Echo Model

Status: Core concept — formed Session 4 Connected to: Multi-scale composition, NK model, fitness landscapes, open-ended evolution, CAS theory

The Model

Echo is a computational model of complex adaptive systems (CAS) designed by John Holland (SFI), implemented by Terry Jones and Stephanie Forrest (1993). Created at Murray Gell-Mann's request to illustrate complex structures produced by natural selection.

Echo is a family of models, not a single model. The core: evolving agents in a resource-limited environment. Agents have "chromosomes" (strings of resource letters) with tags (external, recognizable) and conditions (internal). Agents interact via three mechanisms:

1. Combat — offense/defense tags determine outcome probabilistically. Winner takes loser's resources.
2. Trading — agents exchange excess of their "traded resource." Equalizes resource distribution.
3. Mating — genetic exchange via crossover (bacterial-like, not sexual reproduction).

Fitness is ENDOGENOUS — no external fitness function. An agent's fitness emerges from its interactions with the environment and other agents. (Norman Packard, 1989, first emphasized this.)

Holland's Seven CAS Basics

Holland's theory of CAS (from Hidden Order, 1995) identifies seven essential attributes:

Four Properties:

1. Aggregation — hierarchical organization; meta-agents built from simpler agents. "The emergent phenomena that result are the most enigmatic aspect of CAS."
2. Nonlinearity — aggregate behavior not reducible to sum of components.
3. Flows of resources — open-system throughput; CAS maintain identity despite component turnover.
4. Diversity — populations are diverse; "perpetual novelty is the hallmark of CAS."

Three Mechanisms: 5. Tags — external characteristics facilitating selective interactions, aggregate formation, boundaries. 6. Internal models — tacit (implicit action rules) and overt (explicit lookahead). Allow prediction. 7. Building blocks — lower-level components used in physical construction and cognitive decomposition.

Echo's Failure (Smith & Bedau, 1997)

Smith & Bedau ran thousands of Echo simulations (up to 10^6 generations) at six mutation rates (0.00001 to 0.05). Key findings:

Robust Behavioral Patterns

1. Simple trading ecologies dominate. All agents converge to trading the same single resource.
2. Genome accumulation of traded resource. Each new dominant genome contains more copies of the traded resource. This makes reproduction harder (needs more resource to copy genome), so agents live longer, acquire more resources, support larger populations. Counterintuitive: the "fittest" genome is the one that's HARDEST to replicate.
3. Combat spikes — brief proliferations of fighting agents that decimate trading ecologies, then collapse. Trading resumes with a different resource.
4. No hierarchical aggregation. No evidence of diverse, hierarchically organized adaptive aggregates. Only simple trading ecologies and transitory combat.

The Verdict

"Echo is NOT a complex adaptive system." It possesses the mechanisms (tags, internal models, building blocks) by design, but they were installed by fiat, not emergent. Most critically:

"There is no evidence of the emergence of a diversity of hierarchically organized adaptive aggregates. The only evident aggregations were simple trading ecologies and transitory combat ecologies. Although the number of distinct genotypes grew with the mutation rate, this heterogeneity never created any significant diversity in behavior; it was genotypic diversity without phenotypic diversity."

Smith & Bedau's Proposed 8th CAS Property

They propose that what unifies Holland's seven features is:

"the property of robust, open-ended emergence of hierarchical, adaptive structures"

And they propose an EIGHTH characteristic property of CAS:

"the ability of emergent interacting components to create and flexibly maintain their own boundaries and their capacities for interacting with other components"

This is EXACTLY our thesis. Smith & Bedau independently identified:

• Multi-scale composition (hierarchical adaptive structures) as the missing ingredient
• Autopoiesis (boundary creation and maintenance) as the mechanism

They wrote: "Concretely embodying them in some successor model to Echo is the only way to make them precise and subject them to rigorous scrutiny."

Why This Matters for Our Project

1. External validation. Smith & Bedau (1997) independently arrived at our thesis from a completely different starting point (empirical study of Echo, not ANT/computational irreducibility). This is strong corroboration.

2. The gap is real and documented. Echo's failure is not a design flaw in one model — it's a systematic absence. "As far as we know, no present model actually has the features by which Holland characterizes complex adaptive systems." Every ALife model tested shows the same convergence to simple dynamics.

3. Our synthesis fills their gap. Smith & Bedau proposed the 8th property but never implemented it. Our synthesis provides the implementation path:

   • Stigmergy (Session 3) = the mechanism for emergent components to create boundaries (traces that accumulate and constrain)
   • Autopoiesis (Session 2) = the mechanism for maintaining those boundaries (self-production)
   • The trace→actor crossing (H7) = the phase transition from trace to 8th-property actor
   • ANT translation = the language for the restructuring event

4. Echo's resource accumulation dynamic is instructive. The counterintuitive finding — genomes that are HARDER to replicate dominate because agents live longer — reveals how selection pressure alone can produce unexpected dynamics. This is relevant for our simulation design: we should not assume that "fitter" means "easier to replicate."

Criticisms and Limitations

1. Echo is a family of models, not one model

Holland himself emphasized this. Smith & Bedau studied version 1.3 beta 2. Other versions (Hraber et al., 1997) showed somewhat different dynamics. The conclusion that "Echo is not a CAS" may be version-dependent.

2. Smith & Bedau's parameter space was limited

They varied mutation rate but kept other parameters fixed. They acknowledge: "we never saw any macro-level evidence for more interesting evolutionary dynamics than those described." But they didn't exhaustively search parameter space.

3. The 8th property is speculative

Smith & Bedau explicitly say their suggestions are "really just so many words" until concretely embodied in a successor model. The 8th property is a hypothesis, not a validated finding.

4. Echo's endogenous fitness is still single-scale

Even with endogenous fitness (no external fitness function), Echo operates at one scale. Agents interact with other agents at the same level. There's no mechanism for emergent structures to become new-level actors with new interaction rules.

Empirical Evidence

Smith & Bedau (1997) — extensive

Thousands of runs, up to 10^6 generations, six mutation rates. Quantitative data on population dynamics, genome evolution, interaction frequencies, resource reservoirs. The most thorough empirical study of Echo available.

Hraber, Jones & Forrest (1997) — "The Ecology of Echo"

Published in Artificial Life journal. Shows somewhat richer dynamics than Smith & Bedau found. Smith & Bedau note this discrepancy but argue it's intensity, not qualitative difference.

No empirical studies test the 8th property

No study has implemented Smith & Bedau's proposed 8th CAS property in a successor model. This is the gap our project addresses.

Cross-References

• [[concepts/multi-scale-composition]] — Echo's failure confirms the multi-scale composition gap
• [[concepts/nk-model]] — Both are SFI CAS models; NK provides the fitness landscape framework Echo implicitly uses
• [[concepts/fitness-landscapes]] — Echo's endogenous fitness is a step toward dynamic landscapes
• [[concepts/stigmergy]] — Stigmergy is the mechanism for the 8th property (boundary creation)
• [[concepts/autopoiesis]] — Autopoiesis is the mechanism for boundary maintenance (the 8th property)
• Holland (1995) — Hidden Order — seven CAS basics
• Smith & Bedau (1997) — "Is Echo a Complex Adaptive System?" — the critical evaluation
• Hraber et al. (1997) — "The Ecology of Echo" — somewhat different dynamics