Explorations in Emergence: A Multitiered Research Proposal
Draft proposal
Timothy Burke • Swarthmore College • Fall 2003

(this proposal first appeared at http://www.swarthmore.edu/SocSci/tburke1/emergenceproposal.html)


The topic of emergence is one branch of the sprawling body of contemporary research associated with the study of complexity, a general area of academic inquiry that grew rapidly in the 1980s.

Emergent systems are defined by several characteristics. They begin with simple initial conditions or phenomena and yield steadily over time to complex organization or structures. There is no top-down or controlling agent or leader who shapes the progressive movement from simplicity to complexity. The complexity or forms of order which come from initial simplicities are not visible or apparent within those simplicities at the outset, and therefore the forms of structure or systematicity which emerge over time often subjectively appear to observers to be surprising or unexpected.

Emergent phenomena are shaped by iterative feedback loops, incorporating both positive and negative feedback, and typically involve autonomous agents acting on a background environment using a set of specified, simple rules that are associated with those agents.
A classic demonstration of emergent systems can be found in Mitch Resnick’s computer programs StarLogo and NetLogo, which use a very simple programming language to create simulations governed by emergent principles. One NetLogo simulation called Termites has two types of colored pixels on a black field of pixels, one of them designated as a “termite”, the other designated as a “stick”. The termites are the agents, the black pixels and the sticks are the environment (the black pixels being “empty space”). The termites have three basic rules: 1) each time step, they are to move one pixel in a random direction. 2) If they are proximate to a “stick”, they should “pick it up” (and change color to indicate they are carrying a stick). 3) If they are carrying a stick and they bump into another stick, they should drop the stick they’re carrying in an empty space in a random direction proximate to their position. With any random distribution of termites and sticks at the outset, the termites will in a fairly short time build a large circular pile of sticks, and that pile will from that point forward be a stable and largely unchanging structure within the simulation environment. Once built, it never gets taken apart. In retrospect, it is possible to work out why this happens, but there is no instruction to build a pile, no centrally coordinating agent that understands what a circular pile is, no design or “top-down” process.

Emergence has so far largely appealed to natural scientists looking to explain the evolution of complex systems that appear to lack central control or planning, including immune systems, the brain, and the behavior of social insects. Its appeal to computer scientists and engineers as both a practical and theoretical tool is also substantial. While some social scientists and scholars in the humanities have recognized the relevance and importance of emergence and related topics, explorations of the subject in these disciplinary areas remain fairly tentative and preliminary.

After attending the meetings of the Emergent Systems Working Group at Bryn Mawr College, and being part of a Mellon-funded faculty seminar at Swarthmore, Bryn Mawr and Haverford dedicated to the same topic this academic year, I am strongly convinced that the idea of emergence is not only important to a number of my scholarly interests but actually helps to connect them in ways that nothing else could. However, like a number of my colleagues in these two groups, I am struggling to see how to go beyond using emergence simply as a metaphor or trope in historical analysis and social theory, something that is valuable enough in its own right, but also ultimately something of a dead end.

I would like to use the Mellon New Directions grant to fund five divergent research “modules” , each of which is a limited thought-experiment of some kind designed to apply the idea of emergence and emergent systems to inquiry in the humanities and the social sciences, and to explore in the process the possibilities and limitations of emergence and complexity as ideas. Each module of this larger project is an attempt to empirically explore some of the theoretical preoccupations that have marked my recent scholarly writing in history and cultural studies through the lens of emergent systems and emergent phenomenon. I hope to use each of these modules both as conceptual explorations designed to illuminate difficult theoretical and philosophical questions about history and causation and to work towards practical applications that make use of the idea of emergence.
Several of these modules use Resnick’s NetLogo in the same spirit that he himself has used it, largely as a way to stimulate what he calls “explorations of microworlds”, where the reaction of users of the program are in some ways deemed more interesting than the program itself.

This project realizes the objective of the Mellon New Directions grant in two ways. First, it will help me continue and deepen my involvement with a wide range of scholarship and research in the areas of emergence, complexity and network theory which take me quite far away from my established expertise in African cultural history and American cultural studies—though I expect my new work may ultimately return major new dividends in both of those areas of study. Second, several of the modules in this overall project will require me to learn entirely new methodological practices and research norms, and connect me to disciplinary and research communities with which I have had little or no previous contact, most crucially computer science and psychology.

Module 1) Human Agency, Information and Emergent Simulation: NetLogo with Human Turtles

One of my major preoccupations in my almost completed manuscript on the lives of three Zimbabwean men has been the exploration of the relationship between individual agency and historical causation, between microhistory and macrohistory, between willful, contingent choices and established, determinate structures.

Emergence as an idea inevitably raises similar questions. If complexity and structure in human societies arises from the cumulative actions of many individual human agents, but those actions are based on rules or fixed principles, where does human agency enter in? Can emergence be an applied technology of social change, a dynamic that we can choose to push in particular directions? Such questions seem to belong strictly to the realm of philosophy: it is hard to fight the instinct to shrug in resignation whenever a debate about the existence of free will begins.

In this module, however, I propose a limited investigation of human agency within an emergent system using the program NetLogo as a template. Rather than running the program itself, I propose to ask human subjects to substitute for the “turtles” in NetLogo. “Turtles”, in the program’s nomenclature, are the agents to whom rules are attached, and who enact those rules in turn within the background environment of the program.
Human agents will be instructed to act as surrogate “turtles” in three steps within a single NetLogo simulation design, plotting their moves on a gridded sheet of paper. In each run of the experiment, all participants will be acting simultaneously, but with varying degrees of communication or isolation from each other. At the moment, I plan to use the “Termite” simulation described above.

In the first run of the experiment, human subjects will have no information about their environment beyond a single pixel distance from their assigned turtle. They will be bound by the same ruleset as the “termites” in NetLogo except that they be allowed to choose the direction of their next move. In the second run of the experiment, human subjects will have global information about the entire playing field and a choice of direction in which to move on each time step. In the third run of the experiment, they will have global information, simultaneous communication with all other human players and the choice to move in any direction.

The design of this module is intended to test both how a combination of rules and will or intention might affect the range of structures possible within an emergent system and how the information available to autonomous agents within such a system might change their actions.

Module 2) Sociality and Economy in Persistent World Massively-Multiplayer Computer Games

This module builds on research that I have already conducted into the internal economic behaviors and regimes of value within existing massively-multiplayer persistent world computer games (MMOGs) like Everquest, Ultima Online and Asheron’s Call.

Persistent world computer games change accumulatively over time: both the individual characters controlled by particular players and structures and features of the gameworld as a whole are affected by such changes. Several of the key designers specializing in this type of software have recognized that the dynamic evolution of such games and the communities of players surrounding them have already recognized the intense relevance of emergence as an idea for understanding and designing these games. In my own work, I have argued that they can serve as an unintentional test bed of sorts for the study of emergent phenomenon.

Here I propose to continue my studies of current massively-multiplayer games with an eye to extending the premises of Module 1 into a less-controlled, non-experimental environment. Massively multiplayer games effectively convert human players into rule-based autonomous agents within software environments, but in this case, the ruleset that constrains what players can do in the gameworld is being manipulated by players who bring real-world desires and goals, often incompatible ones, to the act of playing. The question of whether the game itself or the prior instincts and desires of the players are producing particular kinds of persistent structures within these gameworlds is often difficult to answer.

In this module, I will continue ongoing projects of ethnographic participant-observation of several MMOGs and investigate the possibility of limited research access to proprietary quantiative data describing the aggregate actions of players in several such games.

Module 3) Overloading NetLogo: The Consequences of Excessive Complexity

One of the practical dilemmas that existing software platforms like NetLogo or Mathematica that are used to test or explore emergent phenomena often confront is their relative simplicity compared to the vast majority of real complex systems that researchers have regarded as potentially characterized by emergent dynamics or principles in their development. A few existing empirical cases, such as the collective behavior of slime-molds, can be effectively modeled within the NetLogo environment, but for the most part, NetLogo and similar platforms largely offer only suggestive hints about how emergence might work in vastly more intricate cases.

In the end, there is a slight feel of deceptive tautology in many existing NetLogo simulations and other simulations and models that propose to use emergence. The first-time viewer of such simulations finds their results surprising or unexpected, but in retrospect, it is always possible to see why a particular set of rules produced a particular set of behaviors. Resnick has written that he does not regard NetLogo as a simulation, but on the other hand, his particular programs also possess a designed coherence to them.

In his much-discussed work A New Kind of Science, Stephen Wolfram identifies a very small set of cellular automata as being truly “interesting” by virtue of the fact that they produce complex patterns that their pattern or structure continues infinitely without resolving out into a simple repetition, randomness or unitary state.

What I propose to do in this module of my research is to explore a related kind of “interesting” result by setting out to deliberately overburden a NetLogo simulation with excessive complexity. Rather than trying to figure out which rulesets might create the most effective simulation or model of a more complex process, I plan to start with: 1) an environment; 2) a single class of generic objects in the environment and 3) a single class of autonomous agents. I then propose to devise a method for the random, auto-generation of NetLogo-readable rules to attach to the agents. In each iteration of this process, following the addition of a new rule, I will run the simulation.

What I am curious about here is 1) whether truly “unexpected” or, in Wolfram’s parlance, “interesting”, structures emerge when the rulesets are not being designed to simulate or model anything in particular; 2) at what point the number of rules randomly attached to agents produces either endless sequences of dynamically evolving structures or patterns or an indiscriminate mess, whether simulated emergence as it begins to approach the density of complexity that characterizes most real-world systems can continue to be as recognizable or resonant as it is when it is simpler and more designed.

Module 4) Narrative and Organicism: Storywriting After NetLogo and Computer Game Sessions
One of the areas of inquiry that most interests me and relates to my work as a historian within research relating to emergence is the exploration of emergence as a strategy for the automated generation of narrative or storytelling. A number of scholars in different fields have ongoing research projects along these lines: I propose a modest, limited contribution to this area of inquiry. One of the difficulties that some investigators have discovered is that programs which use emergent principles to create narratives cannot “leave out the boring bits” , that they can only string together endless sequences of the root components of narrative. What I propose to do is ask how human subjects make the leap from the action of emergent systems to narrative descriptions of those actions.

Often when observers first see a NetLogo simulation, they rapidly anthropomorphize and narrativize the acton of the turtles, speaking in terms of their desire to accomplish particular goals or create particular structures. What I propose to do in this module is to first ask individual human subjects to watch a short NetLogo simulation and then after a short interval to describe to a researcher the events that they witnessed. Then I propose to do the same thing but with groups of three subjects who watch together and are asked to create a narrative through conversing with each other about what they watch. Finally, for purposes of contrast, I plan to have subjects play a short five-minute run of a computer game which has no pre-set sequence but provisions “narrative components” like charactes, setting and the like and to then have those subjects tell a story of what happened during their play session .

The most important objective in this proposal is to observe how human agents make the leap from emergent simulations to narratives, particularly to see at what junctures they elect to “leave out the boring bits” or otherwise edit or rewrite the sequence of events, possibly with an eye to understanding how a computer-mediated process might duplicate or emulate this kind of editing.

Of the five projects I propose to undertake, this module is the one which most closely builds upon existing work, both by Resnick and by a number of other researchers, and where I may wish to look for opportunities to collaborate with or observe the ongoing work of other researchers.

Module 5) Counterfactual Arguments Through Emergent Simulations

This is the module which I regard as being the most important project I propose to undertake within this proposal but it is also the one that I have the least programmatic sense of how to accomplish.

In their book Growing Artificial Societies, Joshua Epstein and Robert Axtell describe their work with an emergent simulation they called Sugarscape, arguing that this simulation offers the first glimpse of what they regard as a “total social science”, where the making of models to test hypotheses might not inevitably involve gross forms of simplification but could instead incorporate hundreds of variables and demonstrate social processes on a microscale from the bottom-up.

I see something of the same potential in emergent simulations with regard to one specific aspect of the discipline of history. The sub-genre of historical writing called counterfactual or contrapositive scholarship, more popularly known as “What if?” history (as in “What if Germany had won World War II?”), has been around for a long time. However, most scholarly historians regard formal counterfactual writing with great suspicion or active contempt. In part, this is because it tends to be the province of military historians and is usually written in a broad, popular and narrative manner.

However, it is also because formally constructed counterfactuals tend to be highly arbitrary exercises that can be made to confirm the pet arguments of any historian about causality without being subject to the same kinds of methodological tests that other historical studies are potentially open to. When a historian sets out to ask, “Could Hitler have won World War II?” it is generally with a completely fixed idea about the answer: the exploration of the question is only a way to boltress a preconceived argument.

The problem is that broadly speaking all historical analysis, as the philosopher David Lewis has noted, all arguments about causality in human experience are implicitly counterfactual. It is impossible to make any claims in historical scholarship which do not have a counterfactual component unless those claims are rigidly deterministic. Some scholars like Philip Tetlock and Aaron Belkin have proposed rigorous rules or criteria for separating out reasonable and useful counterfactuals from highly fanciful or arbitrary ones.

What I would like to do is find out whether it is possible instead to build a flexible “counterfactual engine” that would use an emergence-based approach to explore a variety of counterhistorical cases. The goal would not be to create simulations which demonstrated a single or rigid hypothesis about causality, but to create a “possibility space” of connected thought experiments, of multiple answers to the question “What if?”, results that could demonstrate and constrain the work of analyzing counterfactuals without being as rigged or preconceived as counterfactual writing tends to be at present. Historians often go into archival research with expectations about what they will find, but equally, they are often surprised at the documents and materials they uncover. I would like to use emergent simulations to achieve the same effects for counterfactual scholarship, to open up the possibility of the unexpected, and to use the results as part of a new approach to counterhistorical writing.

The implementation of this module is largely unfixed, in contrast to the other four. The first step will have to be an exploration of the suitability of existing platforms, including NetLogo and Mathematica.