Reviewed by
David Hales
Delft University of Technology, Netherlands
The first chapter of the book states: The book comprises about 250 pages structured into nine concise chapters, a comprehensive index and around 20 pages of notes which expand on the main text and give relevant references and pointers. It is written in an easy to read non-academic style and assumes no prior knowledge of either physics or social science.
The basic idea of the book is that to understand social phenomena one needs to consider the patterns of interaction rather than the atoms (individuals) themselves. It is argued that complex social outcomes can result from potentially simple rules of behaviour practised by individuals. The emphasis is on modelling agents as simple adaptive learners, imitators, heuristic followers and then observing the social patterns that emerge.
The basic claim in the book is that through the construction of simple models we can understand deep truths about social phenomena in general. The argument is that it has worked in physics so why not here? The idea is that good models often throw out lots of details that are not important to the fundamental process of interest and that through such abstraction one can find great generality. On the other hand, the book is careful to state that models need to be informed by reality. General empirical findings must be taken into account. Simplicity and abstraction does not mean empirical reality takes a back seat.
In Chapter 1, called "think patterns, not people", Buchanan lays out his basic assumptions - that many important social phenomena can be thought of as self-organising processes involving simple behaviours at the individual level but complex outcomes at the social level. It is argued that this is a lot like physics and that physics-like models and approaches can help to understand such social phenomena. A number of examples are given of how unintended and counter-intuitive social phenomena can emerge from simple behaviour but complex interactions.
In Chapter 2, called "the human problem", the idea of applying "science" to human social systems is discussed. A number of common objections are considered and rejected. The argument is richly supported with discussion of contemporary phenomenon and the history of science. Unsurprisingly the conclusion is that physics-like scientific approaches can be applied. This area is potentially an intellectual minefield yet the chapter proceeds smoothly without excluding the main objections - this is an impressive achievement.
In Chapter 3, called "our thinking instincts", Buchanan argues for what he
sees as some basic building blocks of human social behaviour. Here we see a
commitment to empirical fact and plausible simplicity. He, quite properly,
makes a number of damming observations on the inadequacy of rational action
theory as a basis for understanding social or economic behaviour:
"Call it what you will - a house of cards, emperor with no clothes,
whatever - rational economics couldn't possibly have stood up for much
longer, and it hasn't." [p. 50]
In fact, Buchanan presents rational action theory as an example of social
feedback phenomena itself - something that he discussed throughout the book
giving many examples. I particularly liked his following observation:
"One paper I read several years ago argued that it might be rational, for
any individual economist, to stick with the "rational choice" idea even if
he or she knows it is wrong. After all, since it is still the dominant idea,
economists will do better in terms of their career by arguing for it, rather
than by attacking it." [p. 50]
Buchanan argues that the human brain was evolved in a hunter gatherer world and that we should consider this when thinking about how humans might behave. He states that evidence suggests we are not rational calculators but rather "crafty gamblers and adaptive opportunists".
In Chapter 4, called "the adaptive atom", models of financial markets are considered. An overview is given concerning the kinds of statistical properties found in financial markets - such as fat tails, bubbles, clustered volatility. Brian Arthur's El Farol Bar model (Arthur 1994) is discussed and the related Minority Game formulation. In general Buchanan concludes that models composed of simple adaptive agents more readily explain the dynamics and statistical properties observed in real markets than rational action theories.
In Chapter 5, called "the imitating atom", the idea that humans often copy the behaviours of others around them is presented. Numerous counter intuitive social phenomena are explained with reference to imitation in individuals. Granovetters' (1978) threshold model of riots is examined and related to spin glass type models capturing opinion formation based on imitation.
Buchanan neatly finishes this chapter by concluding that simple kinds of imitation may explain superficial patterns and similarities between individuals (weak interactions) but that group and community cooperation and competition must be based on stronger kinds of social interaction.
In Chapter 6, called "the cooperative atom", the human phenomena of
cooperation, self-sacrifice and altruism are advanced. Buchanan argues that
there are many examples of humans behaving in altruistic ways and we need to
tackle it on its own terms rather than explain it away through second order
benefits. Results of experiments involving human subjects in one-shot games
are discussed as are the altruistic punishment game experiments of Fehr and Gatcher
(Fehr and Gatcher 2002). Cooperation and competition are related through the idea of a group
competition approach. Competition between groups over vast periods of our
history, it is claimed, explains why we are partially hardwired to cooperate
within our in-groups yet often highly uncooperative with out-groups - even
to the point of violence. Hence this scheme is advanced as a way to
potentially understand what appear to be irrational altruism and irrational
destructive selfishness:
"The deepest paradox of social physics may be this - we are inherently
skilled at making peace for the same reason that we are skilled at making
war." [p.139]
In Chapter 7, called "together, apart", Buchanan explores further the
implications of the powerful forces that competition between social groups
can exert. He argues that such a perspective might provide an explanation of
nationalistic and genocidal phenomena. The main idea here is that given,
even arbitrary, group markers or labels (such as race, accent, nationality
or religion) adaptation, imitation and group cooperation can lead to highly
negative interactions between groups. Essentially, the powerful intra-group
cooperation mechanisms previously discussed can lead to inter-group
conflict. Presenting results from simulation models by Axelrod and Hammond
(Axelrod and Hammond 2003), it is shown that out-group "prejudice" emerges
in simple models where agents are tagged with different colours (or labels).
Strategies that support cooperate in-groups, of the same colour, spread. As
Buchanan states:
"In effect, labels put a structure on an otherwise unstructured social
reality and thereby make it possible for people to make better decisions by
taking part in "tribes" [p.152].
It is important to stress that this chapter, although illustrated with
references to Bosnia and Nazi Germany, is qualified with a number of caveats
concerning over interpretation of the abstract "cartoon-like" models
presented. However, the basic argument is that the models demonstrate that
highly negative prejudicial behaviour can emerge from simple interactions
between agents when no such prejudicial behaviour exists initially. Rather
that individuals become trapped into group dynamics that they can not
escape. The interpretation of Buchanan, though, is that such phenomena only
get out of hand in human societies when the conventional mechanisms of
social interaction break down (such as economic collapse) and people are
forced to use crude labels to structure their interactions. In addition he
observes that genocide often requires the encouragement of a politically
motivated minority who exploit the situation for political ends:
"So the lesson of social physics, if you will, is that ethnic hatred is a
primitive "mode" of human collective behaviour, akin to the natural
vibrations of a guitar string or the swinging of a pendulum. If this weren't
the case, stoking ethnic hatred would never be an effective political
strategy, as it would push against human tendency and inclination.
Politicians play to ethnic fears because they know fear motivates, perhaps,
more basically and immediately than any other emotion. And, in the right
setting, the opportunistic intelligence of a power-hungry individual can
control the actions of millions." [p.160]
In Chapter 8, called "conspiracies and numbers" both wealth inequality and other phenomena are discussed with references to the power-law distributions that are often found. Simple models by physicists have shown how reinforcement mechanisms can lead to such skewed distributions - repeated successful investments, even from luck, multiply wealth for example. Axtell's firm model is presented (Axtell 2001) indicating how a dynamic theory of the firm might be constructed in which employee dynamics drive firm formation and dissolution. Here again, group formation and cooperation is seen to be a dynamic process far from equilibrium yet firm size distributions settle around the empirically observed levels.
In Chapter 9, called "forward to the past", the book sums up the main
conclusions and unpacks the "scientific approach" to social phenomena that
is being championed. Refreshingly Buchanan goes back to to David Hume and
Adam Smith dispensing with the neo-classical school thus:
"It is safe to say the economists' traditional idea - that we're all
hyperrational calculating machines who unfailingly act in our own
self-interest - has not been among the more productive ideas of science.
One might even say that it stands out as a monument to the incursion of a
completely non-scientific way of thinking into human science." [p.191]
In summary, this is very much a popular science book introducing the general reader to social modelling from a physics point of view. Yet its breadth and readability make it a highly useful starting point for those who wish to dig deeper into the area from any academic background and level. It is richly illustrated with stories and facts situating the models discussed within real world phenomena - this brings the work presented alive and makes it easy for the reader to understand the significance of what is being discussed. Given this is a short book, that can be read in a couple of sittings, Buchanan has done an amazing job. I can recommend the book to anyone with an interest in this area.
JASSS readers may wonder where all the wider more complex agent-based modelling work is - it is not discussed in this book. But it is clear from the start of the book that its focus is on a physics approach and certain kinds of models - those that produce social patterns from very simple agent rules and can be generalised over many phenomena. It therefore does not focus on cognitive (Artificial Intelligence based) agents or those models highly tuned for specific situated social phenomena (such as those produced by participatory modelling). Hence this is not a deficit of the book.
One aspect that did disappoint was the lack of high quality illustrations. There are only a handful of figures, mostly simple line drawings in black and white. However this is a minor gripe.
ReferencesAXELROD R and HAMMOND R (2003) The Evolution of Ethnocentric Behaviour. Available at: http://www-personal.umich.edu/~axe/research/AxHamm_Ethno.pdf
AXTELL R (2001) Zipf Distribution of U.S. Firm Sizes. Science 293, 1818-20.
FEHR E and GACHTER S (2002) Altruistic Punishment in Humans. Nature, 415, 137-40.
GRANOVETTER M (1978) Threshold Models of Collective Behaviour. American Journal of Sociology, 83, 1420-43.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2008