Key Concepts

The ideas AoM is built from — each with what it means, how it connects to the others and why, and where in the book it’s argued. The page is meant to stand on its own: a reader who follows the links should be able to put the framework together for themselves, in their own order, without reading the book first. Linked terms have entries of their own; terms in italics are treated in the chapters and essays named. Alongside the framework’s own ideas are shorter entries for terms from neighboring fields — ethics, systems theory, biology — kept to say how AoM uses them. The page grew out of the book’s glossary and keeps that page’s address.

action

Something an agent or system does that changes its own state or its environment. Every action is an irreversible physical event, so it always leaves some trace in the world that another observer could, in principle, read — though that trace becomes information only for an agent equipped to read it.

actor

Anything that performs an action — but not necessarily an agent. An actor may act with no intention, learning, or stake in the outcome (a falling rock is an actor); its actions leave traces in the world, but those traces carry no meaning and serve no purpose unless the actor is also an agent. (Contrast agent.)

adaptability

The capacity of actors within a system to influence resilience or adapt in response to challenges or opportunities.

agency

What lets a system act on its own behalf: perceive, interpret and predict, and then act for what it values, with the methods it has — its values-model and methods-model refined as it goes by contact with its environment of interaction. Agency comes in degrees, at any grain — a cell, a person, a group acting as one — and it grows: as an agent takes in what reaches it, its models tend toward coherence over a wider context of meaning and a wider scope of effectiveness. We understand it by what a system does, so nothing ties it to one material or one grain (functionalism). (Contrast actor; see agent, self-maintenance.)

agent

An actor that has agency: one that maintains itself — making and keeping its own boundary, which is to have a stake — and that not only acts but takes in what reaches it from its environment of interaction, refining by that contact what it values (values-model) and what works (methods-model), and so behaves in a purposeful, future-oriented way. The bacterium climbing a gradient is the minimal case; a person, or larger collective structures, far larger ones. (Contrast actor.)

Arrow of Morality (AoM)

This book, and the framework it sets out: morality as a direction rather than a destination — coherence over a widening context. AoM is used throughout for the framework itself, not only for the volume.

boundary

In AoM, the boundary of a self-maintaining process: the membrane a cell makes and keeps, and whatever does the same work for any agent, in any material. The agent’s own activity makes it — it isn’t given in advance — and it lasts only while that activity does. The self isn’t what sits inside a boundary; the self is the maintaining, and the boundary is what the maintaining produces (see self-maintenance, self).

The boundary lets things through, and the agent regulates what crosses it: sealed, it starves; too open, it spills; and a boundary whose openings can be adjusted is an agent that can shift its own position within that range. The boundary is what makes a self and an other. Everything the agent meets is other — and since an agent has no direct view of itself, what it notices of its own doing reaches it as other too (see reflexive closure). Where the boundary falls is enacted by what the agent does, including its real part in larger agents, so the extent of a self can move: say “we won” and the boundary has taken in a team (see self, coherent pluralism). Argued in the Prologue (“A cell keeps itself going”) and Chapter 5 (“Selves made of selves”).

circularity

An account that depends on what it accounts for. AoM is circular in several places, and says so: its criterion is itself an agent’s assessment and is held to its own standard; what an agent values is known through what it enacts; agreement on what matters is enacted before it’s agreed; each of these entries is defined through the others. The usual objection assumes every circle is vicious. A circle is vicious only when it’s frozen and sealed — closed in a single moment, with nothing coming in from outside. Unrolled in time and disciplined by contact with what the agent meets, a circle is the ordinary shape of every self-maintaining process: a cell that makes the parts that make it; a dictionary whose words are defined by one another and anchored by readers who learned some of them by pointing; a science that refines its methods by their results; a self that knows its values by what it does. Each comes back to where it was, but later, and changed by what arrived in between. Seen end-on, such a circle looks closed; seen along the axis of time, it’s a spiral.

The test is the test for the counter-dynamic, seen from the other side: a circle is sealed exactly when what arrives is discounted, so the echo chamber is a sealed circle and a sound account is an open one. Reflexive closure is the special case where the circle passes through the agent’s own model of itself. Minimal realism supplies the discipline: something outside the circle pushes back. Self-maintenance is an open circle over time — the same form carried on through changed instances. Judging a process by one moment, asking whether it’s coherent now, is reading the spiral as a circle; that’s why neither coherence nor the difference between widening and capture has a snapshot verdict. Recursion in computing and self-reference in logic are its relatives; there a base case or a grounding does the work that contact does here.

coherence

How well a model’s parts line up: with each other, and with what the agent effectively perceives. Coherence belongs to a model, not to a structure holding together, and it’s what lets an agent make sense of its situation and act reliably. It isn’t one measure. An agent’s values-model coheres over the contexts of meaning-making it meets, and its methods-model over the scopes of effectiveness it acts in; the two are assessed separately, and an agent can advance on one alone. Skill isn’t goodness.

Taking more in gives a model more to line up, so coherence and context can rise together or pull apart. Widening first exposes incoherence that had been kept in separate rooms; then integration repairs it. A temporary loss of coherence is the friction of taking something in, not a sign of narrowing. Coherence that rises as context widens is the direction agents tend to assess as increasing morality (coherence over a widening context); coherence kept by discounting what arrives is the counter-dynamic. There’s no single score for coherence, and one would favor the sealed system, which has the most internal coherence because it discounts the most (see morality). Argued in Chapter 3 (“When the model holds together”) and Chapter 4 (“Getting the rooms together”). See increasing coherence, scope.

coherence over a widening context

The direction AoM describes: what an agent values and how it acts growing more coherent while the context it takes in widens rather than shrinks. Distinct from plain coherence (a property of a model at a moment) and increasing coherence (the refining process), both of which it builds on. Its reverse — coherence bought by narrowing — is the counter-dynamic. Written IC² in shorthand, from the earlier formulation “increasing coherence over increasing context” — the two I-C pairs, squared. (See increasing context.)

coherent pluralism

The shape a group tends to take when its reach outgrows what one center can keep coherent: a network of networks, with coherence kept inside each node and the nodes joined loosely. The center is thin in content, not in reach. It keeps up the conditions under which the nodes can differ, not what they hold; the edges between nodes stay open; and nodes can leave, which is what keeps the center thin.

It’s the answer to the cost of coherence at scale: reconciling every member with every other grows faster than the group does, so coherence is kept local and composed (see nesting). Each node meets part of the world, so the form that makes a large coherence affordable is also how the group’s context widens, and it keeps elimination from outrunning context — preserving the variety that agents’ better answers come from. It reads differently at two grains: at the group’s grain it’s an agent with a thin values-model, acting as one in little more than keeping those conditions; at the nodes’ grain it’s what supports their agency. Neither description cancels the other. Principles emerge along the edges connecting the nodes. A group that one center can keep coherent doesn’t need the form. Argued in Chapter 5 (“The cost of getting bigger”), Coherent Pluralism, Coherence at Scale, and Part II. Contrast the counter-dynamic, which keeps coherence by discounting instead.

collective agency

A collective agent — the book’s we — is an agent made of agents. It has a grain of its own, and at that grain it enacts values and methods of its own — some shared with its members, some emerging at the group’s grain and held by no member — and in neither case a sum of theirs. Its members are those whose models of what matters and what works converge enough with one another’s for the group to act as one; membership is enacted, follows from the models, and is a matter of degree. Shared models alone don’t make one act: acting as one takes coordination — members’ enactments brought together through shared methods, communication and trust — and coordination is itself enacted.

Nesting is the form it tends to take as its reach grows. It isn’t a mind over and above its members: both its enactment and its contact pass through them, so any mismatch between what’s done and what’s met has a group form — an act that exists at the group’s grain and that no member recognizes as ours; many hands, where the coordination was enacted but never perceived as one act. Argued in Chapter 5 (“Selves Made of Selves”), The We, More Than Its Notes, Coherence at Scale and Coherent Pluralism.

consciousness

What an agent reports from the inside — the first-person side of being an agent, as its own self-model presents it. The report is real as a report; what AoM declines is the assumption that it gives direct access to experience, since the self-model, like every model, is made from somewhere (see reflexive closure). AoM treats consciousness not as a second source of worth set beside agency, but as the inside view of the perspectival, indexical self that agency already requires: a perspective that has folded back to include itself among the things it models — the observer that is also the observed. The weight it carries is real, but graded along the same axis as agency — a richer interior to value with, and a richer perspective for other agents to perceive and integrate — not a separate standing a being carries on its own. (Distinguish sentience, the narrower capacity to feel; and the minimal, unfelt value of a self-maintaining cell. Placed in full in the Supplemental essay The View From Inside; on why standing is conferred by the assessing agent rather than riding on consciousness, see Standing, and the Widening Circle.)

consequences

The actual effects an action produces — intended and unintended, near and far, many of them hidden from the agent in advance. In AoM consequences are never the measure of what is good (see consequentialism). They are one part of what reaches an agent from its environment of interaction — the part that follows from acts, its own or others’ — and so one part of the contact by which its models are refined.

consequentialism

The family of views that judge an act by its consequences; utilitarianism — the greatest aggregate well-being — is its best-known form. AoM keeps the core insight, that consequences are where value actually shows up, but declines the fixed, view-from-nowhere ledger: “good outcomes” are always good for agents, over some context and some reach of concern, and the full consequences of anything are mostly hidden in advance. What AoM tracks is not a fixed quantity of good produced but the direction of the process — concern widening or narrowing. Reframed in Chapter 6.

constructivism

The position that an agent’s knowledge, meaning, and values are built — actively put together out of its contact with the world — rather than found lying ready-made in it. One of AoM’s three epistemological foundations, alongside perspectival realism and functionalism. But made isn’t arbitrary: constructions are built against a shared world that pushes back (minimal realism) and are disciplined by what works and by what the agent meets, so a model can still be plain wrong, and better and worse constructions can be told apart. (See meaning-making; developed in Chapter 2, “made is not arbitrary.”)

contact

Whatever reaches an agent from its environment of interaction — the part of the world it acts in and gets acted on by — by any route at all: the effects of its own acts, the effects of other agents’ acts, and what simply arrives. It’s what an agent’s values-model and methods-model are refined by. The consequences of the agent’s own acts are one part of it, not the whole (consequences). And contact isn’t the same as perception: perception is the agent’s reading of what reached it, and whatever the reading left out is contact all the same — met later as surprise or failure, and at the limit as the check on staying viable (self-maintenance). That’s how a bacterium, which reads very little, still gets corrected. Refinement is contact taken in; the counter-dynamic is the same contact discounted. Minimal realism says that something corrects the model, never what; contact is the name for its arriving. Contact has a reach, and nesting changes how it crosses an interface: modeled rather than met.

context

What an agent effectively perceives, as it arrives: the input to its sense-making (what’s going on) and meaning-making (what matters), and what it uses to integrate and refine its values-model and methods-model. Context is processed as it arrives, not stored; what persists of it is in the models. Strip the context away and you don’t reach a purer meaning underneath: meaning is always made within one.

Context increases in two ways. With the agent’s surface of interaction: the larger the surface, the more may be effectively perceived — though a larger reach brings perceptions the agent’s models may not yet recognize. And with the models themselves: refined models recognize more of what arrives, so meaningful interaction tends to increase context. What an agent doesn’t effectively perceive can’t correct its models (minimal realism). Argued in Chapter 2 (“In context”), Chapter 3, and the Prologue (“Faster models”). See increasing context, coherence over a widening context, counter-dynamic.

convergence

The tendency of agents’ values and methods — and so their assessments and their oughts — to come into agreement as their context widens. Agents share a physical inheritance, and biological agents an evolutionary one, and so far as they meet the same world they share a context; together these give them a common basis for agreement, and without some shared context they couldn’t relate at all. What makes agreement tend to follow is synergy: coordination yields what conflict gives up, and as agents effectively perceive this, what they value and how they act tend to converge. The book calls the shape this makes the tree of agreement.

Convergence is enacted before it’s agreed. A larger agent’s values include what its members effectively converge on in what they do together, and values that emerge at the group’s grain and are held by no member; conscious agreement is one form convergence can take, not the thing itself. It isn’t absolute: it can be assessed as meeting a threshold sufficient for acting together, which is what lets a we act as one. Conflicts that can’t be reconciled at the surface of agents’ values can often be reconciled at the deeper, more widely shared values beneath them, which is one way convergence proceeds. Convergence is the group face of normativity, not its source (ought). Agreement reached by narrowing — by discounting what has reached the agents — is the counter-dynamic, not convergence. Argued in Chapter 3 (“The old roots”, “How conflicts actually end”), Chapter 6 (“How strangers agree”), and The Tree of Agreement.

cost of coherence at scale

As an agent’s reach grows, the work of keeping its models coherent grows faster than its reach does, because what has to be reconciled is relationships among what it meets, and relationships multiply faster than the things related. For an individual, the relationships are among what it values, among its methods and the situations they must work in, and between what it expects and what arrives. For a group, they’re among its members’ models: what each member values and does has to fit well enough with the others’ for the group to act as one. The cost is the same kind at both grains, and from outside the two look the same. The same growing surface of interaction that brings the cost also brings the combinations that yield synergy; both come with widening.

An agent can meet the cost in two ways. It can do the work of integration, revising its models until what has reached it fits. For a group at large scale this tends to take the form of nesting: coherence kept within parts and composed where the parts meet, so that no part has to reconcile everything with everything; coherent pluralism is this form for a group whose reach outgrows one center. Or it can keep coherence by discounting what arrives: an individual by not registering what doesn’t fit, whether evidence or a conflicting value of its own; a group by not registering what its members bring to it. The second way is always cheaper in the moment, which is why the counter-dynamic is available at every scale. That the work outruns reach is a tendency under a condition — that an agent’s capacity to integrate grows more slowly than the surface it must integrate — not a law. Argued in Chapter 3 (“When the model holds together”), Chapter 5 (“The cost of getting bigger”) and Coherence at Scale.

counter-dynamic

Coherence kept by discounting what has already reached the agent. When the work of holding a model coherent outruns what the agent can integrate, there are two ways to restore coherence: integrate and refine, or stop taking in. The second is always the cheaper, and it’s the counter-dynamic: the worldview made airtight by discounting whatever would disturb it — the cult, the echo chamber, the hardening ideology. Chapter 6 calls it the arrow running backward. Its context needn’t shrink; it’s enough that what arrives no longer counts.

Its cost is that what’s discounted can’t correct the agent’s models, which then drift from what the agent meets (minimal realism). It’s the reverse of coherence over a widening context, and it’s what lets an agent tell coherence reached by widening from coherence reached by narrowing without appeal to any fixed creed. It happens at every scale: in a person who discounts an inherited drive rather than re-cohering it, in an institution, in a civilization (see Moloch). Because the cheaper road is always open, it can be out-competed but never abolished. It’s most reliably recognized from inside, since only the agent knows what has reached it: is this coherence kept by discounting something that has already reached me? Not attending to what hasn’t yet reached an agent isn’t the counter-dynamic. Argued in Chapter 3, Chapter 5, Chapter 6 (“The arrow runs backward”) and Coherence at Scale.

diversity

In the systems sense (not the social-policy one): the variety and variation within a system that supplies the raw material for adaptation and exploration. Diversity is what lets a system try many things and discover what works; too little starves it of options. AoM treats a healthy spread of perspectives and approaches as part of what keeps coherence growing rather than calcifying. (See resilience, pluralism.)

drives

Deep, largely pre-conscious motivational pushes rooted in how an agent came to be — for living things, their biology and evolutionary heritage: hunger, fear, the pull toward others. The rawest layer beneath the felt and the valued; see values for how the layers compare.

duty-ethics (deontology)

The view that an act is judged by its adherence to duties or rules, regardless of outcome — keep your promises; do not treat a person as a mere means. AoM honors what this gets right: a good rule is compressed coherence, hard-won wisdom so reliable across contexts that we are usually wise to follow it without recomputing from scratch. But it locates the error too — mistaking that compression for bedrock, and forgetting the rule is answerable to the coherence it summarizes and revisable when the territory changes. Reframed in Chapter 6.

dynamic balance

A state of equilibrium characterized not by stasis, but by ongoing negotiation, adjustment, and fluctuation between the competing forces of unity and diversity.

effectiveness

The degree to which an agent’s methods actually bring about the change they aim at — whether “what works” works. Effectiveness isn’t judged against a fixed yardstick but by results in the world, across the scope of contexts a method has to serve. It’s what the methods-model’s coherence is measured over, as context is what the values-model’s coherence is measured over: a method grows more effective as it’s refined against feedback over a widening scope. (See methods, scope.)

emergence

The appearance, when simpler parts combine, of capacities and properties that none of the parts has alone — water’s wetness from hydrogen and oxygen, value from a cell’s self-maintenance, a self from the activity of many. It recurs at every level the book traces: synergy throws up the new, selection keeps what holds together, and each emergent layer becomes the platform for the next. (Prologue; Ch 1.)

emotions

Fast, embodied appraisals that register how a situation bears on what an agent values — value-detection felt as bodily signal (fear, anger, joy) that orients attention and readies action. See values for how the layers compare.

enacted

Done, rather than represented: carried in what an agent does, whether or not it holds any model of it. AoM pairs the word with modeled: what an agent does is enacted; what it makes of what it does and meets, in a model, is modeled — and the modeled side is always partial and open to correction. Everything an agent does is enacted, deliberate choices among the rest, and what it does expresses its nature, which its values-model and methods-model partly determine; the principles and preferences it states are models of that, built afterward, by the agent or by others, from what it does.

The pair keeps coming back. Minimal realism is enacted in self-maintenance, not held as a belief. Membership in a larger agent is enacted by real participation; “I” and “we” are how an agent models itself when it reflects. Convergence is enacted in what agents do together; conscious agreement is one form it can take. As agents’ context widens, what they do tends toward increasing coherence over increasing context; “moral” and “ought” are agents’ assessments of acts, on the modeled side (morality, ought). No agent has direct access to what it enacts, which is why the modeled side stays partial (see reflexive closure). Enacted isn’t the same as unconscious: a choice can be fully deliberate and still show more of what an agent values than the agent can say.

entropy

A count of the microscopic arrangements that fit a given rough description of a system — its temperature, its pressure, where its boundaries lie — and the quantity the second law says tends to increase, because a system left to itself drifts to the descriptions that more arrangements fit. It is not a measure of mess. A structure is a constraint that fewer arrangements fit, and it can form and last wherever energy flows through it and raises the count elsewhere; so structure is not an exception to the second law but one of the ways it plays out (Prologue; Ch 1). (The count is relative to the chosen description, and objective once that description is fixed; this is where the strict thermodynamic sense and the information-theoretic one meet.)

eudaimonia

Flourishing understood as a trajectory rather than a prize: living well in the ongoing sense of an agent whose values grow more coherent across a widening context, and its methods across a widening scope, together, not the reaching of a fixed final state. AoM borrows the classical term but reads it as a direction of travel — in keeping with morality as a direction, not a destination.

EvoDevo

A systems-theory framing in which every complex self-organizing system — cosmic, biological, cultural — is shaped by two opposing but complementary forces:

Evolution (the tree): unpredictable, branching exploration that throws up novelty, variety, and local adaptation — the engine of possibility (genetic mutation, cultural invention, market competition).
Development (the funnel): predictable, convergent constraint that guides systems toward hierarchical levels of complexity and functional maturity — the engine of probability (star formation, an embryo building a heart, the emergence of complex intelligence).

extensibility

Of an account: that it applies to cases that weren’t in view when it was made, without changing what it claims. AoM’s extensibility follows from functionalism: agents are understood by what they do, so the account reaches group and artificial agents, and kinds not yet met, by the same reasoning it applies to a cell or a person. It’s also the form the account’s own growth takes — refinement by contact rather than replacement — and so an instance of what it describes. (See agency, circularity.)

feelings

The consciously experienced side of emotions — what an emotion is like from the inside, once it surfaces into awareness. See values for how the layers compare.

function

What a thing does — what it affords, connects to, makes possible or forecloses. To understand something, on AoM’s view, is to grasp its function: to see its edges, where its powers begin and end. Function comes first; essence, if there is any at all, a distant second. (See functionalism.)

functional abstraction

A description that picks out part of a system by what it does, not by what it’s made of or where in the system the work happens. An agent’s values-model and methods-model are functional abstractions: names for work spread through the agent — in a cell’s chemistry, a brain’s activity, or a machine’s learned parameters — not components housed anywhere inside it. The difference between them is one of function, not of parts. An abstraction of this kind holds at the grain at which it explains: no one can point to a values-model, but an agent is far harder to predict or understand without one. (See functionalism; cf. Daniel Dennett, “Real Patterns,” 1991.)

functionalism

The commitment to understanding things by what they do rather than by some hidden inner essence — to grasp a knife by its cutting, a mind by what it accomplishes, a value by the work it does for an agent. One of AoM’s three epistemological foundations, alongside perspectival realism and constructivism. Its key consequence: if a thing is what it does, the same role can in principle be filled by very different stuff — so minds and agency needn’t be tied to any one material. This is what makes the account extensible: understood by what they do, agents turn up at every grain and in every material, and the account reaches kinds of agent not yet met without changing what it claims. (See function, functional abstraction; developed in Chapter 2, “seeing the edges,” and Chapter 4 on methods.)

good

An agent assesses an act as good so far as it perceives the act as promoting its values in the present situation, with no demand for coherence beyond it. This is the first form of value, the one that arrives with self-maintenance: good or bad for this agent, in its situation, not as a cosmic fact. Good, right-in-principle and moral are one kind of assessment taking in progressively more context, not three stages; good takes in only the situation at hand. Argued in Chapter 1 (“A candle and a cell”) and Chapter 6 (“The three old answers”). Compare right-in-principle and morality, and see the example under morality.

grain

The level of description at which a statement holds. A flock is one thing at the grain of the sky and many things at the grain of the birds, and neither description cancels the other. In AoM the same goes for agents: an individual and a group look the same from outside and different from inside; a larger agent acts as one at its own grain while its members stay agents at theirs; coherent pluralism is an agent at one grain and supports its nodes’ agency at another; selection “against what fails” and selection “for what works” are one process described at the grain of a single interaction and at the grain of a lineage over time. So too persistence and continuity: at the grain of an instance, a self-maintaining form persists for a while and ends; at the grain of the form, it continues through its instances, and it’s continuity, not persistence, that its activity tends toward (self-maintenance). Persistence is often taken to be the most basic thing conserved; that’s a statement at the first grain. Questions that assume one grain must be the right one — is a we really an agent? is coherent pluralism a level of agency or something supporting it? — dissolve once the grain is named.

A grain isn’t simply chosen. A member’s grain is enacted by its real part in a larger agent, though in the agent’s own reflection it’s “chosen”; and an agent’s model of itself has closure at the grain at which it describes itself, which is what reflexive closure means (see also self). AoM’s levels of agency are a carving of a gradient at a grain useful for the account, with some stretches steeper than others, not a set of phase changes in the world. Grain-relativity isn’t relativism: each description still answers to what the agent meets (minimal realism), and a grain is a way of describing one thing, not a license to describe anything. Argued in Chapter 5 (“Selves made of selves”), Coherent Pluralism, and the Prologue (“Wholes made of wholes”).

homeostasis

A system’s holding of its internal conditions within the narrow range it needs to keep functioning, against everything the environment does to push it elsewhere — the body holding its temperature, a cell its chemistry. In AoM it is the minimal form of an agent acting on its own behalf: the self-maintenance from which value first arises. But regulation alone is not self-maintenance. A system can hold its conditions in range with nothing within it maintaining what it is — a thermostat-governed room, or the earth system, whose surface has stayed habitable for billions of years as a by-product of the living things within it, each maintaining itself. Homeostasis is an agent acting on its own behalf only where it serves a self: the cell, not the climate. (See agency, self-maintenance; Ch 1, the cell against the candle.)

identity

What makes an agent the same agent over time, despite the constant turnover of its materials and states. AoM locates identity not in a fixed substance but in the continuity of a process — the ongoing activity of a self maintaining and rebuilding its own boundary and models. You are not the stuff; you are the carrying-on. (See self-identity, nature; cf. Juarrero, Deacon.)

increasing coherence

The ongoing work of refining a model — values-model, methods-model, or picture of the world — by folding in new experience: adjusting, re-fitting, and restructuring it to reduce contradiction and fit it better to what the world returns, so it grows more robust and more widely effective across contexts of meaning and action. The process; coherence is the property it improves, and coherence over a widening context is what it becomes when the context widens as well.

increasing context

The process by which more of what reaches an agent comes to be effectively perceived, and so enters its sense-making and meaning-making and the refining of its models (see context). Increasing context and widening context name the same thing throughout the book. Context doesn’t pile up: it’s processed as it arrives, and what persists of it is in the agent’s models. What increases is how much of what arrives the agent effectively perceives, in two ways: a larger surface of interaction, as when agents whose values and methods effectively converge form a larger agent with a reach no member holds; and models that recognize more of what arrives, as the agent’s interactions refine them.

Group agency holds at two grains. At the grain of the larger agent, its context is its own, not the sum of its members’. At the grain of a member, the member stays an agent, and its context widens to take in what it perceives of its part in the larger whole. Increasing context is described by what’s effectively perceived, not by whether the result is better; that’s for agents to assess (see morality). Argued in Chapter 2 (“In context”), Chapter 3, Chapter 5 and the Prologue (“Faster models”). See coherence over a widening context.

information

Not raw data but difference that makes a difference to an agent — a pattern that shapes its capacity for coherent, purposeful engagement with the world. The same physical signal can be mere noise to one agent and rich information to another, depending on what each can detect and what each has at stake; information is therefore relational, not a stuff the world simply contains. Every action, being an irreversible physical event, leaves a trace that could inform some observer — but a trace becomes information only for an agent equipped to read it. (Compare meaning: information bears on what an agent can do; meaning is how it matters.)

intelligence

An agent’s capacity to build and refine models that work — to perceive, predict, and act effectively across a widening range of contexts. On AoM’s functional reading, intelligence is not a fixed quantity or a human monopoly but a matter of degree and kind: how well, and how widely, a system can make sense of its world and bring about valued change in it. (See agency, methods, effectiveness.)

intentionality

An agent’s capacity to direct its attention and action toward something — an object, a goal, an outcome it represents to itself — as steered by its values and carried out through its methods. The “aboutness” of a mind: the way an agent’s states are of or about things in its world. (See purpose, agency.)

meaning

The significance an agent constructs by interpreting a perception, symbol, or experience in light of its values — how something matters to it, given its purposes and its situation. Meaning is not a property sealed inside objects (the way mass is) but relational: it lives in the meeting between a thing and an agent that cares, and it is always made within a context. Strip away every caring agent and you are left not with a world of unread meanings but with bare patterns and no meaning at all. (Distinguish significance, the degree and direction of mattering; and sense-making, getting right what merely is.)

meaning-making

The active process by which an agent renders its world significant — sorting perceptions and events into what is worth caring about, and how much, by reading them through its values within a context. AoM deliberately pairs it against sense-making (rendering the world intelligible — getting right what is going on): meaning-making feeds the values-model, sense-making feeds the methods-model. The two lean on each other but are not the same work, and an agent can be strong in one and weak in the other.

meaningful growth

Growth wider and more whole at once: an agent’s values growing more coherent over an increasing context, via methods growing more coherent over an increasing scope, the two together, at any scale of agency, from a single agent to groups of any size. Agents assess growth as meaningful by their values; what AoM describes is the tendency of what agents do, as their context widens, toward increasing coherence over increasing context, and the convergence of their assessments of it.

It’s distinct from accumulation: growth in size, reach or power without growth in coherence over context. The paradigm is methods extending an agent’s reach faster than its context grows — the hand outrunning the eye. Accumulation isn’t itself the counter-dynamic; it becomes the counter-dynamic only when the coherence is kept by discounting what arrives. It’s also distinct from ordinary progress, which may widen what an agent can do without changing what it values. Meaningful growth has no ceiling and no finish, and AoM reckons it as distance from an origin, not toward a destination: a destination would need a future context no agent yet has, while how far coherence and context have grown from where they began can be assessed now. Argued in Chapter 7 and Meaningful Growth. See eudaimonia, coherence over a widening context.

metaethics

The branch of philosophy that asks not which acts are right but what moral claims are — whether values are discovered or made, what “good” refers to, whether moral judgments can be true. AoM’s theory (Part 1) takes up these questions: it relocates value from a fixed property of the universe (and from mere human projection) to something that emerges with self-maintaining agents and is refined through coherence over a widening context. (Contrast normative ethics, which asks what to do.)

methods

An agent’s working knowledge of what gets the job done — the skills, strategies, and know-how for bringing about valued change, where what counts as “valued” is set by its current values. Methods are the practical side of an agent: how it actually moves from a less preferred state toward a more preferred one. (See methods-model, effectiveness.)

methods-model

An agent’s working model of what works for bringing about valued change, within a scope of effectiveness. With ongoing experience the methods-model is continuously updated and refined, and, as the agent’s reach widens, tends to grow more coherent over an increasing scope of effectiveness. What counts as working is set by the values-model: methods are set going by a mismatch between what’s perceived and what’s valued, and both models are revised by what the attempt turns up. Like the values-model, it’s a functional abstraction — the difference between the two is one of function, not of parts. (For what the model is a model of, see methods.)

minimal realism

The least that AoM’s realism claims: that there’s something a model doesn’t contain, and that it corrects the model. It isn’t a conclusion an agent reaches or a hypothesis it could set aside. It’s enacted in self-maintenance: to make and keep a boundary is already to be met by what one doesn’t control. A bacterium holds no belief about a world; it maintains itself against one. So minimal realism arises together with the least of agents, neither inferred from experience nor taken on faith. It applies wherever a self is maintained — a cell, an organism, an institution — and pushback from within a self, from its own constituents, is the world pushing back as much as pushback from beyond it. It says that something pushes back, never what: every account of what pushes back, every world-model, stays open to revision. Like reflexive closure, it’s a claim about knowing, not about value. (See perspectival realism, constructivism.)

Moloch (collective action traps)

The book’s name, after Allen Ginsberg’s Howl and Scott Alexander’s Meditations on Moloch, for the trap many agents spring together against the wish of each: the fishery emptied, the arms built that no one can afford to be the last to build — catastrophes with no author. Game theorists call it a collective action or multipolar trap. In AoM’s terms the agents in it usually share a great deal of context; what they don’t share is a self. Each values goods arriving to its own self, and soon. The shared consequence of what they all do is discounted on both counts: it lies in the future, so it’s perceived as less certain, and it isn’t expected to fall directly on the agent that acts — which is what an externality is, and the counter-dynamic in its group form. The gain is private and the loss shared, so an agent that alone holds back bears the whole cost for a fraction of the benefit and is removed from the field; widening alone is selected against, which is why preaching fails. The way out isn’t a hand from outside that adds a cost — that leaves the selves as they were — but a we wide enough that the consequence arrives to the same self as the gain: convergence is how such a self forms, and coherent pluralism its form at scale. Argued in Part II (Held Together, Held Apart) and Moloch, Formally; the structure of why narrowing is downhill is in Coherence at Scale.

morality

AoM doesn’t define what is moral. “Moral,” said of an act, is always an agent’s assessment, and what AoM describes is how agents assess, and how those assessments tend to change and to converge as context widens. The criterion, which is also the book’s thesis:

An agent assesses an act as increasingly moral to the extent that it perceives the act as promoting present but evolving values, increasingly coherent over an increasing context of meaning-making, via present but evolving methods, increasingly coherent over an increasing scope of effectiveness — with the values-model and methods-model continuously refined by contact with the environment of interaction.

Put less completely: agents assess an act as more moral the more they perceive it as promoting values coherent over a wider context, by methods coherent over a wider scope.

The criterion has two axes: what the agent values, coherent over a context of meaning-making, and how it acts, coherent over a scope of effectiveness. Coherence isn’t one measure across both; skill isn’t goodness. The values and methods are the agent’s present ones — as they are, and as they change under contact with what the agent meets — not an ideal set. The assessment is comparative and open-ended: no act is moral full stop, and there’s no finish line. As agents’ context widens, what they do tends, two steps forward and one back, toward increasing coherence over increasing context, and their assessments tend to converge (convergence); that tendency, which agents come to assess as increasing morality, is what the book’s title names. Context and coherence are described without reference to what’s better, so the criterion doesn’t assume what it’s meant to assess. Compare good and right-in-principle, the same assessment over less context; see increasing context, ought.

An example. A friend has asked for help moving house on Saturday; on Friday a better offer comes up. Assessed as good, the question is only whether keeping the promise serves what the agent values right now — and the offer is tempting. Assessed as right-in-principle, the question is whether keeping it fits what the agent values and how it acts, coherently, within the situation as it understands it, held still — and an agent who values being someone whose word holds will say yes. Assessed as increasingly moral, the question widens: the friend’s plans built around the help, what breaking the promise would do to their trust and to the trust their friendships run on, what kind of agent many such choices make. The assessment doesn’t climb from one rung to the next; it takes in more of the context, and it moves as the context does.

nature

What an agent characteristically values, and how it tends to act on those values across different contexts. Know an agent’s nature and its circumstances, and its behavior becomes largely predictable. A nature is shaped both by what the agent is born with and by how it has developed. (See values, values-model.)

nesting

The shape integration tends to take when an agent’s reach outgrows what one center can reconcile: coherence kept within parts and composed where the parts meet, so that no part has to reconcile everything with everything. Each part is an agent at its own grain and a constituent at the next, both at once; grain-relativity is what lets that be said without contradiction. Its price is in how contact crosses an interface: what one part receives of what another met is a compressed account — a report, a price, a bearing read out over a phone — a model standing in for the contact, and what the model leaves out shows up, if at all, as consequences at the group’s grain, later. Nesting is a tendency the cost of coherence produces, not a design imposed: the polycentric commons and the delegated organization are forms observed, not prescribed.

It’s how the cost of coherence at scale is met without discounting what has reached anyone. Thinning what crosses an interface is nesting; discounting what has crossed is the counter-dynamic; from outside the two look alike, and the sealed interface — a part that stops registering what its neighbors send — is the counter-dynamic in nested form, with the metric gamed in place of what it measured as its institutional sign. Nesting spreads contact as well as enactment: a nested group perceives through its parts, which is how its contact keeps pace with its reach at scale. Coherent pluralism is nesting for a group whose reach outgrows one center: thin at the center in content, not in reach. It isn’t hierarchy in the sense of command — composition where parts meet isn’t control from above; nesting is a form found at every grain where the cost is met by integration rather than by discounting. Argued in Chapter 5 (“Selves Made of Selves”), Coherence at Scale and Coherent Pluralism (cf. Elinor Ostrom’s polycentric governance and Stafford Beer’s variety engineering).

normative ethics

The branch of moral philosophy that asks which actions are right or wrong and what sort of person one should be. AoM offers no standard of right or wrong of its own. It describes the process by which agents form and revise their standards, and the direction in which that process tends: increasing coherence over increasing context. Each traditional theory fixes one standard (consequences, rules or character), which AoM reads as a frame held still: what agents find right-in-principle at a given context, revised as their context increases. (Contrast metaethics; see also process ethics.)

ought

AoM doesn’t derive an ought from an is; Hume’s gap is left where he drew it. Nor does it deny that oughts exist. It relocates them: normativity, the pull of an ought, is enacted in agents’ valuing. An agent already acts to promote what it values, by the methods it has. An ought is that agent’s assessment of what to do, and its pull comes from the valuing, not from a fact about the world or a rule outside the agent. It isn’t the pull of any one value. The assessment is driven by the agent’s values-model — everything that matters to it and how those things fit together — and shaped by its methods-model, what works, in context: what the act would bring back from everyone and everything within the agent’s reach. A single strong drive is one input to that assessment. An ought that follows it against the rest of what the agent values, and against what its methods would bring back, is coherence kept by discounting: the counter-dynamic turned inward.

Because the ought lives in valuing, it changes as the valuing does. As an agent’s context widens, its oughts change as its values and methods grow more coherent over it; and as agents’ values and methods converge (convergence), so do their oughts. This holds at two grains. For a single agent, the ought lies in its own valuing, refined by what it meets. For a we, agents joined into a larger agent, normativity tends to follow its members’ convergence, which is the group face of the same thing, not its source. So AoM issues no oughts of its own: it describes how agents’ oughts arise and tend to change, and an agent will tend to act to promote its values whether or not it’s told to. There can be a logical must; there’s no moral one. Argued in Chapter 6 (“Where the ‘ought’ comes from”) and The Is–Ought Relocation.

perspectival realism

The position that knowing is always from somewhere and still of something. Perspectival: every account of the world is made from a situated position, shaped by an agent’s senses, history, concerns and conceptual scheme. There’s no standpoint outside all relations — no place to stand and describe the world from nowhere — because whoever describes it is always among what is described (reflexive closure). Realism: something the agent’s models don’t contain corrects them (minimal realism); the world doesn’t depend on any one perspective, and it pushes back. Different perspectives meet what corrects them, and so far as they meet the same things they can be tested against one another and can be more or less accurate about what they engage with (convergence). So it rejects both the “view from nowhere” and pure relativism: whatever there is to know is met only from inside relations, and that meeting is how knowledge gets corrected. One of AoM’s three epistemological foundations, alongside constructivism and functionalism. (See perspective, Umwelt.)

perspective

The situated vantage from which an agent meets the world — the particular window, fixed by its senses, history, needs, and context, through which alone it has any access to reality. There is no view from nowhere; every model is somebody’s model. To say knowing is perspectival is not to call it arbitrary: different perspectives open onto one shared world and can be tested and corrected against it. (The keystone of perspectival realism; see also Umwelt.)

pluralism

Referring to the variety in perspectives, identities, values, knowledge systems, approaches, and lifestyles, including epistemic diversity, cultural pluralism, and value pluralism.

preferences

Comparative leanings between options that fall out of an agent’s values in a given context — values expressed as ranking, this over that. See values for how the layers compare.

principles

The consciously held, articulable distillations of an agent’s deeper values — “justice,” “fairness,” “freedom” — compressed names for key aspects of the underlying complex. See values for how the layers compare.

process ethics

An approach that locates the moral center of gravity in the ongoing process — how choices are reached, revised, and learned from — rather than in a fixed outcome or rule. AoM is a process ethics in the fullest sense: morality is not a possession but an activity, the continual refinement of what an agent values and how it acts across a widening context. The interpersonal practices often linked with the term (open communication, collaboration, mutual understanding) are one expression of this at the scale of people; the same process runs at every scale, from a single agent to a whole culture. (See morality, coherence over a widening context.)

progress

Directional change in some respect — technical, economic, in capability or reach. Progress in this ordinary sense need not bear on what agents value: it can widen what agents are able to do while what they value stays where it was, or narrows. For the change AoM describes, the book speaks of meaningful growth. And because that growth is open-ended, AoM reckons it as distance from an origin, not distance to a destination: an expected or imagined destination would need a future context no agent yet has, while how far coherence and context have grown from where they began can be assessed now. (See meaningful growth, counter-dynamic.)

purpose

An agent’s built-in orientation toward ends — the end-directedness that lives in how its values-model and methods-model are organized, inclining it to act so as to reach and hold the states it models as valuable. Not a goal stamped on it from outside but one that arises from how the agent is put together. (See teleonomic, teleologic, intentionality.)

reach

How far an agent’s interaction with the world extends: what reaches it and what it can act on, in space, in kind and in time. A bacterium meets the world a few thousandths of a millimeter at a time, over seconds; a tree across meters of soil and air, over centuries; an animal with eyes at a distance, before anything arrives. Each level of organization reaches farther, and a group acting as one has a reach no member holds.

Reach isn’t context. A larger reach brings more perceptions, not necessarily effective ones; context is what the agent’s models actually take up. As reach grows, the agent’s surface of interaction grows faster than reach itself (the possible combinations and relationships multiply), so two things grow faster than the agent: the openings for synergy, and the work of keeping coherent what it meets (the cost of coherence at scale). Methods extend reach, and can extend it faster than the agent’s context grows; Part II calls this the hand outrunning the eye, and it’s accumulation rather than meaningful growth. Care depends on reach as well as on value: two things an agent assesses as equally valuable get different care if one is within its reach and the other isn’t (standing). Reach extends in time as well as space, which is how care extends to descendants (Caring Across Time). Argued in the Prologue (“Faster models”), Chapter 1 (“The outward reach”), Chapter 5, Part II (“The reach of our hands”) and Coherence at Scale.

reflexive closure

The recognition that whoever describes the world is part of what’s described: one more set of relations among the rest, never a vantage point outside them. Every account is therefore from somewhere. A “view from nowhere” would be a relation to everything from no position at all — not a view that’s hard to reach, but no view. Closure is what makes AoM’s realism perspectival, and it joins two things often held apart: models are constructed, because they’re made from a position, and disciplined, because that position is itself in the world that pushes back. Closure here means there’s no outside, not that self-description is ever complete: an agent describing itself still does so from somewhere, so some blind spot always remains. It’s a claim about knowing, not about value; nothing about what matters follows from it alone. (See perspectival realism, perspective, constructivism; cf. Humberto Maturana, “Anything said is said by an observer,” and Carlo Rovelli’s relational quantum mechanics.)

resilience

The capacity of a system (social, ecological, epistemic) to absorb disturbances and reorganize while retaining essential function, structure, identity, and feedbacks

right-in-principle

An agent assesses an act as right-in-principle so far as it perceives the act as promoting its values, coherent over a given context of meaning, via methods coherent over a given scope of effectiveness — the context held still. Coherence is now asked for, but inside a frame that isn’t allowed to grow.

This is the assessment that consequentialism, deontology and virtue ethics each make: each holds one frame still and takes it for the whole of morality. AoM keeps the assessment and drops the claim that it’s the whole; moral is the same assessment over an increasing context. A right, in the sense of an entitlement, is a different thing: a construction a we makes and can grant, withhold or revise, where right-in-principle is an assessment any agent makes of an act (see Rights, and the Widening We). Argued in Chapter 6 (“The three old answers”). Compare good and morality, and see the example under morality.

scope

The breadth of contexts across which an agent’s methods stay effective — how widely “what works” actually works. Methods coherent over a narrow scope succeed only in the situation they were tuned for; methods coherent over a wide scope carry across many. Scope is the methods-side counterpart to context on the values side, and the criterion pairs them: an agent assesses an act as increasingly moral as it perceives the act promoting values coherent over an increasing context of meaning-making, via methods coherent over an increasing scope of effectiveness (see morality). (See methods, effectiveness.)

self

A process considered as the one doing the maintaining: something that holds its own boundary, models a world, and sorts that world into what helps it and what harms it. There’s no substance underneath the activity — you are not the stuff; you are the carrying-on (see identity) — but there is a position: the place from which mine, here, and at stake have anything to point at (see consciousness).

Two things about selves run through the rest of the book. They nest: cells make up bodies, bodies make up families, crews, institutions, nations — selves made of selves — and the bigger one is real, not just a way of talking. A beehive does what no bee does; a city acts in ways no citizen intends; a research field knows things no scientist holds in one head. And the boundary is a setting rather than a fixture: where you stop, and how far your we reaches, can move, and can grow. Say “we won” and it has taken in a team of strangers; a soldier dies for a country, a self that includes far more than one body.

Because a self is a shape and not a kind of stuff, the story doesn’t stop at biology. Teams, institutions, and capable enough artificial systems can take that shape too.

(Distinguish self-identity — the agent’s model of itself, which can be wrong about any of this — and self-maintenance, the activity by which a self is a self at all. Worked out in Chapter 5, Selves Made of Selves.)

self-identity

An agent’s model of itself — the part of its world-model that represents the agent to itself: its boundary, its history, and what it takes itself to value. Where identity is the fact of continuity, self-identity is that continuity modeled from the inside; it can be more or less accurate, and it shapes how the agent acts. (See identity.)

self-maintenance

The activity of a process that has to act to keep itself in being — and, in AoM, the door value comes in by. A candle flame is self-sustaining: it holds its shape far from equilibrium, draws in its own fuel, even steadies itself after a puff of air. A living cell does something more: it makes and keeps its own boundary, the membrane marking where it ends and the world begins, and it acts on its own behalf, steering toward what sustains it and away from what destroys it. With that, the world divides in two, help and harm, and value appears: not painted on by an observer, but arising in something with a stake, in whatever material does this. The flame maintains a process; it doesn’t maintain a self. The self-producing organization that makes the difference is autopoiesis, “self-making” (Humberto Maturana and Francisco Varela, Autopoiesis and Cognition, 1980). Maintaining a self and having a stake are one fact, not a sequence, and the least of agents begins with them.

The boundary a self-maintaining process keeps lets things through, and it can fail two ways: sealed off from exchange, the process starves; too open, it spills; a membrane whose channels open and close is a process that can shift its own position within that range. What the activity tends toward reads at two grains: at the grain of the instance, help and harm are weighed against this cell’s continuing; at the grain of the form, self-maintenance tends toward the continuity of the form through its instances — repair, reproduction, the lineage — not the persistence of any one. Self-maintaining things rarely just hold station: the same restlessness that sends the bacterium up a gradient sends them to probe the edges of what their position makes possible, and that reaching is where context begins to widen. Two measures shouldn’t be confused. Help and harm are weighed against the self’s continuing; better and worse are assessed by an agent according to the criterion (morality), coherence over a widening context. They can come apart: a process can keep itself in being superbly while discounting everything that would widen it (until its world changes); see counter-dynamic. Argued in Chapter 1 (“A candle and a cell”) and the Prologue (“A cell keeps itself going”). See identity, homeostasis, minimal realism; cf. Evan Thompson, Mind in Life, on the flame-versus-organism contrast.

sense-making

The work of rendering a situation intelligible — getting right what is going on (that the shape in the grass is a snake, that the ice will hold). AoM deliberately separates this from meaning-making, which renders a situation significant (what matters). The two lean on each other but are not the same: a con artist makes excellent sense of you while making nothing of your welfare. Sense-making is the line that runs, two chapters on, into the methods-model; meaning-making runs into the values-model. (Used more narrowly here than in enactivism, where “sense-making” covers what AoM calls meaning-making.)

sentience

The capacity to feel — to have experiences with a positive or negative tone, such as pleasure and suffering. AoM distinguishes it sharply from the minimal, unfelt value of a self-maintaining cell (which sorts help from harm without feeling anything) and treats sentience as a later, richer development that weighs heavily in moral concern — not as a standing the subject emits on its own, but as it is perceived and valued within some agent’s widening context. (See consciousness; Ch 1‘s caution that to value is not yet to feel.)

shared values

Encompassing collective goals, social cohesion, common frameworks, ethical consensus, and coordinated action necessary for societal functioning. (identified as commonality at some meaningful level of the hierarchy)

significance

How much, and in what direction, something matters to an agent — the weight and likely consequence it attributes to a perception or event, read through its values within a context. Significance is the direct output of meaning-making: where meaning is that something matters, significance is how much and to what end — the quantity that lets an agent rank and choose.

standing

Whether something counts at all in an agent’s valuing. In AoM, standing is conferred: anything an agent counts has standing as that agent counts it, in degrees, because value is always relative to some agent and there’s no view from nowhere from which standing could be read off what’s counted. There’s no difference in kind among what can be counted — human or artificial agents, non-human animals, trees, rivers, a language, the environment — and no special channel, such as consciousness, by which something gains standing. Agents tend to be assessed as of more value than non-agents, but not necessarily.

Conferred doesn’t mean optional. What has already reached an agent can’t be dropped from its count without discounting what has arrived, and that’s the counter-dynamic; so the same move that makes standing conferred is what lets agents assess cruelty and exclusion as narrowing rather than as a permissible drawing of the circle. Agency bears on whose acts the criterion assesses, not on what can be counted. Standing answers only whether something counts. How much it counts (worth) is a further assessment by the same agent, and how much it counts to that agent in practice (care) depends on reach as well as on value: care starts close, and as context widens the circle widens without flattening. Agreement among agents affirms standing; it doesn’t ground it. Contrast equal consideration of interests, which grades from nowhere. Argued in Chapter 5 (“Why you love your child most”), Chapter 6, and Standing, and the Widening Circle.

synergy

The combined effect of parts acting together that none could produce alone — an outcome that is genuinely new, or simply greater than the sum. Synergy is the engine of emergence and a key driver of evolution: from molecules to cells to societies to ecosystems, things that combine into a working whole gain advantages that selection can favor. (See emergence; Ch 1.)

teleonomic

Apparently purpose-like behavior that is really the product of evolution and mechanism, not of any goal held in mind — the heart that is ‘for’ pumping blood without intending anything. In a part, such as the heart, the purpose is only apparent; in a whole that maintains itself, it is enacted and the agent’s own, even when no goal is held in mind — the bacterium climbing a gradient. Represented purpose grows out of this. (Contrast teleologic.)

teleologic

Pertaining to genuine ends — behavior directed by a represented goal an agent actually holds, not merely the as-if purpose of a teleonomic process. AoM is interested in how the second grows out of the first: how purpose that is enacted becomes purpose that is also represented. (Contrast teleonomic; see purpose, intentionality.)

Umwelt

The part of its surroundings that an agent’s senses and needs make available to it — the whole of what it can detect and act upon, and so the only world it has. A tick’s Umwelt has three signals; a dog’s is a vast landscape of scent; ours feels seamless and total but is itself a narrow band, assembled by the brain into what we take for the world. Every Umwelt feels complete from the inside; none is the world entire. (The on-ramp to perspectival realism; Ch 1.)

values

What matters to an agent, and how much, relative to its stake: the many-sided, naturally layered and fine-grained complex that meaning is drawn from and that moves the agent to do what it does in any given context. Value arrives with self-maintenance, in any material, and isn’t the same as feeling. In AoM, “values” always refers to the effective elements of the agent’s values-model — its actual, working model of what matters, always present but evolving — never to values as fixed contents in the everyday sense. “An agent’s values” names what we imagine as discrete elements of that model; “values” is often used for the model itself. The two ways of speaking name one thing.

Values are layered. For a person, food and shelter sit closer to the core than music; within music they might lean to rock over classical, and within classical to Bach over Beethoven. The deep values are few and widely shared, the shallow ones many and often in conflict, which is why a conflict that can’t be settled at the surface can often be settled underneath (see convergence). Contact with what the agent meets corrects the values-model. An agent has no direct view of its own values-model: it watches what it does, as others do, and what it learns becomes part of its model of the world (see reflexive closure, enacted). So an agent can’t simply state its values. Its principles — “justice,” “fairness,” “freedom,” “love” — are consciously recognized distillations of parts of the underlying complex, models of it built afterward. (How the neighbors compare: drives, emotions, and feelings are the felt and motivating layers; values are what matters; preferences and principles are values expressed as ranking and as spelled-out rules.) Argued in Chapter 3 (“Not a list but a model”, “The deep and the shallow”).

values-model

An agent’s working model of what matters to it and how much, layered and fine-grained: the process that does the work of valuing, always present and always evolving. “Values” names its effective elements, or the model itself; the two ways of speaking name one thing (see values). Contact with what the agent meets corrects it. The agent has no direct view of it, and learns of it as it learns of anything else — from what it does (see reflexive closure). Like the methods-model, it’s a functional abstraction: a name for something done throughout the agent, not a part located in it. As agents’ context widens, their values-models tend to grow more coherent over it (coherence over a widening context).

virtue-ethics

The view that locates morality in the character of the agent — the virtuous habits and traits from which good action flows — rather than in rules or outcomes. AoM reads character as an agent’s values-model and methods-model grown stable: the settled dispositions that make good action reliable. It honors the insight while marking the limit — real virtuous people disagree, and no virtue stands available to umpire the others — which is why character still needs the wider process to adjudicate. Reframed in Chapter 6.

world-model

An agent’s settling record of what the world turned out to be like — and, in AoM, a derived structure rather than a third defining one. Agency is defined first by self-maintenance, and then by its present but evolving values-model and methods-model, each of which constrains the other. The world-model is the sediment of those two models’ working against the world: the building record of what resisted, laid down wherever value-directed action was surprised. What the agent notices of its own doing enters it too, since the agent meets itself as other (see boundary, reflexive closure); and it’s where what persists of context, which is processed as it arrives and not stored, comes to rest. So it isn’t something agency is defined by but something agency continuously builds — which is why modeling the world is sense-making, the line that runs into the methods-model, and never a third axis of its own. (See self-identity — the part of the world-model that represents the agent to itself.)