Every system model requires a boundary — a decision about what is inside the system and what is treated as exogenous (outside the system, impinging on it as an input but not governed by it). No real system is truly closed; the question is always where to draw the line, and that decision is consequential.
Meadows returned to this point repeatedly in her teaching: choosing the system boundary is not a neutral technical act. It is a theoretical and political choice that determines what causes are considered and what causes are invisible. A model of a city's traffic system might draw its boundary to include roads, signals, and drivers — but exclude land-use patterns, housing prices, and the economic forces that determine where people live and work. The model will then find solutions within the variables it includes, and miss the more powerful leverage points outside them.
The critique works both ways. Boundaries too narrow miss important feedback: environmental economists who excluded ecosystem services from economic models systematically underestimated the costs of growth. But boundaries too wide produce unmanageable complexity and introduce speculation where data is thin. groping-in-the-dark-1982 was partly a reflection on what the mit-system-dynamics-group had learned about this tradeoff in the years after limits-to-growth-1972.
limits-to-growth-1972 drew its boundary at the global level — a controversial choice that was both the model's greatest strength and a source of criticism. By including the full feedback between population, capital, food, resources, and pollution at global scale, it captured dynamics that nation-level models missed. Critics argued the model ignored the capacity for substitution, technological change, and price signals within the boundary. Meadows and her collaborators responded in beyond-the-limits-1992 and limits-to-growth-30-year-update-2004 with updated analyses that retained the global scope while refining the internal structure.
A practical implication Meadows drew: whenever a system seems to be behaving strangely — producing unexpected outcomes, resisting intervention — one diagnostic question is whether the relevant feedback loops are actually captured within the model's (or the analyst's) boundary. "Policy resistance" often occurs because the corrective loops are outside the boundary of the policy-maker's mental model. The intervention changes something, but the change triggers a response from the larger system that was not in view.
Meadows also noted that boundaries shift over time and across actors. A corporation draws a boundary that includes its own profits but excludes its pollution. A regulator draws a boundary that includes the regulated industry but may exclude the communities affected by it. Different stakeholders draw boundaries differently because their interests and perspectives differ — and the resulting system models embed these differences as structural assumptions. Making boundaries explicit and examining whose interests they serve is part of what Meadows meant by iceberg-model thinking: surfacing the structural and paradigmatic assumptions beneath the visible behavior.
The sustainability-institute and balaton-group work extended this to indicator design: what is inside the boundary of what we measure determines what we see, and what we see determines what we manage.