Holland observed that tagging operates at multiple levels simultaneously. An agent's primary tag determines its broadest category of interaction. Secondary tags create finer-grained patterns within that category. Tertiary tags determine specific interaction details. The layered hierarchy allows systems to be organized at the macro level while remaining flexible at the micro level — a structural requirement for genuine adaptation.
The Berkeley study documented what Holland's framework predicts: when AI tools enter a functionally-tagged organization, the functional tags begin to dissolve. Engineers build interfaces. Designers write code. Product managers prototype end-to-end features without engineering or design. The translation costs that once required functional mediation collapse, and with them the interaction boundaries the organization was built around. Delegation decreased. Boundaries blurred. The tags that had structured the organization's interaction patterns were being rewritten by the tools' capabilities.
Holland's framework specifies both the opportunity and the danger. Dissolving rigid functional tags increases interaction diversity, which increases potential for emergence. The engineer understanding backend systems who now builds user interfaces creates an internal interaction between domains previously mediated through lossy handoffs — and emergent solutions that bridge the domains become possible. The danger is that tag dissolution without replacement produces noise rather than productive diversity. Tags exist for reasons. The pre-AI functional tags were blunt instruments but they prevented the system from dissolving into a state where everyone does everything shallowly.
This is the organizational equivalent of Holland's edge of chaos principle. Too much tagging structure and the system is frozen — rigid silos preventing cross-pollination. Too little tagging and the system is noise. The productive zone is narrow and must be actively maintained. Vector pods — small groups whose job is to decide what should be built rather than to build it — are a retagging experiment that shifts the primary tag from 'execution skill' to 'cross-domain judgment.' Holland would have noted that such retagging is continuous work, not a one-time design.
Holland identified tagging formally in Hidden Order (1995) as one of seven properties whose joint presence defines complex adaptive systems. The concept drew on his immune system studies with collaborators at the Santa Fe Institute and on the economics literature on price systems as information coordinators.
The concept's generality — applying with identical structural logic across biological, economic, and computational domains — was precisely Holland's contribution. Earlier traditions had studied markers in each domain separately; Holland showed they were instances of a single phenomenon.
Tags structure interaction space. They determine not what agents are but what they can do with each other.
Multiple levels simultaneously. Primary, secondary, and tertiary tags create hierarchical interaction architectures.
Tags evolve. In complex adaptive systems, the tagging structure itself is subject to adaptive pressure.
AI dissolves functional tags. When translation costs collapse, the interaction boundaries built around them lose their structural function.
Retagging is continuous work. The appropriate tagging structure for AI-augmented organizations is not a fixed architecture but an adaptive process.