FRONTIER · FORMALISATION

Option Space — A Candidate Formalisation

This is the keystone work from step 1 of the v0.2 push plan. It responds directly to Lens 4's sketch (reachable configurations under bounded exergy budget) and Lens 6's measurability attack. It is NOT the final formalisation — it is a candidate written so you have something concrete to react to rather than a blank page.

Status: Draft. Needs your voice, your judgment, and probably at least one mathematically-literate critic's eyes before promotion.

Format on promotion: Probably either (a) a new essay chapter titled "The Measurement Question" or (b) an appendix titled "Option Space: A Candidate Formalisation." Option (b) keeps the main essay accessible and gives interested readers a place to go. My instinct: (b).

Round 3 — Observer-relative + chess-moves additions (2026-04-29)

A companion draft now sits at drafts/observer-relative-option-space-and-chess-moves.md. It does not replace this draft — it supplements it. Three specific updates worth flagging before reading the rest:

  1. Open question #5 (the observer). This draft flags "Option space for whom?" as open. The companion gives it a structural answer via Wolfram's Observers Like Us: the relevant slice is R_living(C, B, T) — reachable configurations in which life-persisting observers continue to exist and make distinctions. "For whom" becomes "for the class of life-persisting observers," which is principled rather than arbitrary.

  2. Open question #1 (computing R). This draft notes that R is computationally intractable. The companion partially dodges the problem by shifting from global ω(C) evaluation to local Δω(move) evaluation. Most policy and configuration changes are moves evaluable directionally — closing or opening R_living at given horizons — even when the global measure isn't computable.

  3. The §"What this doesn't solve" measurability concession. That section concedes intractability honestly. The companion strengthens the response: global ω is non-computable, but local Δω is evaluable directionally. The Lens 6 attack lands on the global frame; the chess-moves frame is a separate (local) mode of evaluation that the attack does not touch.

The four-candidate-measures section below still holds — it covers global ω comparisons. The chess-moves frame is a separate mode, not a replacement. Both modes live alongside each other.

The forest-vs-monoculture worked example below gets reframed in the companion as a move evaluation (clearcut closes R_living at multi-decade horizons) rather than a state comparison (forest dominates monoculture in the partial order). Same direction, different frame.

This draft retains structural precedence; the companion is supplementary until both are reviewed for promotion.

Why a formalisation is needed

Lens 6's most urgent attack (#2 in the ranked list): "Option space has no unit, no measurement procedure, no error bars — yet you propose it replaces GDP. Give me the unit."

Lens 3 independently: "A dissertation defending option space would need a non-circular formal definition, a worked example, and a bridge to existing operationalisable concepts (Sen's capabilities, real options, Assembly Index)."

Lens 1: If CE rides on DP §4.2, option space is the forward dual of DP's trajectory cost functional.

All three converge on the same point: without a formalisation, the central concept is rhetorical.

The option-space-measurability proposition (added 2026-04-23, status open) is the correct admission. But the admission opens a gap. This draft proposes filling it — not by claiming the gap is closed, but by staking a candidate formalisation the field can attack.

The setup

Let:

The reachable set from C given budget B over horizon T:

R(C, B, T) = { C' : C' is reachable from C using exergy ≤ B in time ≤ T,
               under physical and biophysical constraints,
               and under some set of admissible human decisions }

This set is finite in principle (discrete physical states, bounded by entropy considerations) but astronomically large in practice.

The option space measure

The reachable set R is a mathematical object. To get a scalar "option space measure" ω(C, B, T), we need to reduce R to a number (or, more honestly, a partial ordering over configurations).

Several candidate measures. Each has strengths and failure modes.

Candidate 1: Cardinality

ω₁(C, B, T) = |R(C, B, T)|

Count reachable configurations.

Candidate 2: Diversity-weighted

ω₂(C, B, T) = Σ_{C' ∈ R} w(C', R)

where w(C', R) weights each reachable configuration by its distinctness from the others.

Candidate 3: Assembly-weighted

ω₃(C, B, T) = Σ_{C' ∈ R} AI(C')

where AI(C') is the Assembly Index (Cronin) of C'.

Candidate 4: Compositional

ω₄(C, B, T) = sup { Σ_i L(Cⱼ) : {Cⱼ} ⊂ R is an achievable trajectory }

where L is some local utility or "livingness" measure.

Recommendation

Start with ω₃ (Assembly-weighted reachable set). Reasons:

  1. It's the measure most directly supportable by Distinction Physics (distinctions-as-primitive gives a path to counting construction steps).
  2. It's the measure that makes the CE/AT bridge genuinely novel rather than rhetorical.
  3. It's the measure that handles the Lens 6 "forest vs database of trees" case correctly — the forest's high-AI future states (complex ecosystems) weight more than the database's low-AI future states (more files).
  4. It's formally definable even if currently non-computable, which is more than can be said for the current rhetoric.

Treat ω₃ as the current candidate, explicitly open, with the expectation that the right measure will emerge from empirical work and mathematical refinement.

The partial ordering

Whatever ω is chosen, value comparisons should be read as a partial order, not a total one:

C ≽ C' iff ω(C, B, T) ≥ ω(C', B, T) for all relevant (B, T)

Some pairs (C, C') will be incomparable — C is better under some budgets and horizons, C' under others. This is a feature, not a bug. It matches the intuition that there is no single "best" configuration across all contexts.

Implication for P22 (viable-objective): The viable-objective ratio is not a scalar to be maximised but a partial order to be respected. "Maximise durable flourishing per unit of bounded throughput" reads as scalar; honest form is "prefer configurations that dominate in the partial order, accept tradeoffs where they don't."

This is a substantive revision to P22. Consider.

A worked example: forest vs. industrial-agriculture configuration

Configuration A: Diverse forest ecosystem

Configuration B: Industrial monoculture on same land

Informal ω₃ comparison:

R(A) contains many high-AI future configurations: the complex ecosystem states, the multi-species human-settlement states, the adaptive-to-climate-change states. Many of these would require substantial construction history if starting from scratch — they're Assembly-heavy.

R(B) contains fewer high-AI configurations. Most reachable states are low-AI: another monoculture, a degraded field, a simpler plantation. The clearcut has narrowed the reachable set in a way that exergy alone cannot restore — some pathways require stored biological information (seed banks, mycorrhizal networks) that was destroyed.

ω₃(A) >> ω₃(B) — the forest dominates on the partial order.

This is the kind of comparison CE needs to be able to make rigorously, not just illustratively. The example shows the direction.

What this doesn't solve

Unresolved issues

  1. How to compute R(C, B, T). In principle: a search over physical trajectories. In practice: computationally intractable for systems of interest. Requires approximation, heuristics, or bounds.

  2. How to compute AI(C') for socio-economic configurations. Cronin's AI is defined for molecules via construction graphs. Extension to institutions, skills, infrastructures requires a generalised notion of construction. Possibly: minimum number of decision-steps (in some primitive alphabet) to build the configuration. This is speculative.

  3. How to handle stochasticity. Many reachable configurations are possible only probabilistically. Expected option space vs worst-case option space are different measures. Risk-aversion becomes a parameter.

  4. How to handle human agency. R(C, B, T) depends on what decisions humans make. This is either a feature (option space respects agency) or a bug (it collapses into subjective choice). Probably: condition on "admissible" decisions — those compatible with physical survival and some minimal ethics — and measure R over that restricted set.

  5. How to handle the observer. Option space for whom? Humanity? A particular economic actor? The biosphere? Different observers have different reachable sets. The framework should own this.

Items to flag as open

All five above should be explicitly open questions. Honest scope for v0.2 is: "here's a candidate formalisation, here's what it resolves, here's what it doesn't." Not: "we have a measure."

Connection to DP §4.2 (the optional deepening)

Per v0.2 direction, DP is offered as optional vocabulary, not required grounding. For readers who want it:

DP §4.2 defines a trajectory cost functional over distinction-space. The forward dual of that functional is precisely R(C, B, T): the set of trajectories starting from C, bounded by cost B, reaching time T. Option space is then a measure on that reachable set.

This is the Lens 1 bridge. It's available. It's not required. Readers without DP exposure can read the formalisation as pure non-equilibrium thermodynamics + ecological economics; readers with DP exposure see the additional structure.

Status as a proposition

Currently option-space-measurability is the only proposition covering this. It's open, which is correct.

After promotion of this draft, consider:

What to do with this

  1. Read it. Probably with skepticism. It's my best articulation of Lens 4's sketch, not yours.
  2. React. Which candidate measure feels right? Are the open questions the right ones? Is Assembly-weighted the move, or is there a better candidate I haven't considered?
  3. Promote as either appendix or new chapter once it feels like your voice. Expect substantial rewriting.
  4. Cite it from P3 (value-option-space) and P22 (viable-objective) once promoted.

The measurability attack is fatal to the current draft. This is the draft that makes it non-fatal. Worth time.

Length note

This is ~300 lines. On promotion, probably 2-4x that, with worked examples, clearer math, and your framing. An appendix of 1000-2000 lines is defensible; this is the kind of technical depth academic readers will specifically look for to judge the work.

The main essay stays accessible; the appendix carries the technical weight.