Notes on Clark Chapter 4
("Collective Wisdom: Slime-Mold Style")
Econ 308: Agent-Based Computational Economics
- Last Updated: 16 June 2006
- Latest Course Offering: Spring 2006
- Course Instructor:
-
Professor Leigh Tesfatsion
tesfatsi AT iastate.edu
-
Syllabus for Econ 308
- Basic Reference:
- Andy Clark, Being There: Putting Brain, Body, and World Together
Again, MIT Press, Cambridge, MA, 1998 (paper), ISBN: 0-262-53156-9
Basic Concepts
- Self-Organization and Emergence (Clark, p. 73)
- "A self-organizing system is one in which some kind of
higher-level pattern emerges from the interactions of multiple
simple components without the benefit of a leader, controller, or
orchestrator."
- Direct Emergence (Clark, p. 73-75)
- "(D)irect emergence... relies largely on the properties
of (and relations between) the individual elements, with
environmental conditions playing only a background role."
- Examples: (1) Traffic jams; (2) The compression of
the molecules of a gas in a container, leading to the emergence of
the macro properties "temperature" and "pressure."
- Indirect Emergence (Clark, p. 73-75)
- "(I)ndirect emergence... relies on the interactions
of individual elements but requires that these interactions be
mediated by active and often quite complex environmental
structures."
- Example 1 (pp. 75-76): Nest-building behavior of
termites, in which arches are slowly constructed as a result of the
propensity of each individual termite to make mudballs
(incorporating its chemical trace) and to drop them where chemical
traces from other termites are strongest.
- Example 2 (pp. 76-77): Ship navigation -- an overall
system of artifacts, agents, natural world, and spatial organization
coming together to solve the problem of keeping the ship afloat and
headed in the desired direction. "The overall (ship-level) behavior
is not controlled by a detailed plan in the head of the captain."
- Rational Reconstruction (Clark, p. 80)
- "(T)he practice of casting each problem immediately in terms
of an abstract input-output mapping and seeking an optimal solution
to the problem thus defined."
- Embodied Active Cognition (Clark, p. 81-82)
- Clark describes three key features of this methodology: (a)
Real-world real-time focus; (b) Awareness of decentralized solutions
and the possibility of emergent solutions; (c) An extended vision of
cognition and computation spread out in space and time.
- Classical Cognitivism (Clark, p. 83)
- "(D)epiction of the mind in terms of a central logic engine,
symbolic databases, and some peripheral `sensory' modules. Key
characteristics of this vision include these ideas: (1) memory as
retrieval from a stored symbolic database; (2) problem solving as
logical inference; (3) cognition as centralized; (4) the environment
as (just) a problem domain; and (4) the body as input device."
- Connectionist (Artificial Neural Network) Revolution (Clark,
p. 83-84)
- Replaces first three tenets of classical cognitivism with
the following: (1)* memory as pattern re-creation; (2)* problem
solving as pattern completion and pattern transformation; and (3)*
cognition as increasingly decentralized. However, Clark claims that
this radical rethinking of the nature of the inner cognitive
engine has been marred by "residual classicism" because it has been
largely accompanied by a tacit acceptance of the classical
marginalization of body and world.
- Emergentist Perspective (Clark, p. 83-84)
- "To see adaptive success as inhering as much in the complex
interactions among body, world, and brain as in the inner
processes bounded by skin and skull." The emergentist perspective
thus augments connectionist views by "a vision of the environment as
an active resource whose intrinsic dynamics can play important
problem-solving roles and the body as part of the computational
loop."
Key Issues
1. Role of self-organization and emergence in human affairs?
(Clark, p. 73): "Collectives of humans, too, exhibit forms of
emergent adaptive behavior. The biological brain, which parasitizes
the external world... so as to augment its problem-solving
capabilities, does not draw the line at inorganic extensions.
Instead, the collective properties of groups of individual agents
determine crucial aspects of our adaptive success."
2. Importance of distinction between direct and indirect emergence?
(Clark, p. 74): "The difference ... concerns the extent to which we
may understand the emergence of a target phenomenon by focusing
largely on the properties of the individual elements (direct
emergence), versus the extent to which explaining the phenomenon
requires attending to quite specific environmental details."
3. How does harmonization of brains, bodies, and worlds come about?
(Clark, Section 4.4, pp. 77-80): "...an important part of the answer
is clearly `through evolution.'" Examples: Ship navigation
team facing an unexpected challenge; Optimal placement of footpaths to
connect a complex of already-constructed buildings.
4. What's wrong with "rational reconstruction"?
(Clark, Section 4.5, pp. 80-81): Clark argues that rational
reconstruction can "mislead in several crucial ways," paraphrased as
follows: (a) It can obscure opportunistic strategies that involve
acting upon or otherwise exploiting the real world as an aid to
problem solving; (b) it invites a view of cognition as passive
computation in which the important role of epistemic action
falls through the cracks; and (c) it obscures the role of history in
constraining the space of biologically plausible solutions.
5. What challenges are posed by the "embodied active cognition"
alternative to rational reconstruction?
(Clark, p. 82): "(T)he study of embodied active cognition clearly
presents some major conceptual and methodological challenges.
- The problem of tractability: How are we to isolate
tractable phenomena to study?
- The problem of advanced cognition: How far can we
(should we) hope to go with a decentralized view of the mind?
Surely there is some role for central planning in advanced
cognition.
- The problem of identity: Where does all this leave
the individual person? Does it imply that the individual brain and
the individual organism are not proper objects of scientific study?
6. How can we respond to these concerns?
(Clark, p. 83) "The key to integrating the facts about advanced
cognition with the vision of embodied active cognition lies, I shall
suggest, in better understanding the roles of two very special
external props or scaffolds: language and culture."
Questions Arising from In-Class Discussion
1. Comparing and contrasting the termite and navigation cases as examples
of indirect emergence
Kunal Daftari's Question: How are these examples similar? How are
they different? Some dimensions along which comparisons can be made concern:
(a) stationary vs. nonstationary environments; (b) degree of hierarchical
structure (do the termites have a "captain"? what does being a captain of a
ship mean in terms of control over ship activities?); (c) the "completeness"
of the rules that govern the two systems (is every contingency covered?),
which is closely related to the "soft assembly" versus "hard assembly"
distinction made by Clark in previous chapters; and (d) the origin of the
rules that govern the two systems (genetic vs. socially constructed).
2. Considering direct and indirect emergence for economic systems
Do properties of economic systems arise from "direct" emergence?
"indirect" emergence? What do these concepts really mean for economic
processes with self-aware agents? Do expressed "behaviors" and the "rules of
the game" co-evolve together for economic systems? How much is truly
plastic, and how much is determined by the constraints of "human nature"?
3. Time constraints
Jordan Swanson's Question: How do time constraints affect the way
agents solve problems (e.g., the "optimal" choice of chess moves in a
real-time game)? the way systems (e.g., a team of ship navigators)
self-organize in the face of an unexpected danger?
4. A need for social science?
Jordan Swanson's Question: How much can be learned about ants from
studying ONE ant? about people by studying INDIVIDUAL people in isolation
from each other?
5. The baseball analogy
Jen Diaz's interesting use of a baseball team as an example of a mixed
hard and soft-assembled system capable of self-organization within the
constraints of a socially constructed rule set (which has been evolving over
time).
6. Identical vs. Heterogeneous Agents
Andy Ruff's Interesting Question: How does self-organization
differ when the underlying agents are IDENTICAL versus when they are
intrinsically different or when they can learn to be different (e.g., through
division of labor)?
7. "Edge of chaos" considerations?
In what kinds of environments do soft-assembled systems work
best? hard-assembled systems? Do systems evolve from one form to the other
over time by self-organization? by design? Can they move in either
direction? Is there an "optimal" degree of soft-assemblage for a social
system in terms of permitting change, growth, development, or "progress"?
Copyright © 2006 Leigh Tesfatsion. All Rights Reserved.