January 27, 2004

Reading: Johnson, "Emergence"

Steven Johnson taps self-organization theory to relate the behaviors of slime mold, ants, Sims and WTO protesters in his readable introduction to the emergence aspects of complexity theory, "Emergence: The Connected Lives of Ants, Brains, Cities and Software" (2001). One of the more illuminating works in Wired's Howard Dean Reading List mentioned in an earlier post at this site. (more ... )

"The Myth of the Ant Queen"
Johnson reminds us that despite the title, ant queens are not authority figures. Ant colonies grow and collectively develop solutions to environmental challenges, with no one directing anything. Instead, individual behavior, marked by scent trails that grow stronger with repeated use, reinforces itself through positive feedback. These scent trails create the colony's tangible form of cumulative "memory."

For thousands of years, cities (like ant colonies) have grown and organized themselves into complex arrangements that kept large numbers fed, housed, protected and employed, all without central planning. In cities over the centuries, pathways, hubs and trade-specific marketplaces have emerged from repeated use and persist, again as a form of cumulative memory that guides old and new dwellers to resources. He calls a city a "pattern amplifying machine; its neighborhoods are a way of measuring and expressing the repeated behavior of larger collectivities -- capturing information about group behavior, and sharing that behavior with the group." Johnson, p. 40.

Johnson recommends us to the work of the American scientist Warren Weaver (1894 to 1978), one of the founders of information theory, who divided scientific inquiry into three camps:
1) study of simple systems with few variables
(e.g. physics)
2) study of "disorganized complexity" with a great many variables
(e.g. probability theory)
3) study of "organized complexity" with many interrelated variables
(e.g. complexity theory).

Johnson observes that the study of complexity and emergence have taken off in recent years in part because of the development of the affordable computing power required to study it. Weaver's work influenced Jane Jacobs in her writing of "The Death and Life of the Great American Cities," (1992) Johnson quotes Jacobs' observation that cities have "innate abilities for understanding, communicating, contriving and inventing what is required to combat their difficulties."

For another example, he points to the work of Oliver Selfridge at M.I.T., (one of the founders of Artificial Intelligence studies and author of the classic 1958 paper "Pandemonium") who developed a model to teach a computer to learn pattern recognition, relying on a distributed, bottom-up intelligence. He did so by creating swarms of simple programs ("demons") that reported to higher-level demons that tallied the confidence votes of the swarm. Through a scheme Selfridge described as natural selection, "right" guessing demons were reinforced and "wrong" guessing demons disregarded. Later researchers' experiments combined randomly generated programs with feedback and found that increasingly successful software "emerged" through similar natural selection.

Work of researchers like Weaver, Jacobs and Selfridge combined, cumulated and fed back upon each itself. The study of organized complexity emerged as a distinct school of thought recognized by several Nobel prizes and best selling books like Hofstadter's "Godel, Escher, Bach," (1999) Gleick's "Chaos" (1988) and Waldrop's "Complexity," (1992) as well as popular simulation games like Sim City.

"Street Level"
Johnson tells us about Deborah Gordon's studies of ant colonies that suggest five fundamental principles for building local knowledge into macro-intelligence and adaptability:
* More is different
-- a critical mass of instances is needed to observe emergent behavior
* Ignorance is useful
-- emergent systems get unwieldy if components are complicated
* Encourage random encounters
-- they enable adaptation and discovery
* Look for patterns in the signs
-- pattern detection leads to circulation of meta-information
* Pay attention to your neighbors
-- "local information can lead to global wisdom"

Gordon also learned that to understand ant colonies, one must study them over the decades of their lifespan as their global behavior emerges. "The persistence of the whole over time -- the global behavior that outlasts any of its component parts -- is one of the defining characteristics of complex systems." Johnson, p. 82.

But does not human volition distinquish humans from ants? Johnson contends that human consciousness is limited to a scale that deprives it of the millenial scale of ant-like decision patterns. Thinking of a city as a "superorganism," Johnson says that it triumphs over other social forms because it possesses "a kind of emergent intelligence: an ability to store and retrieve information, to recognize and respond to patterns in human behavior. We contribute to that emergent intelligence, but it is almost impossible for us to perceive that contribution, because our lives unfold on the wrong scale." Johnson, p. 100.

"The Pattern Match"
Johnson extends his discussion to the concept of "learning" as something not limited to organisms with conscious awareness. As one instance of non-conscious learning, he points to an organism's development of antibodies in response to exposure to viruses or bacteria. "Its about altering a system's behavior in response to ... patterns in ways that make the system more successful at whatever goal its pursuing." Johnson, p. 104. Cities learn also, he writes, being "patterns in time" that keep their shapes over centuries. He reminds us of Lewis Mumford (1895 to 1990), who described a city as "a structure specifically equipped to store and transmit the goods of civilization."

Is the World Wide Web also learning? Johnson thinks not. Although it self-organizes like a city into stable macro-shapes, it is not adaptative, due to the lack of natural selection, he writes. It could be made so, he suggests, if hyperlinks were bi-directional instead of one-directional. All emergent systems are built out of reciprocal feedback. "Self-organizing systems use feedback to bootstrap themselves into a more orderly structure." Johnson, p. 121. He points to developing systems (using Alexa as an example) that build connections between websites based on user traffic (much like the scent trails of ants and the paths and ways of an evolving city). Such feedback systems may, Johnson writes, lead to adaptive behavior equatable to learning.

[DKS query: Does the development of bi-directional linking tools, such as Movable Type's Trackback, and the evolution of the self-referential "blogosphere" provide the feedback loop to which Johnson refers as essential for the Web to "learn?"]

"Listening to Feedback"
As an example of the effects of distributed control and positive feedback, Johnson points to The media handling of the 1998 "Naughtygate" scandal about President Bill Clinton and several women including Gennifer Flowers, Monica Lewinsky, Katherine Willey. Years before, Johnson contends, the allegations would probably not have reached the level of national news, and could have disappeared completely. But independent news sources like the Drudge Report had emerged, and CNN's desire to be competitive led it to allow local stations to air stories without network clearance.

Led by Drudge's "outing" of a Newsweek story squashed by its publishers, those independent sources broke the story. Despite attempts by the White House to dismiss it, the report was picked up by others, and the self-reinforcing power of positive feedback made the story *about* the story an item the national media had little choice but to carry. See e.g., Howard Kurtz, "With Time to Fill, Media Stretch Limits," Washington Post (Jan. 24, 1998; Page B01).

Johnson uses this case as an illustration of how positive feedback amplifies its own signal, to the limits of the media to carry it. Negative feedback, on the other hand, enables equilibrium despite changing and unpredictable external conditions. Negative feedback enables a complex system to become adaptive and to achieve homeostasis, a state of self-regulating stability. One can adjust the rules of the feedback system in ways that affect the resulting outcome, in order to foster desired values, suggests Johnson.

"Control Artist"
Supercomputing guru Danny Hillis built a software system that evolved or learned ways to sort numbers, Johnson tells us. However, the efficiency of the system peaked out with repeated iterations, as it hit "false peaks" in the software "fitness landscape." Hillis overcame that obstacle by introducing "predator" routines that constantly threatened the evolved software routines. The higher the sorting programs climbed in efficiency, the more challenging the predators became. The resulting "arms race" produced increasing competent programs as the system functioned more like an environment than an organism. See Kevin Kelly's Out of Control, Chap. 15 for a bit more on Danny Hillis' studies.

"The Mind Readers"
Humans and chimps are among the very few organisms that have evolved an awareness of that other's minds are also aware, Johnson tells us. 150 is the usual practical limit to the number of other minds that one can track, he says. To get over that limit, we have learned to mentally model people in terms of their categorization into neighborhoods, groups, classes or political parties, he suggests.

Though a bottom-up organization can produce brilliant innovation, it is itself (predictably) unpredictable, making it risky as a format in which to operate a business. Some, says Johnson, see the real battle of the near future as between hierarchical forces and decentralized forces. A similar dialectic has been observed between capitalism and the noncapitalism of innovators.

On politics, Johnson offers his views. "In fact, the needs of most progressive movements are uniquely suited to adaptive, self-organizing systems: both have a keen ear for collective wisdom; both are naturally hostile to excessive concentrations of power; and both are friendly to change. For any movement that aims to be truly global in scope, making it almost impossible to rely on centralized power, adaptive self-organization may well be the only road available." Johnson, p. 224.

As examples, he points to the NGO swarms at the World Trade Organization meeting in Seattle in 1999, which some saw as hopelessly unfocused. "What they fail to recognize is that there can be power and intelligence in a swarm, and if you're trying to do battle against a distributed network like global capitalism, you're better off becoming a distributed network yourself." Johnson, p. 226. See also "The Non-Governmental Order: Will NGOs Democratise, or Merely Disrupt, Global Governance?" Economist (Dec. 11-17, 1999).

Control, or lack thereof, is a key sticking point, as Johnson reminds us in the last chapter, when he says that "understanding emergence has always been about giving up control, letting the system govern itself as much as possible, letting it learn from the footprints." Johnson, p. 234.


Posted by dougsimpson at January 27, 2004 03:30 PM | TrackBack