Emergent Systems Working Group

February 26 and March 4, 2004
Tim Burke

Emergence: What It’s Good For (Maybe)

(the outline of this talk first appeared at http://www.swarthmore.edu/socsci/tburke1/marchemerge.html)

1. Of all the gin joints in all the world, emergence had to walk into this one, or why the topic interested me in the first place

GAFOSS: two tribes of australopithecines. Discovering that I swing both ways

My humanistic impulses: narrative, meaning, interpretation, experience, consciousness, individual agency, contingency, unexpected results, respect for unknowability. A thousand stories in the naked city, and a multitude of truths. The geographer and the ephemeral flower. Romanticist reaction to the hubris of hard social science. History as Mr. Toad's Wild Ride. Down with Utopia.

But: truth matters. Empiricism matters. Reason matters. The distinction between bad or weak academic work and good academic work is real and correlates to its rigorous pursuit of truth; some humanistic inquiry reduces that distinction entirely to a matter of taste or fashion or power/knowledge. Social science knows (really knows) things that humanistic inquiry doesn't and can't know: patterns, structures, regularities, systems, underlying forces, causality. The content of ethics is non-arbitrary and meaningfully transhistorical. Material constraints and social facts on the ground matter. Up with policy, up with planning, up with the good society.

The histories I want

 

a history where individual agency is interesting in and of itself, but where it also meaningfully shapes what happens at larger scales of human experience-not as typical or collective or representative, a single example of an absent corporate whole, but in its individuality and idiosyncracy.

an acknowledgement that unexpected results and genuine newness appear in human history all the time, but also the ability to relate what is new to what came before. (Generative social science tends to make everything interesting into an externality.)

a history that has room for particularity and locality without making them into a theoretical fetish object, an exclusive epistemological choice. Relating particularity and locality to systematicity and generality without subjugating either to the other .

a history that does not insist on the equivalence of initial conditions and later consequences (Marxism's "last instance": the reducibility of the superstructure to the base; Freudian psychotherapy's developmentalist history of the individual). There's reductionism and then there's reductionism. Looking for the good heuristic kind, rejecting the bad ontological kind.

A good metaphor is good enough. No it's not.

Has emergence solved my problems? Can this mind be saved?

 

Not really, but it has helped considerably, both metaphorically and empirically.

Suggests a new way to talk about the irreducibility of later conditions to initial conditions while also insisting that they are always related

Relates small idiosyncratic actions and scales to large systemic consequences and scales without flattening particularity and locality

A new way to think about the appearance of novelty and newness in a system without insisting on total disjuncture or disconnection.

A good metaphor and yet also tantalizingly empirical.

The 2 major sticking points:

 

The velvet determinism of emergence. (Different starting conditions, different results; same starting condition, same results.) Is this an artifact of the relative simplicity of the models of emergent complexity that we have, or this an essential feature of the concept?

The problem of scalability, of knowing that stigmergic effects could be viewed as autonomous agents at the next level of complexity and yet being unable to encompass that within any given discussion of an emergent system.

How I'm applying emergence: a new history of indirect rule in Africa

 

a. What was indirect rule?
b. Mahmood Mamdani's sensible but terribly wrong idea
c. Robinson and Gallagher's old theory of peripheral causation and its problems
d. Emergence to the rescue: local actors, systematic results. The 1890-1910 window. The colonial state as stigmergic.

 

---

 

Why the Grobstein/Burke thought-experiment on NetLogo and human agents is still important to me

Santa Fe Summer Institute does "Boids" with blindfolded humans

scales of determinism?

 

while perhaps human agents acting like turtles are in turn determined by human brains, human society, the narrow set of particular expectations they bring to participating in the exercise, and so on, the boundaries that this 'determination' places upon them are so capacious that they approach free will as an approximation. (Wolfram's argument, as I understand it). And so the systemic results of a rule-based simulation are potentially wildly different and much more variable than they are with software agents who must follow all rules.

A cheap way to tackle the scalability issue? Comparison with the problem of MMOG economies: human agents in a rule-constrained environment. Is what happens the total complexity of human beings in a modern society or is the relative simplicity of the ruleset?

The counter-factual engine: new specs, goals, problems

 

Autonomous agents distinct from an environment

Agents have energetic limitations and fixed lifespans

Heterogenous population of agents (many different agents with different rules)

Genetic implementation of agent rulesets (agents can pass on rules to later agents; rules can be changed by exchange between agents; ideally, new rules can come into being through generational transmission and change). Requires imposition of fitness landscape on agents.

Probablistic execution of rules, in relation to fitness landscape and environment; e.g., heterogenous behavior within any given class of agents. Deliberately imperfect optimization by agents, e.g., probability may dictate that agent will execute non-optimal rule

Environments where stigmergic effects can acquire persistent rulesets of their own over time; "scalability of complexity" up into the environment.

Two test beds: the Atlantic world, 1350-1800 and China and environs 900-1600.

 

Origins and evolution of the Atlantic system, especially the slave trade; the "Why didn't China industrialize?" problem.

Problem of setting: neither of these are isolates.

Problem of initial conditions: what classes of agents, how many, what rules, what fitness landscape

Would episodic or event history be a better test bed?

The quixotic goal: comparative quantifying of contingency through mapping the total possibility space of macrohistorical landscapes.

Additions, revisions, extensions are encouraged in the Forum and/or at emergent.brynmawr.edu

Participants for September 23, 2003: Jan Trembley, Karen Greif, Jim Marshall, Tim Burke,
Anne Dalke, Mark Kuperberg, Hannah Wilhelm, Doug Blank, Al Albano, Paul Grobstein (10)

Participants for September 30: Karen Greif, Jan Trembly, Paul Grobstein, Jim Marshall, Mark Kuperberg, Doug Blank,
Ted Wong, Anne Dalke, Al Albano, Hannah Wilhelm, Tim Burke, Alan Baker, Emily Kahoe, Jason Coleman (14)