part_of::[[Miller & Drexler (1988) Comparative Ecology]]
This paper appeared alongside “Markets and Computation: Agoric Open Systems” and “Incentive Engineering for Computational Resource Management” in the same Huberman volume. Together the three papers constitute what became known as the Agoric Papers — a coordinated intellectual program proposing market mechanisms as the organizing principle for open computational systems.
“Comparative Ecology” occupies a specific role in that program: it provides the evolutionary justification. Where “Markets and Computation” develops the economic mechanisms and “Incentive Engineering” addresses resource pricing, “Comparative Ecology” answers the prior question: why would market-based evolution produce better computational systems than alternatives? The paper builds the argument from first principles through comparative analysis, then hands off to the companion papers.
The term “agoric” — from the Greek agora, meaning marketplace — signals the paper’s commitment to voluntary exchange as an architectural principle, not just an implementation strategy. The point is not that markets happen to be efficient, but that voluntary exchange is structurally distinct from predation, command, and parasitism in ways that make it the correct foundation for open systems.
The paper’s analytical structure rests on comparing three evolutionary systems:
Biological ecosystems favor predation and arms races. Biological ESSs tend toward “teeth and armor” — strategies that succeed by extracting resources from others. The mutual escalation dynamic means that evolutionary progress is largely zero-sum: each adaptation provokes a counter-adaptation. This makes biological ecosystems an unreliable model for productive computational evolution.
Axelrod’s iterated prisoner’s dilemma tournaments showed that cooperative strategies can dominate even without enforcement. The key result — that “nice and retaliatory” (tit-for-tat) dominates — established that mutual cooperation is stable when interactions repeat and memory exists. This is a step toward voluntary exchange but lacks the full property rights and currency framework needed for computational implementation.
Market economies extend Axelrod’s result: trade is symbiotic rather than zero-sum, property rights prevent coercion, and conserved currency prevents parasitic loops. The paper argues that direct computational markets — where these properties are implemented architecturally rather than socially — produce the most favorable ESS: “productive and wary.”
The comparison is not merely academic. Each system has a different failure mode, and the paper’s contribution is identifying which failure modes are architectural (can be designed out) versus inherent. Biological predation is inherent to fitness competition without property rights; EURISKO-style parasitism is architectural (it arises from missing currency conservation).
The paper’s most durable technical contribution is the identification of object encapsulation as a computational implementation of property rights. The argument runs:
This is a significant move: it identifies a domain where the theoretical preconditions for market efficiency hold more cleanly than in human markets. The paper does not claim that computational markets are perfect — it claims that some of the frictions that make human markets imperfect can be eliminated by architecture.
The cryptographic extension adds the external dimension: encapsulation handles internal access control, but public-key cryptography handles external identity and currency integrity. Together they create a complete property rights framework at the software level.
The paper’s encapsulation argument is a direct precursor to capability security. Mark Miller’s later formalization of the Principle of Least Authority — that each software component should hold only the authority it needs to perform its function — rests on the same architectural intuition: authority should be held structurally, transferred voluntarily, and not obtainable by theft.
The connection is not just historical. Capability security is the security-theoretic generalization of the encapsulation-as-property-rights claim. Where “Comparative Ecology” makes the economic argument (proper encapsulation enables market efficiency), capability security makes the security argument (proper encapsulation enables safe composition of untrusted components). Both arguments share the architectural move of treating access rights as structural rather than social.
Miller’s 2006 dissertation “Robust Composition” is the formal development of the capability security program. The intellectual lineage from this 1988 paper through the E language and Caja project to modern secure JavaScript environments runs through the encapsulation argument introduced here.
The direct market concept — where computational agents negotiate, pay for services, and hold resources through cryptographically enforced property rights — is structurally similar to what Nick Szabo would formalize as “smart contracts” in the early 1990s. Szabo explicitly referenced agoric systems in his 1994 writing, showing direct familiarity with the framework.
The paper does not use the term “smart contract,” but the concept is present: agreements enforced by computational mechanisms rather than legal codes, with property rights implemented as physical laws rather than social conventions. The contribution is architectural — it established the conceptual vocabulary that later work could name.
The American Information Exchange (AMiX), which Miller led in the late 1980s, may be the first operational smart-contracting system. AMiX was one of the acknowledged influences on Szabo’s thinking. The path from this paper’s theoretical ecology to AMiX to Szabo to blockchain smart contracts is traceable, if not always formally cited.
The evolutionary timeframe problem. The paper assumes that direct computational markets will evolve productive strategies, but does not address the timeframe or path dependence. Biological evolution succeeds over millions of generations; computational markets might reach local optima quickly and require external perturbation to escape. The paper does not analyze convergence properties.
The property rights assumption. The argument that encapsulation implements property rights assumes that the encapsulation is correct and complete. In practice, software systems have bugs, and a bug that breaks encapsulation breaks the property rights framework. The paper treats encapsulation as a solved problem rather than an ongoing engineering challenge.
Voluntary exchange and power asymmetries. Market ecosystems are more symbiotic than biological ones when participants have comparable power. If one agent controls resources that others need to exist, “voluntary” exchange becomes coercive. The paper does not address power concentration or monopoly dynamics within direct markets.