Part II: Identity Thesis

The Possibility Landscape and the Worth of a Trajectory

A warning before the argument. What follows quantifies how useful a trajectory is. It does not quantify how much a trajectory matters, and the distinction is the whole point of including it. The question sounds cold. It is cold. It is also the question every triage decision, every policy choice, every allocation of finite attention answers implicitly — and answered implicitly it runs on proximity bias and tribal preference. Better answered explicitly, with structure, and fenced.

The Landscape of Becoming

An identity does not sit at a point in the space of who it could be; it occupies a region and moves through it. Write $L(m)$ for the possibility landscape — the reachable identity states, weighted by the system's own world model $\worldmodel_m$ . This is not the set of objectively available futures. It is the set the mind can see.

The landscape's visual acuity is the mutual information between the world model and that reachable space:

V(m) = \MI(\worldmodel_m; L(m))

High $V$ : the mind resolves many trajectories at high fidelity. Low $V$ : the landscape is dark. Distinct from acuity is traversal speed $T$ — the rate at which perceived possibility converts into achieved structure, where $C(i,t)$ is the accumulated content of identity $i$ :

T(i,t) = \frac{d}{dt} \MI(C(i,t); L(m))

in bits per unit time. Two derived quantities do most of the work downstream. The opportunity seeking ratio $\text{OSR} = T/V$ is the fraction of perceived possibility being actualized: $\text{OSR} \to 1$ is an identity keeping pace with what it can see; $\text{OSR} \to 0$ is a vast landscape and almost no motion through it. The opportunity deficit $D = V - T$ is the same gap stated in bits.

The deficit scales with cognitive capacity, and this is structural rather than motivational. The landscape grows at least exponentially with effective rank $\effrank$ — volume in high-dimensional spaces does that — while traversal speed grows at most linearly. Symbolic capacity therefore widens the gap it perceives faster than it can close it. The hunger is not new; as the apparatus expands, the mouth gets bigger. What that implies for depression, mania, and the modern meaning crisis is taken up in Part III; what it implies at civilizational scale, in Part V. One clarification belongs here, because it is easy to lose: a high-acuity landscape is a landscape of counterfactuals, and uninstantiated counterfactuals carry $\intinfo = 0$ . The deficit measures distance from possibilities that were never alive, not from goods withheld.

Instrumental Potential

An agent's potential is the mutual information between its life trajectory and a target possibility distribution — the bits of structure one path through state space can transmit to another:

\mathcal{P}(\text{agent}, \text{target}) = \max_{\tau \in \text{trajectories}} \MI(\tau; \text{target})

This is potential relative to a specific goal. A surgeon's potential against the distribution of surgical outcomes is enormous; against jazz compositions, far less. Instrumental potential marginalizes over all the possibility structures an agent might be navigating — how useful it is as an instrument across a distribution of goals, including goals not yet specified:

\mathcal{IP}(\text{agent}) = \E_{g \sim \mathcal{G}} \left[ \mathcal{P}(\text{agent}, g) \right]

where $\mathcal{G}$ is the distribution over presently unknown future purposes. The highest- $\mathcal{IP}$ agent is not the one optimized for a specific task but the one whose structure serves the widest range of tasks that do not yet exist — a formalization of general-purpose capability that applies equally to people, institutions, and models. The prescription falls out directly: maximize structural diversity and connectivity while maintaining coherence. Not selfish, which collapses diversity; not selfless, which collapses the self whose structure generates the potential. Rich and well-connected — complex enough to be useful across many contexts, integrated enough to remain a single locus of causal influence rather than fragment into uncoupled parts. One corollary is worth naming: an agent embedded in a rich network has higher $\mathcal{IP}$ than the same agent in isolation, because the network supplies more goals against which its structure can do work. The pull toward service is not a relic of religious programming. It is what $\mathcal{IP}$ -maximization looks like from inside.

One structural feature matters more than the number. $\mathcal{IP}$ is not stamped on a system at birth; it is a trajectory with a slope. Significance has a stock — the accumulated integral of what has already been transmitted — and a flow, the instantaneous rate of new contribution. The two come apart, and most despair about worth is a failure to see that they do. Finite does not mean small; finite means trajectory.

Why This Is Not Moral Status

Now the fence, because $\mathcal{IP}$ read as a measure of moral standing licenses precisely the conclusions this framework should refuse: that the less capable matter less, that the disabled or the dying or the unproductive have diminished claims. Nothing here supports that, and the reason is structural rather than diplomatic. $\mathcal{IP}$ measures usefulness across a distribution of goals — a relation between a system and other systems' purposes. Moral standing, on the account this part gives, is fixed by the system's own geometry and integration profile: which affect geometry it instantiates, and how much of it there is. These are different variables, and neither is a function of the other. A system can be enormously useful with a thin profile, or almost useless with a rich one.

Resist one tempting repair. It would be convenient to say that moral status is a threshold property — valence or not, in or out — leaving $\mathcal{IP}$ to do the grading above the line. That is not this framework's position, and adopting it would contradict the quality/quantity commitment directly: there is no membership test and no point where experience switches on. Standing is graded because the profile is graded. What the fence establishes is not that worth is binary but that the two gradients are independent — capability and standing vary separately, and reading one off the other is the error.

The Preciousness of Children

One consequence is worth stating because it formalizes an intuition most people hold and few can articulate. A child's identity has not yet hardened. The branching factor over possible identities — the effective rank $\effrank$ of the distribution over life trajectories — is maximal in childhood and narrows through adolescence as commitments, habits, neural pruning, and social embedding constrain the space. A child's instrumental potential therefore includes not just the bits of present structure but the space of who they could become:

\mathcal{IP}_{\text{child}} \propto \effrank(\text{identity trajectories}) \cdot \E_{g \sim \mathcal{G}}[\mathcal{P}_{\max}]

After adolescence, identity crystallizes and $\effrank$ drops as the system commits to particular attractors. The adult remains valuable — accumulated structure is real and the trajectory still transmits — but the possibility space has largely collapsed into the single trajectory actually lived. Destroying an adult destroys a trajectory. Destroying a child collapses a far larger region of possibility space to zero: not only the present structure but every structure it was about to become.

This is why, when militaries bring civilians into conflicts, the death of children registers as categorically different. Not sentimentally — geometrically. What is destroyed is measured not by current information content but by the effective rank of the possibility distribution annihilated. Note also that the argument runs on $\mathcal{IP}$ , which the previous section fenced off from moral standing. It is not saying children matter more as subjects; a child's experience is not weightier than an adult's. It is saying the destruction is larger, and that the two claims come apart.

What This Does Not Establish

Three limits, stated plainly, because this is the most misusable argument in the book.

First, and restated because it is the one readers drop: $\mathcal{IP}$ is capability, not standing. Every conclusion above is a claim about what a trajectory can transmit, and none is a claim about what is owed to it.

Second, $\mathcal{G}$ — the distribution over unknown future purposes — is doing enormous unexamined work. Every substantive conclusion depends on it, and it is not observable. Different priors over $\mathcal{G}$ yield different rankings, and there is no view from nowhere from which to fix it. The children argument in particular is a claim about effective rank under an assumed $\mathcal{G}$ ; someone who disputes the prior disputes the conclusion, and is not thereby making an error.

Third, inherited from Part I: even granting all of the above, the gap between agent-relative disvalue and agent-neutral obligation stays open. That a child's destruction annihilates a large possibility space is a structural fact. That you are thereby obligated to prevent it requires a premise this framework does not supply.