If instead of treating self-sustainment and non-domination as settlement-internal properties we treat them as properties of a whole Earth–space system—where off-world sites, launch providers, and Earth jurisdictions share or trade life-support, legal authority, and emergency capacity—how does this system-level lens change which combinations of locations (orbit, Moon, Mars, asteroids) look most justifiable to scale up, and does it reveal any architectures (for example, highly interdependent orbital–lunar–Earth grids) that are ethically worse than their parts because they lock humanity into fragile or easily captured interdependencies that settlement-level metrics would miss?

Answer

System-level thinking generally favors modest, modular orbital–lunar combinations and small robotic/crew asteroid nodes, and makes large, tightly coupled Mars-heavy grids look ethically weak and fragility-prone.

1. How a system lens reshapes which combos look justifiable

• Orbit + Earth (no Moon/Mars)

  • Looks strongest: short latency, many independent providers, easy oversight.
  • Good for: services, experiments, limited habitats.

• Orbit + Moon + Earth (modular)

  • Still acceptable if: multiple launch/comm providers, stockpiles and backups on each side, clear legal splits.
  • Best role: small lunar industry/logistics feeding mostly Earth and some orbital habitats.

• Orbit + Mars + Earth (no Moon)

  • Weaker: long latency, narrow launch chains, high exit cost.
  • Justifiable only at small scales and with strong caps; not as a core grid.

• Orbit + Moon + Mars + asteroids (dense grid)

  • Ethically risky if life-support, power or governance are tightly shared.
  • Safer if each node can degrade to safe-mode with local buffers and slow trade, not collapse.

• Asteroid nodes

  • Best as mostly robotic with small, rotating crews.
  • System value: materials, science; not population cores.

2. Architectures that become worse-than-sum

Ethically weak patterns:

• Single chokepoint stacks

  • E.g., one mega-operator controlling launch, comms, and critical spares for orbit + Moon.
  • Risk: one failure or capture endangers all nodes; residents have no real exit or voice.

• High mutual dependence without buffers

  • E.g., orbitals depending on lunar power; lunar bases depending on orbital water/air; neither holding long buffers.
  • Local metrics might look fine, but system can’t tolerate a major link loss.

• Governance capture grids

  • Shared corporate or national control across nodes, with arbitration, audits, and courts all aligned to that controller.
  • System-level domination even if each settlement has nominal local councils.

• Mars-as-hub architectures

  • Using Mars as an early central hub for population or governance (e.g., Martian council with veto over cis-lunar rules).
  • Distance and exit costs make any abuse or secession hard to remedy.

3. System-level metrics to prefer or avoid

Prefer combos where, at system scale:

• No single node or provider is a hard single point of failure.
• Each major node can survive months in isolation without mass casualties.
• Legal and oversight authority are plural and cross-checking.

Avoid combos where:

• Large populations depend on one fragile off-world segment.
• Earth’s poorest or least powerful are structurally locked out of voice or benefit while bearing non-trivial shared risk.
• Emergency aid or exit from any major node depends on a single political or commercial actor.

4. Net effect on scaling decisions

• Easiest to justify scaling: orbital habitats and infrastructure, with multiple providers and strong Earth oversight.
• Cautiously justifiable: limited lunar industrial/logistics sites tightly coupled to orbit and Earth, but not to Mars for core life-support.
• High concern: architectures where Martian cities, or dense Mars–Moon–orbit trade grids, become central to human survival before they are independently safe and well-governed.

So a system-level view shifts ethics toward small, redundant, weakly coupled networks, and flags strongly interdependent orbital–lunar–Mars grids as potentially worse than their individually acceptable parts.