Padgett, John F., The emergence of simple ecologies of skill: a hypercycle approach to economic organization, forthcoming in The economy as a complex evolving system, ed by Brian Arthur, Steven Durlauf, and David Lane, Nov 1995.

Present theory assumes the prior existence of the entities whose features it tries to explain.

In the constant flow of people through organizations, the collectivity typically is not renegotiated anew but component parts are transformed and molded into the ongoing flow of action. But which sets of component parts and of collective actions are mutually reproduceable?

Hypercycle theory is used to explain the origin of organisms on earth, where each step in the process reinforces the next steps reproduction and growth. The author attempts to find conditions for these hypercycles in organizations. Collective order is the stabilization of chains of action-reaction sequences, which fold back on each other to keep themselves alive. It turns out there are dynamic constraints on what firms are possible within life.

The organization of skill

Each element in an "economic hypercycle" is an action-capacity or skill which resides in a person until activated by action with another agent in their environment. Work is an orchestrated sequences of actions and reactions, the sequence which produces some collective result. The full set of sequences is the human "technology" of the firm.

This representation emphasizes form, not content. Only linear chains of action sequences can be investigated easily. Unlike chemical hypercycles, many kinds of skills can reside in individuals. These skills can change through interaction with others. The mix of skills within individuals is variable. Never-used skills are fogotten, and often-used skills will "reproduce". Rates of skill reproduction depend on how often they are requested.

Thus one issue is how different organizational ecologies induce different degrees of complexity in the skill mixes of the individuals that comprise them.

Padgett uses simulations to test two ways that hypercyles can be varied:

1. Length of hypercycle chain

2. Mode of skill reproduction

* source only reproduction (skill of iniator is only reproduced)

* target only reproduction (only skill of interaction recipient is reproduced)

* joint reproduction (both parties skllls is reproduced)

The first measures the complexity of the technology trying to be seeded in an organization. The second measures distribution of reward. Source only reproduction is "selfish" learning, target only learning is "altruistic or teacher" learning, and joint is when both groups benefit.

In this model social network effects are largely ignored. In this model they interact only with their four nearest neighbors in 2D space.

Model

1. Assume a 10 by 10 grid of individuals

2. Assume a set of "action capacity" elements or skllls distributed randomly across the grid as (1,2), (1,2,3), etc. with set variations between individuals

3. Assume that certain combinations are required for action. Elements 1 and 2 activate each other, 2 and 3 etc until the last element and 1 activate each other.

4. Choose at random a skill looking for action located. Intiate interaction randomly with one of the four neighbors. If a compatible element exists the interaction is successful and joint action occurs, if not it is inert and nothing occurs.

5. If successful interaction occurs, learning within the individual follows according to the specified learning rules.

6. The total volume of elements is kept constant, which means an old one is killed off randomly as a new one is reproduced.

7. The steps are repeated until equilibrium behavior begins to emerge.

Step 6 is not realistic, but helps to investigate structural effects.

This model resembles more self-organizing play more than formally-dictated work.

The simplest hypercycle: two elements only

In this case the target only reproduction creates a more complicated ecologie than the source only reproduction. In general, at equilibrium, there are clusters of alternative single-speciality nodes. The target-only and joint cluster sizes were much bigger than source only.

One finds that no spatially embodied social ecology is possible under a purely "selfish" source-only reproduction/learning regime. However in target only or joint groups, the dynamics cause depleted individuals to build skllls and bloated individuals to lose some.

These trends occur in the spatial world but are not different in the non-spatial world.

In target-only learning individuals reproduce others, while in selfish modes they reproduce only themselves. The first is self-stimulating, the second ultimately is not. In the non-spatial arrangement there are no individuals and hence no self or other.

The general conclusion is that constraints on interaction partners permit past learning to be "stored" in current behavior. This storage induces reproductive variance across individuals, because localized histories vary. In human terms that once a spatial constraing imposes a meaning for self, hypercycles of interaction causes an asymmetry between selfishness and altruism.

The Three-Element Hypercycle

Interestingly, in the target-only system of three elements, most simulations produced a stable core of a 6-step paired solution, which then established a stable core for a larger equilibrium cluster of interacting individuals. There is no stable ecology in the source-only system.

The three-element ecologies are different. People in the core are not specialized, while "parasites" on the outside usually have only one skill. This equilibria is established with a smaller cluster size. The joint cycle was not as good and was proven to not be a true hypercycle.

This happens because target-only reproduction forces all skill-elements to be maintained in an organization, whereas joint reproduction does not preserve all skills except by chance. Thus altruistic catalysis is more effective than "fairness" in reward in maintaining complex technologies.

Higher-Element Hypercycles

Like the 3 element model

* there is assymetry between source only and target only cycles

* joint systems do not preserve the full hypercycle

* odd number element schemes require paired coupling to survive

* "live"ecologies usually involve multiple skillled individuals at their core with parasitic single skilled people in the periphery

As you get more elements the likelihood of finding stable arrangement declines sharply.

Increasing the element volume increased the chances of finding a reproductive ecology by increasing the number of skllls per individual.

Interestingly, stable systems are found above 5 elements in a spatial system but not in a non-spatial system. Locking auto-catalytic skll sequences into physical space -- that is, into embodied individuals in interaction -- seems absolutely crucial for complicated skill sequences to emerge and be dynamically stable.

Conclusion

Dynamic barriers ot technological complexity can be transcended one global reproductive variation and behavioral "memory" are created via physical spatial arrangements. Social technologies of action-reaction rule sequences produce individuals as they wend themselves back and form across sites. The "seeding" of work into social interaction among distinct locational units (phenotypes) permits both more complex-skill sets and more multi-faceted indivdiuals to become alive.

Thus altruism is more crucial to the life of dynamic firms than most social scientists believe, not because the selfish sacrifice for the collective good, but because of the reproductive logic of chemistry.