.

The main implications of this model is that network structure is important, but the traditional measures of centrality, like betweenness and degree, are not. For instance the authors show examples of networks where the player with the highest betweenness and degree is the least valued. They also show an example of two networks with the same number of players, but the one with the more unequal degree distribution is actually more egalitarian in distributing value. They extend and continue having fun with their new model showing that all bipartite networks have a unique extremal distribution and extend the model allowing groups of players within k-distance of each other to break off.

This was just a discussion paper, but it does bring up some interesting points. For instance, if value is determined by those you interact with then it is certainly possible that someone several steps disconnected from me can disconnect me from value creation. Because each player has the choice of defecting to their own network if it benefits them we find the most central player, contrary to much network theory, is not the wealthiest. However, as the authors point out, this model is far too abstract to be empirically verified. It's value is in the theory and method, methinks.

* - in mathese that means you can draw a straight line and intersect the function at two points.

** - clique is defined totally differently than what I'm used to... sigh.

Swarm Intelligence

2009-05-14T09:54:00.001-06:00

Swarm Intelligence
James Kennedy and Russell C Eberhart

This was a really cool book. I will have to get the last few chapters later on, since I focused on the first half of the book. The first half is about human society as an information processor. Swarm optimization is essentially a model of social learning that has optimization applications, but was originally conceived as a method of exploring theories of social cognition. The book starts by discussing social theory, brains, language, information, culture, etc, and eventually moves in to the more technical application of coding and implementing swarm optimization. The emphasis of this book is not on the individual but on the social. The interactions between actors is what creates social intelligence and culture and that is what they hope to model and (since one of them is an engineer) exploit.

This book reminded me of the reasons I first went into social science. I wanted to explore the evolution of culture as a thing. The book starts by discussing modeling intelligence and what intelligence really is. They integrate ideas such as emergence and evolutionary computation. Then they move into culture suggesting that 'truth evolves' specifically since it relies on fuzzy systems like the brain to interpret symbols and produce meaning. The next chapters give an introduction to evolution and optimization, we even get some brief commentary on the creation / evolution debate*. Eventually we get to a topic I tend to find very interesting: flocks, herds, and swarms, and treating social behavior as optimization.

The idea is that even random (sorry, stochastic) movement can induce learning. A snail moving about a space is essentially searching for food. I often feel social scientists (well, sociologists in particular) are hesitant to draw parallels between other biological phenomena of organization and social learning and human methods of social learning. Kennedy and Eberhart don't even hesitate, and contrary to other points in the book where they justify evolution or justify modeling, we get little justification for the use of sociobiology. Ants are such an amazing evolved system for problem solving and decision making. The chapter on herds and swarms illustrates how ants problem solve, how this system can be modeled, and what it can tell us about how humans process and act on information.

On a very local level humans appear chaotic and unpredictable, but at the right scale human groups tend to behave in a very fluid manner. I've always taken the view that organisms respond and adapt to their own environment, but what if the environment is made of other organisms like them? Maybe all that complexity in human behavior is not so much that humans individually are unpredictable, but that we can't control the multitude of influences on behavior. Herbert Simon suggests, "A man, viewed as a behaving system, is quite simple. The apparent complexity of his behavior over time is largely a reflection of the complexity of the environment in which he finds himself" (100). Especially if that environment is other humans reacting to other humans. The book then moves into models and science behind social influence (and we see Axelrod come up again, damn he has made SO many influential models of human behavior... so many).

At the end we get our math and our code and our engineering. The particle swarm optmization model uses most of the ideas from the preceding chapters. The particles are networked together to each other and seek consistency with those they are connected to. They also try to optimize their own fitness while still maintaining that consistency of those they are connected to. In this way, if you put a particle swarm in a space the particles tend towards optimal solutions, and since they are stochastic like gnats they also explore the space around optimal solutions which leads to even better solutions.

They end the book by saying, "It is almost unbelievable, even to us, that the computer program that started out as a social-psychology simulation is now used to optimize power grids in Asia; develop high-tech products in the United States; and to solve high-dimensional, nonlinear, noisy mathematical problems" (427). The suggest that study of this 'paradigm' is in its infancy and is far from exhausted. This book gives a fantastic and incredibly cool look at social science, modeling, evolution, information, culture, and computation. I definitely could not spend enough time digging into it, as I would desperately like to code my own particle swarm optimization and experiment with it. Eventually... I have to keep focused on what I have, I guess.

* - "The disagreement between religious advocates and evolutionary scientists comes down to this: the creationists know how life began on earth, and the evolutionists don't." (82)