Computational social science

Is it possible to elucidate complex social outcomes using computational tools? Can we overcome some of the issues for social explanation posed by the fact of heterogeneous actors and changing social environments by making use of increasingly powerful computational tools for modeling the social world? Ken Kollman, John Miller, and Scott Page make the affirmative case to this question in their 2003 volume, Computational Models in Political Economy. The book focuses on computational approaches to political economy and social choice. Their introduction provides an excellent overview of the methodological and philosophical issues that arise in computational social science.

The subject of this book, political economy, naturally lends itself to a computational methodology. Much of political economy concerns institutions that aggregate the behavior of multiple actors, such as voters, politicians, organizations, consumers, and firms. Even when the interactions within and rules of a political or economic institution tion are relatively simple, the aggregate patterns that emerge can be difficult to predict and understand, particularly when there is no equilibrium. It is even more difficult to understand overlapping and interdependent institutions.... Computational methods hold the promise of enabling scholars to integrate aspects of both political and economic institutions without compromising fundamental features of either. (kl 27)

The most interesting of the approaches that they describe is the method of agent-based models (link, link, link). They summarize the approach in these terms:

The models typically have four characteristics, or methodological primitives: agents are diverse, agents interact with each other in a decentralized manner, agents are boundedly rational and adaptive, and the resulting patterns of outcomes comes often do not settle into equilibria.... The purpose of using computer programs in this second role is to study the aggregate patterns that emerge from the "bottom up." (kl 51)

Here is how the editors summarize the strengths of computational approaches to social science.

First, computational models are flexible in their ability to encode a wide range of behaviors and institutions. Any set of assumptions about agent behavior or institutional constraints that can be encoded can be analyzed. 

Second, as stated, computational models are rigorous in that conclusions follow from computer code that forces researchers to be explicit about assumptions. 

Third, while most mathematical models include assumptions so that an equilibrium exists, a system of interacting political actors need not settle into an equilibrium point. It can also cycle, or it can traverse an unpredictable path of outcomes. 

The great strength of computational models is their ability to uncover dynamic patterns. (kl 116)

And they offer a set of criteria of adequacy for ABM models. The model should explain the results; the researcher should check robustness; the model should build upon the past; the researcher should justify the use of the computer; and the researcher should question assumptions (kl 131).

To summarize, models should be evaluated based on their ability to give insight and understanding into old and new phenomena in the simplest way possible. Good, simple models, such as the Prisoner's Dilemma or Nash bargaining, with their ability to frame and shed light on important questions, outlast any particular tool or technique. (kl 139)

A good illustration of a computational approach to problems of political economy is the editors' own contribution to the volume, "Political institutions and sorting in a Tiebout model". A Tiebout configuration is a construct within public choice theory where citizens are permitted to choose among jurisdictions providing different bundles of goods.

In a Tiebout model, local jurisdictions compete for citizens by offering bundles of public goods. Citizens then sort themselves among jurisdictions according to their preferences. Charles M. Tiebout's (1956) original hypothesis challenged Paul Samuelson's (1954) conjecture that public goods could not be allocated efficiently. The Tiebout hypothesis has since been extended to include additional propositions. (kl 2012)

Using an agent-based model they compare different sets of political institutions at the jurisdiction level through which policy choices are made; and they find that there are unexpected outcomes at the population level that derive from differences in the institutions embodied at the jurisdiction level.

Our model departs from previous approaches in several important respects. First, with a few exceptions, our primary interest in comparing paring the performance of political institutions has been largely neglected in the Tiebout literature. A typical Tiebout model takes the political institution, usually majority rule, as constant. Here we vary institutions and measure performance, an approach more consistent with the literature on mechanism design. Second, aside from an example used to demonstrate the annealing phenomenon, we do not explicitly compare equilibria. (kl 2210)

And they find significant differences in collective behavior in different institutional settings.

ABM methodology is well suited to the kind of research problem the authors have posed here. The computational method permits intuitive illustration of the ways that individual preferences in specific settings aggregate to distinctive collective behaviors at the group level. But the approach is not so suitable to the analysis of social behavior that involves a higher degree of hierarchical coordination of individual behavior -- for example, in an army, a religious institution, or a business firm. Furthermore, the advantage of abstractness in ABM formulations is also a disadvantage, in that it leads researchers to ignore some of the complexity and nuance of local circumstances of action that lead to significant differences in outcome.

Is public opinion part of a complex system?

The worrisome likelihood that Russians and other malevolent actors are tinkering with public opinion in Western Europe and the United States through social media creates various kinds of anxiety. Are our democratic values so fragile that a few thousand Facebook or Twitter memes could put us on a different plane about important questions like anti-Muslim bigotry, racism, intolerance, or fanaticism about guns? Can a butterfly in Minsk create a thunderstorm of racism in Cincinnati? Have white supremacy and British ultra-nationalism gone viral?

There is an interesting analogy here with the weather. The weather next Wednesday is the net consequence of a number of processes and variables, none of which are enormously difficult to analyze. But in their complex interactions they create outcomes that are all but impossible to forecast over a period of more than three days. And this suggests the interesting idea that perhaps public opinion is itself the result of complex and chaotic processes that give rise to striking forms of non-linear change over time.

Can we do a better job of understanding the dynamics of public opinion by making use of the tools of complexity theory? Here is a summary description of complex systems provided by John Holland in Complexity: A Very Short Introduction:

Complexity, once an ordinary noun describing objects with many interconnected parts, now designates a scientific field with many branches. A tropical rainforest provides a prime example of a complex system. The rainforest contains an almost endless variety of species—one can walk a hundred paces without seeing the same species of tree twice, and a single tree may host over a thousand distinct species of insects. The interactions between these species range from extreme generalists (‘ army’ ants will consume most anything living in their path) to extreme specialists (Darwin’s ‘comet orchid’, with a foot-long nectar tube, can only be pollinated by a particular moth with a foot-long proboscis—neither would survive without the other). Adaptation in rainforests is an ongoing, relatively rapid process, continually yielding new interactions and new species (orchids, closely studied by Darwin, are the world’s most rapidly evolving plant form). This lush, persistent variety is almost paradoxical because tropical rainforests develop on the poorest of soils—the rains quickly leach all nutrients into the nearest creek. What makes such variety possible? (1)

Let's consider briefly how public opinion might fit into the framework of complexity theory. On the positive side, public opinion has some of the dynamic characteristics of systems that are often treated as being complex: non-linearity, inflection points, critical mass. Like a disease, a feature of public opinion can suddenly "go viral" -- reproduce many times more rapidly than in previous periods. And the collective phenomenon of public opinion has a feature of "self-causation" that finds parallels in other kinds of systems -- a sudden increase in the currency of a certain attitude or belief can itself accelerate the proliferation of the belief more broadly.

On the negative side, the causal inputs to public opinion dynamics do not appear to be particularly "complex" -- word-of-mouth, traditional media, local influencers, and the new factor of social media networks like Twitter, Weibo, or Facebook. We might conceptualize a given individual's opinion formation as the net result of information and influence received through these different kinds of inputs, along with some kind of internal cognitive processing. And the population's "opinions" are no more than the sum of the opinions of the various individuals.

Most fundamentally -- what are the "system" characteristics that are relevant to the dynamics of public opinion in a modern society? How does public opinion derive from a system of individuals and communication pathways?

This isn't a particularly esoteric question. We can define public opinion at the statistical aggregate of the distribution of beliefs and attitudes throughout a population -- recognizing that there is a distribution of opinion around every topic. For example, at present public opinion in the United States on the topic of President Trump is fairly negative, with a record low 35% approval rating. And the Pew Research Center finds that US public opinion sees racism as an increasingly important problem (link):



Complexity theorists like Scott Page and John Holland focus much attention on a particular subset of complex systems, complex adaptive systems (CAS). These are systems in which the agents are themselves subject to change. And significantly, public opinion in a population of human agents is precisely such a system. The agents change their opinions and attitudes as a result of interaction with other agents through the kinds of mechanisms mentioned here. If we were to model public opinion as a "pandemonium" process, then the possibility of abrupt non-linearities in a population becomes apparent. Assume a belief-transmission process in which individuals transmit beliefs to others with a volume proportional to their own adherence to the belief and the volume and number of other agents from whom they have heard the belief, and individuals adopt a belief in proportion to the number and volume of voices they hear that are espousing the belief. Contagion is no longer a linear relationship (exposure to an infected individual results in X probability of infection), but rather a non-linear process in which the previous cycle's increase leads to amplified infection rate in the next round.

Here is a good review article of the idea of a complex system and complexity science by Ladyman, Lambert and Wiesner (link, link). Here is a careful study of the diffusion of "fake news" by bots on Twitter (link, link). (The graphic at the top is taken from this article.) And here is a Ph.D. dissertation on modeling public opinion by Emily Cody (link).

Complexity and contingency

One of the more intriguing currents of social science research today is the field of complexity theory. Scientists like John Holland (Complexity: A Very Short Introduction), John Miller and Scott Page (Complex Adaptive Systems: An Introduction to Computational Models of Social Life), and Joshua Epstein (Generative Social Science: Studies in Agent-Based Computational Modeling) make bold and interesting claims about how social processes embody the intricate interconnectedness of complex systems.

John Holland describes some of the features of behavior of complex systems in these terms in Complexity:

• self-organization into patterns, as occurs with flocks of birds or schools of fish  
• chaotic behaviour where small changes in initial conditions (‘ the flapping of a butterfly’s wings in Argentina’) produce large later changes (‘ a hurricane in the Caribbean’)  
• ‘fat-tailed’ behaviour, where rare events (e.g. mass extinctions and market crashes) occur much more often than would be predicted by a normal (bell-curve) distribution  
• adaptive interaction, where interacting agents (as in markets or the Prisoner’s Dilemma) modify their strategies in diverse ways as experience accumulates. (p. 5)

In CAS the elements are adaptive agents, so the elements themselves change as the agents adapt. The analysis of such systems becomes much more difficult. In particular, the changing interactions between adaptive agents are not simply additive. This non-linearity rules out the direct use of PDEs in most cases (most of the well-developed parts of mathematics, including the theory of PDEs, are based on assumptions of additivity). (p. 11)

Miller and Page put the point this way:

One of the most powerful tools arising from complex systems research is a set of computational techniques that allow a much wider range of models to be explored. With these tools, any number of heterogeneous agents can interact in a dynamic environment subject to the limits of time and space. Having the ability to investigate new theoretical worlds obviously does not imply any kind of scientific necessity or validity— these must be earned by carefully considering the ability of the new models to help us understand and predict the questions that we hold most dear. (Complex Adaptive Systems, kl 199)

Much of the focus of complex systems is on how systems of interacting agents can lead to emergent phenomena. Unfortunately, emergence is one of those complex systems ideas that exists in a well-trodden, but relatively untracked, bog of discussion. The usual notion put forth underlying emergence is that individual, localized behavior aggregates into global behavior that is, in some sense, disconnected from its origins. Such a disconnection implies that, within limits, the details of the local behavior do not matter to the aggregate outcome. Clearly such notions are important when considering the decentralized systems that are key to the study of complex systems. Here we discuss emergence from both an intuitive and a theoretical perspective. (Complex Adaptive Systems, kl 832)

As discussed previously, we have access to some useful “emergence” theorems for systems that display disorganized complexity. However, to fully understand emergence, we need to go beyond these disorganized systems with their interrelated, helter-skelter agents and begin to develop theories for those systems that entail organized complexity. Under organized complexity, the relationships among the agents are such that through various feedbacks and structural contingencies, agent variations no longer cancel one another out but, rather, become reinforcing. In such a world, we leave the realm of the Law of Large Numbers and instead embark down paths unknown. While we have ample evidence, both empirical and experimental, that under organized complexity, systems can exhibit aggregate properties that are not directly tied to agent details, a sound theoretical foothold from which to leverage this observation is only now being constructed. (Complex Adaptive Systems, kl 987)

And here is Joshua Epstein's description of what he calls "generative social science":

The agent-based computational model— or artificial society— is a new scientific instrument. 1 It can powerfully advance a distinctive approach to social science, one for which the term “generative” seems appropriate. I will discuss this term more fully below, but in a strong form, the central idea is this: To the generativist, explaining the emergence2 of macroscopic societal regularities, such as norms or price equilibria, requires that one answer the following question:  

The Generativist's Question 

*     How could the decentralized local interactions of heterogeneous autonomous agents generate the given regularity?  

The agent-based computational model is well-suited to the study of this question since the following features are characteristics. (5)

Here Epstein refers to the characteristics of heterogeneity of actors, autonomy, explicit space, local interactions, and bounded rationality. And he believes that it is both possible and mandatory to show how higher-level social characteristics emerge from the rule-governed interactions of the agents at a lower level.

There are differences across these approaches. But generally these authors bring together two rather different ideas -- the curious unpredictability of even fairly small interconnected systems familiar from chaos theory, and the idea that there are simple higher level patterns that can be discovered and explained based on the turbulent behavior of the constituents. And they believe that it is possible to construct simulation models that allow us to trace out the interactions and complexities that constitute social systems.

So does complexity science create a basis for a general theory of society? And does it provide a basis for understanding the features of contingency, heterogeneity, and plasticity that I have emphasized throughout? I think these questions eventually lead to "no" on both counts.

Start with the fact of social contingency. Complexity models often give rise to remarkable and unexpected outcomes and patterns. Does this mean that complexity science demonstrates the origin of contingency in social outcomes? By no means; in fact, the opposite is true. The outcomes demonstrated by complexity models are in fact no more than computational derivations of the consequences of the premises of these models. So the surprises created by complex systems models only appear contingent; in fact they are generated by the properties of the constituents. So the surprises produced by complexity science are simulacra of contingency, not the real thing.

Second, what about heterogeneity? Does complexity science illustrate or explain the heterogeneity of social things? Not particularly. The heterogeneity of social things -- organizations, value systems, technical practices -- does not derive from complex system effects; it derives from the fact of individual actor interventions and contingent exogenous influences.

Finally, consider the feature of plasticity -- the fact that social entities can "morph" over time into substantially different structures and functions. Does complexity theory explain the feature of social plasticity? It does not. This is simply another consequence of the substrate of the social world itself: the fact that social structures and forces are constituted by the actors that make them up. This is not a systems characteristic, but rather a reflection of the looseness of social interaction. The linkages within a social system are weak and fragile, and the resulting structures can take many forms, and are subject to change over time.

The tools of simulation and modeling that complexity theorists are in the process of developing are valuable contributions, and they need to be included in the toolbox. However, they do not constitute the basis of a complete and comprehensive methodology for understanding society. Moreover, there are important examples of social phenomena that are not at all amenable to treatment with these tools.

This leads to a fairly obvious conclusion, and one that I believe complexity theorists would accept: that complexity theories and the models they have given rise to are a valuable contribution; but they are only a partial answer to the question, how does the social world work?

Designing and managing large technologies

What is involved in designing, implementing, coordinating, and managing the deployment of a large new technology system in a real social, political, and organizational environment? Here I am thinking of projects like the development of the SAGE early warning system, the Affordable Care Act, or the introduction of nuclear power into the civilian power industry.

Tom Hughes described several such projects in Rescuing Prometheus: Four Monumental Projects That Changed the Modern World. Here is how he describes his focus in that book:

Telling the story of this ongoing creation since 1945 carries us into a human-built world far more complex than that populated earlier by heroic inventors such as Thomas Edison and by firms such as the Ford Motor Company. Post-World War II cultural history of technology and science introduces us to system builders and the military-industrial-university complex. Our focus will be on massive research and development projects rather than on the invention and development of individual machines, devices, and processes. In short, we shall be dealing with collective creative endeavors that have produced the communications, information, transportation, and defense systems that structure our world and shape the way we live our lives. (3)

The emphasis here is on size, complexity, and multi-dimensionality. The projects that Hughes describes include the SAGE air defense system, the Atlas ICBM, Boston's Central Artery/Tunnel project, and the development of ARPANET. Here is an encapsulated description of the SAGE process:

The history of the SAGE Project contains a number of features that became commonplace in the development of large-scale technologies. Transdisciplinary committees, summer study groups, mission-oriented laboratories, government agencies, private corporations, and systems-engineering organizations were involved in the creation of SAGE. More than providing an example of system building from heterogeneous technical and organizational components, the project showed the world how a digital computer could function as a real-time information-processing center for a complex command and control system. SAGE demonstrated that computers could be more than arithmetic calculators, that they could function as automated control centers for industrial as well as military processes. (16)

Mega-projects like these require coordinated efforts in multiple areas -- technical and engineering challenges, business and financial issues, regulatory issues, and numerous other areas where innovation, discovery, and implementation are required. In order to be successful, the organization needs to make realistic judgments about questions for which there can be no certainty -- the future development of technology, the needs and preferences of future businesses and consumers, and the pricing structure that will exist for the goods and services of the industry in the future. And because circumstances change over time, the process needs to be able to adapt to important new elements in the planning environment.

There are multiple dimensions of projects like these. There is the problem of establishing the fundamental specifications of the project -- capacity, quality, functionality. There is the problem of coordinating the efforts of a very large team of geographically dispersed scientists and engineers, whose work is deployed across various parts of the problem. There is the problem of fitting the cost and scope of the project into the budgetary envelope that exists for it. And there is the problem of adapting to changing circumstances during the period of development and implementation -- new technology choices, new economic circumstances, significant changes in demand or social need for the product, large shifts in the costs of inputs into the technology. Obstacles in any of these diverse areas can lead to impairment or failure of the project.

Most of the cases mentioned here involve engineering projects sponsored by the government or the military. And the complexities of these cases are instructive. But there are equally complex cases that are implemented in a private corporate environment -- for example, the development of next-generation space vehicles by SpaceX. And the same issues of planning, coordination, and oversight arise in the private sector as well.

The most obvious thing to note in projects like these -- and many other contemporary projects of similar scope -- is that they require large teams of people with widely different areas of expertise and an ability to collaborate across disciplines. So a key part of leadership and management is to solve the problem of securing coordination around an overall plan across the numerous groups; updating plans in face of changing circumstances; and ensuring that the work products of the several groups are compatible with each other. Moreover, there is the perennial challenge of creating arrangements and incentives in the work environment -- laboratory, design office, budget division, logistics planning -- that stimulate the participants to high-level creativity and achievement.

This topic is of interest for practical reasons -- as a society we need to be confident in the effectiveness and responsiveness of the planning and development that goes into large projects like these. But it is also of interest for a deeper reason: the challenge of attributing rational planning and action to a very large and distributed organization at all. When an individual scientist or engineer leads a laboratory focused on a particular set of research problems, it is possible for that individual (with assistance from the program and lab managers hired for the effort) to keep the important scientific and logistical details in mind. It is an individual effort. But the projects described here are sufficiently complex that there is no individual leader who has the whole plan in mind. Instead, the "organizational intentionality" is embodied in the working committees, communications processes, and assessment mechanisms that have been established.

It is interesting to consider how students, both undergraduate and graduate, can come to have a better appreciation of the organizational challenges raised by large projects like these. Almost by definition, study of these problem areas in a traditional university curriculum proceeds from the point of view of a specialized discipline -- accounting, electrical engineering, environmental policy. But the view provided from a discipline is insufficient to give the student a rich understanding of the complexity of the real-world problems associated with projects like these. It is tempting to think that advanced courses for engineering and management students could be devised making extensive use of detailed case studies as well as simulation tools that would allow students to gain a more adequate understanding of what is needed to organize and implement a large new system. And interestingly enough, this is a place where the skills of humanists and social scientists are perhaps even more essential than the expertise of technology and management specialists. Historians and sociologists have a great deal to add to a student's understanding of these complex, messy processes.

(Martin Filler's review in the News York Review of Books of three recent books on the massive project in lower Manhattan to rebuild the World Trade Center illustrates some of the political and organizational challenges that stand in the way of large, complex projects; link.)

Systems management and the War on Poverty

One of the important developments in engineering and management thinking since World War II is the value of approaching large problems as systems rather than simply as a sum of separable components. Designing a ballpoint pen is very different from designing an aircraft or a fire control system; in the latter cases there are multiple functionalities and components that need to be incorporated, each associated with specific engineering and material disciplines. It was recognized during World War II that it is much more effective to treat the product and the design and manufacturing efforts as systems so that it is possible to conform components to synergistic and mutually supportive inter-relationships.

Agatha Hughes and Thomas Hughes organized a group of leading researchers to reflect upon the history of systems engineering and management, and the chief results are included in their 2000 volume, Systems, Experts, and Computers: The Systems Approach in Management and Engineering, World War II and After. The contributors include experts (and participants) in the history of the development of complex military systems during World War II -- for example, radar-controlled fire control systems for anti-aircraft use (David Mindell); experts like Donald MacKenzie on the incorporation of computing into the control of complex technologies (for example, the pathbreaking SABRE system for airline reservations); and experts on expertise such as Gabrielle Hecht, who provides an essay on post-war French technology management.

Here is how Hughes and Hughes describe the systems approach in their introduction to the volume:

Practitioners and proponents embrace a holistic vision. They focus on the interconnections among subsystems and components, taking special note of the interfaces among the various parts. What is significant is that system builders include heterogeneous components, such as mechanical, electrical, and organizational parts, in a single system. Organizational parts might be managerial structures, such as a military command, or political entities, such as a government bureau. Organizational components not only interact with technical ones but often reflect their characteristics. For instance, a management organization for presiding over the development of an intercontinental missile system might be divided into divisions that mirror the parts of the missile being designed. (2)

Hughes and Hughes provide a narrative that is intended to show the origins of systems engineering in operations research during World War II, and in the rapid development of highly complex technology systems needed for weaponry during the war (automated fire control, for example). In their telling of the story, the development of the digital computer during and after the war was a critical component of the development of the systems approach and the increasingly complex technologies and systems that the approach stewarded into existence. (See earlier posts on the development of ENIAC; link, link.) Much of this research took place within government and military organizations such as OSRD (Office of Scientific Research and Development); but private companies like RAND and MITRE soon emerged to take on contracts from military agencies for large-scale systems projects (5). And the research and development process itself came to be treated as a "system", with new software developed to support project planning and management. One important example was the PERT (Program Evaluation Review Technique) software system, developed by Booz, Allen & Hamilton (10).

Of particular interest here is the light the volume sheds on the efforts by the Johnson administration to apply systems thinking to the large social problems the country faced in the early 1960s, including especially poverty and urban problems (16) (link). David Jardini's essay "Out of the blue yonder: The transfer of systems thinking from the Pentagon to the Great Society, 1961-1965" explores this effort to transfer these systems methods to the social field. "[The chapter] argues that the construction and implementation of the Great Society social welfare programs and their analytical methods can be found at the core of Great Society policy making" (312). 

It emerges that a central political and policy disagreement that determined the course of events was a fundamental disagreement about centralization versus community involvement in social welfare policy. Policy leaders like Robert McNamara preferred to see the nation's social welfare policies to be managed and monitored centrally; affected communities, on the other hand, wanted to have greater control over the programs that would affect them. These disagreements converged on the question of the role of CAPs (Community Action Program) in the implementation and management of policy initiatives on the ground. Should CAPs serve as effective venues for local opinions and demands, or should they be sidelined in favor of a more top-down administrative organization?

The first CAP program guide, for example, suggested that local organizations provide "meaningful opportunities for residents, either as individuals or in groups, to protest or to propose additions to or changes in the ways in which a Community Action Program is being planned or undertaken." In fact, protest and confrontation were viewed by many CAP organizers as at least therapeutic means for the poor to vent their frustrations. (339)

But Johnson's administration was not interested in providing a venue for community advocacy and protest, and quickly sought to find ways of managing social welfare programs to reduce the level of activism they stimulated. The solution was the extension of the PPB (Planning-Programming-Budgeting) model from defense systems administration to the Great Society. But, as Jardini observes, this hierarchical system of control is poorly adapted to the problem of designing and administering programs that affect vast groups of people who can see its effects and can have very different ideas about the appropriateness of the policies being conveyed. "In this sense, the DOD is a poor model for the democratic ideal many Americans hold for their government institutions" (341).

This example illustrates an important tension that runs through many of the essays in the volume concerning the political significance of systems engineering and management. The volume gives support to the idea that systems management is an expert-driven and non-democratic way of organizing complicated human activities. What Robert McNamara brought to Ford Motor Company and the Department of Defense was a hierarchical, analytical, expert-driven system of management that sought to replace decentralized decision-makers with an orderly process driven from the top. For some purposes this may be a reasonably effective way of organizing a large effort involving thousands of agents. But for purposes like social reform it has a fatal flaw; it makes it almost impossible to create the level of buy-in at the local level that will be crucial for the success of a large project.

(I remember asking Tom Hughes in 1999 or so what he thought about the massive "Big Dig" project in Boston, then approaching completion and affecting many neighborhoods and thousands of residents. He commented that he felt that we should not judge the success of the project on the basis of whether it came in under budget; in fact, he suggested that this would show that the project designers and managers had not done enough to modify and adapt the project to gain support from the communities that the project affected.)

Von Neumann on the brain

image: representation of a mammalian brain neural network 

After World War II John von Neumann became interested in the central nervous system as a computing organ. Ironically, more was probably known about neuroanatomy than about advanced digital computing in the 1940s; that situation has reversed, of course. Now we know a great deal about calculating, recognizing, searching, and estimating in silicon; but relatively less about how these kinds of processes work in the setting of the central nervous system. At the time of his final illness von Neumann was preparing a series of Silliman Lectures at Yale University that focused on the parallels that exist between the digital computer and the brain; these were published posthumously as The Computer and the Brain (CB) in 1958. This topic also comes in for substantial discussion in Theory Of Self Reproducing Automata (TSRA) (edited and published posthumously by Arthur Burks in 1966). It is very interesting to see how von Neumann sought to analyze this problem on the basis of the kinds of information available to him in the 1950s.

Much of CB takes the form of a rapid summary of the state of knowledge about digital computing machines that existed in the 1950s, from Turing to ENIAC. Almost all computers today possess the "von Neumann" architecture along these lines.

Alan Turing provided some of the mathematical and logical foundations of modern digital computing (link). He hypothesized a very simple computing device that consisted of a tape of indefinite length, a  tape drive mechanism that permitted moving the tape forwards or backwards one space, and a read-write mechanism that could read the mark in a tape location or erase and re-write the mark in that location. Here is a diagram of a Turing machine:

(Fascinatingly, here is a photo of a working model of a Turing machine (link):)

Turing's fundamental theorem is that any function that is computable at all is computable on a Turing machine; so a Turing machine is a universal computing machine. The von Neumann architecture and the computing machines that it spawned -- ENIAC and its heirs -- are implementations of a universal computing machine. 

From the time of Frege it has been understood that mathematical operations can be built up as compounds of several primitive operations -- addition, subtraction, etc.; so, for example, multiplication can be defined in terms of a sequence of additions. Programming languages and libraries of subroutines take advantage of this basic logic: new functions are defined as series of more elementary operations embodied in machine states. As von Neumann puts the point in CB:

More specifically: any computing machine that is to solve a complex mathematical problem must be “programmed” for this task. This means that the complex operation of solving that problem must be replaced by a combination of the basic operations of the machine. Frequently it means something even more subtle: approximation of that operation—to any desired (prescribed) degree—by such combinations. (5)

Key questions about the capacities of a computing machine, either electro-mechanical or biological, have to do with estimating its dimensionality: how much space does it occupy, how much energy does it consume, and how much time does it take to complete a given calculation? And this is where von Neumann's analysis took its origin. Von Neumann sought to arrive at realistic estimates of the size and functionality of the components of these two kinds of computation machines. The differences in scale are enormous, whether we consider speed, volume, or energy consumption. Fundamentally, neurons are more numerous by orders of magnitude (10^10 versus 10^4); slower by orders of magnitude (5 msec vs. 10^-3 msec); less energy-intensive by orders of magnitude (10^-3 ergs vs.10^2 ergs); and computationally less precise by orders of magnitude. (Essentially he estimates that a neural circuit, either analog or digital, is capable of precision of only about 1%.) And yet von Neumann concludes that brains accomplish computational problems faster than digital computers because of their massively parallel structure -- in spite of the comparative slowness of the individual elements of computation (neurons). This implies that the brain embodies a structurally different architecture than sequential digital computing embodied in the von Neumann model.

Von Neumann takes the fundamental operator of the brain to be the neuron, and he represents the neuron as a digital device (in spite of its evident analog electrochemical properties). A neuron transmits a pulse. "The nervous pulses can clearly be viewed as (two-valued) markers.... The absence of a pulse then represents one value (say, the binary digit 0), and the presence of one represents the other (say, the binary digit 1)" (42). "The nervous system has a prima facie digital character" (44).

In their introduction to the second edition of CB the Churchlands summarize von Neumann's conclusion somewhat differently by emphasizing the importance of the analog features of the brain: "If the brain is a digital computer with a von Neumann architecture, it is doomed to be a computational tortoise by comparison... [But] the brain is neither a tortoise nor a dunce after all, for it was never a serial, digital machine to begin with: it is a massively parallel analog machine" (kl 397). However, it appears to me that they overstate the importance of analog neural features in von Neumann's account. Certainly vN acknowledges the analog electro-chemical features of neural activity; but I don't find him making a strong statement in this book to the effect that analog features contribute to the better-than-expected computational performance of the brain. This seems to correspond more to a view of the Churchlands than to von Neumann's analysis in the 1950s. Here is their view as expressed in "Could a Machine Think?" in Scientific American in 1990:

First, nervous systems are parallel machines, in the sense that signals are processed in millions of different pathways simultaneously. The retina, for example, presents its complex input to the brain not in chunks of eight, 16 or 32 elements, as in a desktop computer, but rather in the form of almost a million distinct signal elements arriving simultaneously at the target of the optic nerve (the lateral geniculate nucleus), there to be processed collectively, simultaneously and in one fell swoop. Second, the brain’s basic processing unit, the neuron, is comparatively simple. Furthermore, its response to incoming signals is analog, not digital, inasmuch as its output spiking frequency varies continuously with its input signals. Third, in the brain axons projecting from one neuronal population to another are often matched by axons returning from their target population. These descending or recurrent projections allow the brain to modulate the character of its sensory processing. (link, 35)

In considering the brain von Neumann reached several fundamental observations. First, the enormous neural network of the central nervous system is itself a universal computing machine. Von Neumann worked on the assumption that the CNS could be "programmed" to represent the fundamental operations of arithmetic and logic; and therefore it has all the power of a universal computational machine. But second, von Neumann believes his analysis demonstrates that its architecture is fundamentally different from the standard von Neumann architecture. This observation is the more fundamental. It derives from von Neumann's estimates of the base speed rate of calculation available to neurons in comparison to vacuum tubes; a von Neumann machine with components of this time scale would take eons to complete the calculations that the brain performs routinely. And so this underlines the importance of the massively parallel computing that is accomplished by the biological neural network. Ironically, however, it has proven challenging to emulate massively parallel neural nets in digital computing environments; here is an interesting technical report by Paul Fox that identifies communication bandwidth as being the primary limiting factor for such emulations (link). 

(Tsutomu Miki explores some of these issues in Brainware : Bio-Inspired Architecture and Its Hardware Implementation.)

Computational models for social phenomena

There is a very lively body of work emerging in the intersection between computational mathematics and various fields of the social sciences. This emerging synergy between advanced computational mathematics and the social sciences is possible, in part, because of the way that social phenomena emerge from the actions and thoughts of individual actors in relationship to each other. This is what allows us to join mathematics to methodology and explanation. Essentially we can think of the upward strut of Coleman’s boat — the part of the story that has to do with the “aggregation dynamics” of a set of actors — and can try to create models that can serve to simulate the effects of these actions and interactions.

source: Hedstrom and Ylikoski (2010) "Causal Mechanisms in the Social Sciences" (link)

 

Here is an interesting example in the form of a research paper by Rahul Narain and colleagues on the topic of modeling crowd behavior ("Aggregate Dynamics for Dense Crowd Simulation", link). Here is their abstract:

Large dense crowds show aggregate behavior with reduced individual freedom of movement. We present a novel, scalable approach for simulating such crowds, using a dual representation both as discrete agents and as a single continuous system. In the continuous setting, we introduce a novel variational constraint called unilateral incompressibility, to model the large-scale behavior of the crowd, and accelerate inter-agent collision avoidance in dense scenarios. This approach makes it possible to simulate very large, dense crowds composed of up to a hundred thousand agents at near- interactive rates on desktop computers.

Federico Bianchi takes up this intersection between computational mathematics and social behavior in a useful short paper called "From Micro to Macro and Back Again: Agent-based Models for Sociology" (link). His paper focuses on one class of computational models, the domain of agent-based models. Here is how he describes this group of approaches to social explanation:

An Agent-Based Model (ABM) is a computational method which enables to study a social phenomenon by representing a set of agents acting upon micro-level behavioural rules and interacting within environmental macro-level (spatial, structural, or institutional) constraints. Agent-Based Social Simulation (ABSS) gives social scientists the possibility to test formal models of social phenomena, generating a virtual representation of the model in silico through computer programming, simulating its systemic evolution over time and comparing it with the observed empirical phenomenon. (1) 

 And here is how he characterizes the role of what I called "aggregation dynamics" above:

Solving the complexity by dissecting the macro-level facts to its micro-level components and reconstructing the mechanism through which interacting actors produce a macro-level social outcome. In other words, reconstructing the micro-macro link from interacting actors to supervenient macrosociological facts. (2)

Or in other words, the task of analysis is to provide a testable model that can account for the way the behaviors and interactions at the individual level can aggregate to the observed patterns at the macro level.

Another more extensive example of work in this area is Gianluca Manzo, Analytical Sociology: Actions and Networks. Manzo's volume proceeds from the perspective of analytical sociology and agent-based models. Manzo provides a very useful introduction to the approach, and Peter Hedstrom and Petri Ylikoski extend the introduction to the field with a chapter examining the role of rational-choice theory within this approach. The remainder of the volume takes the form of essays by more than a dozen sociologists who have used the approach to probe and explain specific kinds of social phenomena.

Manzo provides an account of explanation that highlights the importance of "generating" the phenomena to be explained. Here are several principles of methodology on this topic:

• P4: in order to formulate the "generative model," provide a realistic description of the relevant micro-level entities (P4a) and activities (P4b) assumed to be at work, as well as of the structural interdependencies (P4c) in which these entities are embedded and their  activities unfold;
• P5: in order rigorously to assess the internal consistency of the "generative model" and to determine its high-level consequences, translate the "generative model" into an agent-based computational model;
• P6: in order to assess the generative sufficiency of the mechanisms postulated, compare the agent-based computational model's high-level consequences with the empirical description of the facts to be explained (9)

So agent-based modeling simulations are a crucial part of Manzo's understanding of the logic of analytical sociology. As agent-based modelers sometimes put the point, "you haven't explained a phenomenon until you've shown how it works on the basis of a detailed ABM." But the ABM is not the sole focus of sociological research, on Manzo's approach. Rather, Manzo points out that there are distinct sets of questions that need to be investigated: how do the actors make their choices? What are the structural constraints within which the actors exist? What kinds of interactions and relations exist among the actors? Answers to all these kinds of question are needed if we are to be able to design realistic and illuminating agent-based models of concrete phenomena.

Here is Manzo's summary table of the research cycle (8). And he suggests that each segment of this representation warrants a specific kind of analysis and simulation.

This elaborate diagram indicates that there are different locations within a complex social phenomenon where different kinds of analysis and models are needed. (In this respect the approach Manzo presents parallels the idea of structuring research methodology around the zones of activity singled out by the idea of methodological localism; link.) This is methodologically useful, because it emphasizes to the researcher that there are quite a few different kinds of questions that need to be addressed in order to successfully explain a give domain of phenomena.

The content-specific essays in the volume focus on one or another of the elements of this description of methodology. For example, Per-Olof Wikstrom offers a "situational action theory" account of criminal behavior; this definition of research focuses on the "Logics of Action" principle 4b.

People commit acts of crime because they perceive and choose (habitually or after some deliberation) a particular kind of act of crime as an action alternative in response to a specific motivation (a temptation or a provocation). People are the source of their actions but the causes of their actions are situational. (75)

SAT proposes that people with a weak law-relevant personal morality and weak ability to exercise self-control are more likely to engage in acts of crime because they are more likely to see and choose crime as an option. (87)

Wikstrom attempts to apply these ideas by using a causal model to reproduce crime hotspots based on situational factors (90).

The contribution of Gonzalez-Bailon et al, "Online networks and the diffusion of protest," focuses on the "Structural Interdependency" principle 4c.

One of the programmatic aims of analytical sociology is to uncover the individual-level mechanisms that generate aggregated patterns of behaviour.... The connection between these two levels of analysis, often referred to as the micro-macro link, is characterised by the complexity and nonlinearity that arises from interdependence; that is, from the influence that actors exert on each other when taking a course of action. (263)

Their contribution attempts to provide a basis for capturing the processes of diffusion that are common to a wide variety of types of social behavior, based on formal analysis of interpersonal networks.

Networks play a key role in diffusion processes because they facilitate threshold activation at the local level. Individual actors are not always able to monitor accurate the behavior of everyone else (as global thresholds assume) or they might be more responsive to a small group of people, represented in their personal networks. (271)

They demonstrate that the structure of the local network matters for the diffusion of an action and the activation of individual actors.

In short, Analytical Sociology: Actions and Networks illustrates a number of points of intersection between computational mathematics, simulation systems, and concrete sociological research. This is a very useful effort as social scientists attempt to bring more complex modeling tools to bear on concrete social phenomena.

International relations and complexity theory

Hilton Root has published some very interesting ideas about systems thinking in international relations theory in Dynamics among Nations: The Evolution of Legitimacy and Development in Modern States. Here he offers an approach to social, political, and economic change through a set of ideas that are not yet strongly integrated into IR theory — the perspective of complexity theory, worked out in a clear and useable form.

The three sources of theoretical argument which he introduces -- complexity theory, social network theory, and evolutionary ecology -- represent a significant innovation in comparative history. The novel approach Root takes consists of three large ideas: that social systems at all levels display “adaptive complexity”; that the structure of the social networks (governance systems, information systems, economic inter-dependencies) that are embedded in a specific society have important and unexpected consequences for the behavior of the system; and that complex social developments have much in common with “landscape ecology”, by which he means that there are multiple next steps that can be taken at any point leading to an improvement of performance.

His fundamental claim is that communities, states, and international systems need to be understood as dynamic systems with emergent properties. A society is not simply the linear sum of the behaviors of its component systems.

The system of international relations, like most complex ecosystems, such as the nervous system or a rain forest, is yielding to its rules of complexity. In complex systems, a central administrator rarely guides the collective behaviors that characterize development processes. The system itself has a collective behavior that depends on all its parts. Rather than convergence toward a dominant model, or “global optimum,” the interactive dynamics are coevolutionary; their interactions result in reciprocal and evolving change. (2)

One consequence of these ideas is that international relations and economic and political development processes show substantial path dependency and contingency. Another consequence is that some leading metaphors for large-scale historical change are implausible and misleading: in particular, modernization theory, “uniqueness of the West,” and “end of history.” Finally, Root argues that we should expect substantial variation in the strategies and structures that nations choose, given their own geopolitical environments.

Competition in highly interdependent global environments produces far greater local variation and diversity of structures and strategies than modernization theory ever anticipated. (3)

The book uses numerous episodes from the political, military, and economic histories of Europe and Asia to illustrate and validate the approach he takes. As a particularly interesting example of this, Root interprets Napoleon’s decision to invade Russia, not as folly, but as an intuition of the nodal character of the traditional European state system (126 ff.). He also makes repeated use of periods in Chinese imperial history to illustrate his notion that system dynamics and the structure of the governance network create very powerful obstacles to innovation and change.

So what does Root mean by “complexity”? His central concept is that of a “complex interactive adaptive system” (CIAS) within a heterogeneous environment. Here is a useful description of international relations through the lens of CIAS theory.

A network is comprised of agents. The agents interact according to shared and evolving rules of behavior that in turn define the larger environment or system. That behavior generates continuous feedback loops that enable agents to learn and to adjust their behaviors to others’ actions, thereby re-creating the system in which they operate. Complex adaptive systems are created by interactions and communications of self-adjusting agents. Continuous “feedback” motivates agents to re-evaluate their positions. Because agents are constantly reacting to other agents’ behaviors, nothing in the environment is ever fixed or finite. In order to fully understand the impacts of these agents, their behaviors must be understood as they interact with the broader system. (16)

A key analytical idea the author brings forward repeatedly is the notion of “co-evolution”. This concept captures one important aspect of a complex interactive adaptive system. CIAS’s show two types of unpredictability. First, the mutual interactions of the parts lead to “chaotic” courses of development of the system, as A, B, and C interact to produce unexpected outcome D. But second, the “adaptive” part introduces another kind of indeterminacy, as organisms, actors, and institutions change their characteristics in face of changes in the environment. So the properties of A, B, and C are not fixed over time; rather, selection and purposive adaptation lead to organisms and actors who respond differently over time to ecological opportunities and threats. 

Features of uncertainty, time framing, rule change, and novel behavior all contribute to a set of system characteristics: unpredictability, path dependency, and sensitivity to initial conditions. And Root believes that these factors have important implications about the feasibility of reducibility or micro- to macro- reconstruction:

When a state’s interactions shift from being locally based to being regionally or nationally based, its behaviors change across the network and the greater system. Thus a general theory of the system cannot be deduced from the properties of its constituent parts, just as the universe cannot be reconstructed from the fundamental laws of physics. (31)

Root's treatment of “New Institutional Economics” in Chapter 5 is important for several reasons. Most important, he demonstrates the harm that comes from incorporating a questionable theory of change into a comprehensive agenda for policy. The guiding idea of “creating institutions of good governance” as a panacea for slow economic growth and widespread poverty led policy makers to ignore other important causal factors, including locally rational but myopic strategies pursued by sub-actors. Root seems to agree with Dani Rodrik in concluding that NIC is limited when it comes to serving as a guide for positive policy design:

Assessing the legacy of new institutional economics, Harvard economist Dani Rodrik concludes that beyond “a very aggregate level of generality,” these ideas do not provide much policy guidance. (81)

Instead of looking for a general theory that can be used by centralized planning ministries to guide their economic and social policies, Root favors a more evolutionary approach: allow for a diversity of development experiments at the middle level of society, and then favor those experiments that appear to have the best results.

Chinese planners never attained the celebrity status of their Indian peers, but by trying multiple paths and starting with smaller interventions from the top, they found a better way to determine what worked. After Deng declared the opening of the Chinese economy, he instituted a multi-level process that facilitated both change and stability, and strengthened social organization and social learning through local experimentation. (108-109)

(Contrast this with the “single experiment” approach associated with land collectivization in the 1950s, resulting in massive agricultural failure and famine during the Great Leap Forward.)

Root's treatment of Imperial China’s history is intriguing but controvertible. His central premise is that China’s Imperial system was a hierarchical network of control, and systems like this are substantially less resilient and open to change than multi-nodal networks. The interpretation is reminiscent of the theory of Oriental despotism: an all-powerful imperial system suppressed both challengers and change-agents. But contemporary China historians would probably give the Imperial system more credit in terms of its degree of flexibility in face of challenges. Take peasant uprisings. The state was generally successful in its response to large peasant rebellions, even if the military response was often flat-footed. The Taiping Rebellion is an example that probably supports the author’s interpretation best, since it was local militias organized and funded by local gentry which were most successful in opposing the Taipings. But China’s history is littered with hundreds of peasant and ethnic uprisings, and its military eventually prevailed in most of them.

One way of reading Root’s book is as a guidebook for administrators in a time of complexity. Root correctly emphasizes the difficulty or impossibility of “solving” a set of social and political problems simultaneously, and the parallel difficulty of making confident predictions about medium- or long-term consequences of various policy interventions. Second best, in his account, is an evolutionary approach: try a diversity of approaches, and cautiously increase the volume of those approaches that seem to work best. But even this approach is uncertain; evolutionary processes lead to dead-ends that are unforeseen in earlier stages of the process.

(See this post about decision-making under conditions of deep uncertainty; link. And here is a series of earlier posts about social complexity; link.)

The heterogeneous social?

image: screenshot from video, "A Bird Ballet"

I've argued in several places that we need to think of the social world as being radically heterogeneous (link, link, link). There are multiple processes, things, structures, and temporalities at work, and what we perceive at a moment in time in the social world is a complex composite of these various things. The social world is not a unified system; it is not a manifestation of a unified underlying process; it is not a unity at all.

What does this claim about the social world mean in concrete terms? And what are the implications for the social sciences? Consider a few examples of complex social wholes:

• the industrial revolution, 1700-1850
• the rise of Al Qaeda, 1970-2001
• urbanization in China, 1600-1700
• Chicago as a functioning city, 2000
• the University of Illinois, 1971
• being Muslim in Toronto, 1990 

These examples are themselves heterogeneous. Some are extended historical processes; others are synchronic sets of social facts; others are institutions and social environments at a time; yet others are states of social identities at a time. But the fact about heterogeneity that I want to focus on here is internal: for each social phenomenon, there are heterogeneous components and sub-processes that make it up and that generally have their own dynamics and properties.

First, where is the heterogeneity in these examples?

The industrial revolution is not one thing; it is rather a confluence of developments in technology, markets, habitation, ideology, labor practices, scientific institutions, natural resources, and numerous other social features that change over time. And the outcomes of "industrial revolution" are not uniform over regions, nations, sectors, or industries. Different parts of Britain had different experiences; and these experiences and outcomes are in turn different from those in Sweden or Italy.

Likewise, early-modern urbanization of Chinese cities is a the result of a complex ensemble of processes. We can summarize the outcome by a measure of the percentage of people living in cities greater than 100,000 at a certain moment in time. But the causes, processes, environmental factors, and institutions through which this transformation took place were highly diverse; and the cities that resulted were diverse as well. (G. William Skinner charts out much of this diversity in a number of works; The City in Late Imperial China.)

Or take Chicago in 2000. The social whole is a composite of population, institutions, political processes, demographic transitions, transportation networks, employment systems, and policing practices -- and many other factors I haven't mentioned. And if we were to ask a question along these lines -- why did Chicago come to function in 2000 in the fashion that it did? -- we would have to consider all of these processes and their composite effects, and their interactions with each other. There is no single answer to the question, "what is Chicago and how does it work?".

Being Muslim at a time and place is likewise deeply heterogeneous. Individuals, families, sub-groups, and institutions differ -- from Iowa to Ontario, and within communities and across mosques. Individuals differ in ways that are both personal and institutional. So there is no single identity that is "Muslim in Toronto"; rather, there is an ensemble of people, groups, and social organizations which in the composite represent "the many identities of Muslims in Toronto."

In fact, it seems to me that heterogeneity comes into each of these examples in a variety of ways. There are:

• multiple causes at work
• multiple expressions of ethnic / cultural identity
• multiple purposes and understandings on the parts of participants
• multiple sub-institutions with different profiles and dynamics
• multiple outcomes or macro-characteristics that are denoted by the term

So the constitution and dynamics of social phenomena reflect diverse kinds of things and processes.

So where does "science" come into this picture? Is it possible to have a scientific understanding of a heterogeneous phenomenon?

Here is one possible strategy. We might hope that the sub-components of heterogeneous entities might have separable dynamics of development; so even though the city simpliciter does not have an inherent dynamic of development or functioning, its subsystems do. In this case we might say that a scientific analysis of the whole involves a separate scientific theory of the components and a synthetic effort to show how they interact.

But this approach is perhaps too generous to the power of analysis; it seems to presuppose that we can disassemble a complex and heterogeneous whole into a discrete set of reasonably homogeneous components, each of which can be treated scientifically and separately. The thesis above, though, was fairly comprehensive: "all social phenomena are heterogeneous". So that seems to imply that the results of analysis lead us to a set of components that are themselves heterogeneous -- a heterogeneity regress! And this paradoxical conclusion actually seems to be true in a very practical sense: when we disaggregate "Chicago" into "political institutions," "policing institutions," "economic institutions / market system", and the like -- we again encounter social units that have internal variation and heterogeneity.

Could we at least argue that analysis reduces complexity to a certain extent, and that the components are more amenable to scientific and causal theorizing than the whole? This more modest claim does seem to be defensible. Take the processes underlying "industrial revolution". It is possible to offer a reasonably rigorous study of the development of scientific knowledge and the institutions through which knowledge is created and disseminated, in ways that are less complex that the whole with which we began. Likewise, we can offer specialized study of the "making of the English working class" that includes some of the factors that influenced labor and politics during the period -- thereby making a contribution to a better understanding of the complex whole, industrial revolution.

In an odd way this line of thought seems to bring us back to one of the oldest debates in the history of philosophy going back to the pre-Socratic philosophers: does "nature" have a "nature"? The atomists believed that the complexity of the observed world depended ultimately on the simple properties of the components; whereas philosophers like Heraclitus maintained that nature consisted of "flux" all the way down.

(The  video mentioned at the top, "A Bird Ballet," is beautiful and surprising. But I'm not certain that it fully illustrates the point I'm making about the social world. The ensemble of starlings depicted here shows a startling reality of shifting shapes and motions over time. The viewer is led to ask, how did this ensemble of thousands of organisms come to have this graceful and shifting dynamic?" So far it is a good analogy to the social. But an animal behavior specialist is likely to be able to give us a pretty simple explanation of how the individual-level flight behavior interacts across birds in flight, and results in the swarming behavior documented here. In this respect the swarm is simpler than the "heterogeneity all the way down" picture that I'm putting forward for complex social phenomena. Still, it is a powerful example of "wholes" that are less unified than they first appear.)

Simon on complexity

Herbert Simon's The Sciences of the Artificial - 3rd Edition provided an alternative model for thinking about society. We can think of social institutions as partially designed and selected for their organizational properties; so they are different from proteins and planetary systems.  Simon is also an important contributor to the study of complexity. So his new chapter in the 1996 edition of the book, "Alternative Views of Complexity," is worth reading carefully. Here is how he motivates this new chapter in SA:

The preceding chapters of this book have discussed several kinds of artificial systems. The examples we have examined -- in particular, economic systems, the business firm, the human mind, sophisticated engineering designs, and social plans -- range from the moderately to the exceedingly complex (not necessarily in the order in which I have just listed them). These final two chapters address the topic of complexity more generally, to see what light it casts on the structure and operation of these and other large systems that are prominent in our world today. (169)

It turns out that there isn't much new in the 1996 chapter, however. In fact, most of its content is taken from his pathbreaking 1962 article, "The Architecture of Complexity" (link). The new chapter 7 and renumbered chapter 8 largely incorporate the content and sometimes the language of the 1962 article. And this is interesting, because it implies that Simon's primary ideas about reduction, composition, and inter-level interactions were largely already formed in 1962.

There are a few ideas and themes that are new to the 1996 version. One is a more specific periodization of thinking about complexity theory in the twentieth century.  The 1996 version identifies three phases of theorizing about complexity and "whole systems".

1. Biological emergence theory (post World War I)
2. Cybernetics and systems theory (post World War II)
3. Contemporary complexity theory (post 1960s)

Simon is skeptical about the tendency towards irreducible holism that was associated with the earlier two phases of thinking in both versions; in the 1996 chapter he favors a "weak" interpretation of emergence: a commitment to ...

... reductionism in principle even though it is not easy (often not even computationally feasible) to infer rigorously the properties of the whole from knowledge of the properties of the parts. In this pragmatic way, we can build nearly independent theories for each successive level of complexity, but at the same time, build bridging theories that show how each higher level can be accounted for in terms of the elements and relations of the next level down. (172)

This "pragmatic holism" is already contained in the 1962 version (link). So this doesn't represent new ground in 1996. But Simon's use of this idea to criticize several false starts in the field of complexity research is valuable.

Simon finds some of the central concepts of the third phase to be more promising for the study of social phenomena. The mathematics and physics of chaotic behavior (where simple low-level processes can aggregate to wildly variant higher-level outcomes), simulations of evolution through computational models (genetic algorithms), and the exploration of cellular autonoma (the game of life) all come in for favorable comments. (The Lorenz attractor illustrated here is a common example of chaotic behavior.)

One idea that is not contained in the 1962 version is that of causal non-linearity. Non-linearity is a problem for the "near decomposability" view that Simon wanted to take of complexity in the 1962 version, because it casts doubt on the ability to disentangle causal influences deriving from inter-connected subsystems. Small differences in initial conditions can lead to large differences in outcome. This is a key aspect of chaos theory and the varieties of turbulent phenomena that provide the best examples of chaotic systems. And this casts some doubt on one of the central conclusions of the 1962 paper:

The fact, then, that many complex systems have a nearly decomposable, hierarchic structure is a major facilitating factor enabling us to understand, to describe, and event to "see" such systems and their parts. Or perhaps the proposition should be put the other way round. If there are important systems in the world that are complex without being hierarchic, they may to a considerable extent escape our observation and our understanding. (477)

This is a decidedly pre-chaos understanding of the nature of complex systems. I have the impression that many contemporary complexity theorists would reject the idea that social processes are commonly the result of "nearly decomposable, hierarchic structures". So it is a genuine change for the mathematics of chaos theory to be included in the 1996 version. Complexity research has moved forward since 1962, and Simon recognizes this in the 1996 chapter.

What we don't find here is any discussion of whether actual social processes and systems display chaotic behavior in this well defined sense. And we don't see Simon shifting his position on "nearly decomposable" systems.

Are there examples of social processes and phenomena that display chaotic characteristics over time? Take the occurrence of massive street demonstrations as an example; are there aspects of chaos in the technical sense involved in the outbreak of street mobilization? Do small, apparently random events have large effects on the eventual outcome?

It would appear that this is the case when we look at the cases of uprising and passivity in different cities during the Arab Spring of 2011. Some social scientists have tried to understand the likelihood of uprising as an increasing function of economic crisis, regime weakness, and regime brutality. This implies a linear assumption about the causal role of these three forces. But it seems plausible to speculate that random events like a broken phone chain, an Internet outage, or the defection of a key leader could push the process of mobilization into a different direction. Moreover, it seems that contemporary research on social complexity pays a lot of attention to non-linearity, path-dependency, and sequential processes of social mobilization -- leaving a lot of room for the kinds of turbulent effects that are observed in traffic flow, storm generation, and water dripping from a leaking tap. This is the kind of work that is described in Scott Page and John Miller, Complex Adaptive Systems: An Introduction to Computational Models of Social Life.

So oddly enough, it seems that one could fairly say that Simon's views of social complexity -- as expressed in the 1996 third edition of  The Sciences of the Artificial as well as in his groundbreaking "Architecture of Complexity" in 1962 -- are significantly incomplete, given the way that complexity theorists are now thinking about social processes. Simon did not incorporate the guiding assumptions of "complex adaptive systems theory" into his own thinking, and remained convinced of the adequacy of the ideas of hierarchical systems and nearly decomposable systems as late at 1996.  His own approach to social complexity remains a phase two approach, not a phase three approach.

(The graph at the top of this post is offered as an interpretation of a highly path-dependent social process. The reader is asked to consider each path as a hypothetical development from a common origin, with small stochastic variations in the situation occurring over time. Imagine the starting position is "large city, economic hardship, weak state, lots of repression", time is the x axis, and the y axis measures civil unrest. Some of those variations push the path towards a high outcome (blue), and some towards a low outcome (magenta). The great majority of outcomes fall within a short distance of the starting position. So the most likely outcome is "not much change", but there are unlikely but diametrically different outcomes possible as well.)

Simulating social mechanisms

A key premise of complexity theory is that a population of units has "emergent" properties that result from the interactions of units with dynamic characteristics. Call these units "agents".  The "agent" part of the description refers to the fact that the elements (persons) are self-directed units.  Social ensembles are referred to as "complex adaptive systems" -- systems in which outcomes are the result of complex interactions among the units AND in which the units themselves modify their behavior as a result of prior history.

Scott Page's Complex Adaptive Systems: An Introduction to Computational Models of Social Life provides an excellent introduction. Here is how Page describes an adaptive social system:

Adaptive social systems are composed of interacting, thoughtful (but perhaps not brilliant) agents. It would be difficult to date the exact moment that such systems first arose on our planet -- perhaps it was when early single-celled organisms began to compete with one another for resources.... What it takes to move from an adaptive system to a complex adaptive system is an open question and one that can engender endless debate. At the most basic level, the field of complex systems challenges the notion that by perfectly understanding the behavior of each component part of a system we will then understand the system as a whole. (kl 151)

Herbert Simon added a new chapter on complexity to the third edition of The Sciences of the Artificial - 3rd Edition in 1996.

By adopting this weak interpretation of emergence, we can adhere (and I will adhere) to reductionism in principle even though it is not easy (often not even computationally feasible) to infer rigorously the properties of the whole from knowledge of the properties of the parts. In this pragmatic way, we can build nearly independent theories for each successive level of complexity, but at the same time, build bridging theories that show how each higher level can be accounted for in terms of the elements and relations of the next level down. (172).

This formulation amounts to the claim of what I referred earlier to as "relative explanatory autonomy"; link. It is a further articulation of Simon's view of "pragmatic holism" first expressed in 1962 (link).

So how would agent-based models (ABM) be applied to mechanical systems? Mechanisms are not intentional units. They are not "thoughtful", in Page's terms. In the most abstract version, a mechanism is an input-output relation, perhaps with governing conditions and with probabilistic outcomes -- perhaps something like this:

In this diagram A, B, and D are jointly sufficient for the working of the mechanism, and C is a "blocking condition" for the mechanism. When A,B,C,D are configured as represented the mechanism then does its work, leading with probability PROB to R and the rest of the time to S.

So how do we get complexity, emergence, or unpredictability out of a mechanical system consisting of a group of separate mechanisms? If mechanisms are determinate and exact, then it would seem that a mechanical system should not display "complexity" in Simon's sense; we should be able to compute the state of the system in the future given the starting conditions.

There seem to be several key factors that create indeterminacy or emergence within complex systems. One is the fact of causal interdependency, where the state of one mechanism influences the state of another mechanism which is itself a precursor to the first mechanism.  This is the issue of feedback loops or "coupled" causal processes. Second is non-linearity: small differences in input conditions sometimes bring about large differences in outputs. Whenever an outcome is subject to a threshold effect, we will observe this feature; small changes short of the threshold make no change in the output, whereas small changes at the threshold bring about large changes. And third is the adaptability of the agent itself.  If the agent changes behavioral characteristics in response to earlier experience (through intention, evolution, or some other mechanism) then we can expect outcomes that surprise us, relative to similar earlier sequences. And in fact, mechanisms display features of each of these characteristics. They are generally probabilistic, they are often non-linear, they are sensitive to initial conditions, and at least sometimes they "evolve" over time.

So here is an interesting question: how do these considerations play into the topic of understanding social outcomes on the basis of an analysis of underlying social mechanisms? Assume we have a theory of organizations that involves a number of lesser institutional mechanisms that affect the behavior of the organization. Is it possible to develop an agent-based model of the organization in which the institutional mechanisms are the units? Are meso-level theories of organizations and institutions amenable to implementation within ABM simulation techniques?

Here is a Google Talk by Adrien Treuille on "Modeling and Control of Complex Dynamics".

The talk provides an interesting analysis of "crowd behavior" based on a new way of representing a crowd.