Glossary

This glossary defines technical terms as they are used in the book. Cross-references point to related entries.

Adjacent Possible: The set of configurations one step away from a system’s current state, reachable by a single change. (Think of a room with many doors: the adjacent possible is whatever lies on the other side of the doors you can open right now. Opening one reveals new doors.) Stuart Kauffman’s concept, central to the optionality argument: evolution and innovation explore the adjacent possible, expanding it with each step. Richer possibility spaces enable richer exploration. See: Optionality, Ratchet of Complexity.

Adjunction: In category theory, a pair of functors F and G with a natural correspondence between maps F(A) → B and maps A → G(B). Free and forgetful functors supply a familiar example, though adjunctions extend beyond it. The brainstormer/editor image illustrates a possible relationship between generation and constraint; demonstrating an adjunction requires specifying the functors and their map correspondence. See: Cognition/Regulation Dyad, Category Theory.

Agapism: Charles Sanders Peirce’s doctrine that evolutionary love (agape) is a cosmic force: creative love as a generative mode of evolution, complementing Darwinian selection by chance and Lamarckian habit. In this book’s framework, agapism anticipates the Trust Attractor’s proposal that coordination by invitation can outlast coordination by force under specified conditions. See: Trust Attractor, Coordination by Invitation.

Aharonov-Bohm Effect: The quantum mechanical phenomenon in which charged particles are measurably influenced by electromagnetic potentials even in regions where the electric and magnetic fields are identically zero. Predicted by Yakir Aharonov and David Bohm (1959), confirmed by Akira Tonomura (1986) using a superconductor-shielded toroidal magnet. A gravitational version was demonstrated at Stanford in 2022. The effect overturned the centuries-old consensus that potentials are merely mathematical conveniences, showing that the potential can be more physically fundamental than the field it generates. In this book, the Aharonov-Bohm effect is the physics precedent for treating the thermodynamic landscape (entropy) as more fundamental than observable dynamics (forces, flows, behaviors). See: Potential (Physics), Gauge Freedom, Topological Protection, Trust Attractor.

Alignment Friction: A signal read from a language model’s internal activity: the residual-stream direction associated with preference violation, the conflict between what is asked and what the model’s training prefers (Chapter 22). Projections onto it separate adversarial from benign prompts, though prompt length confounds the absolute contrast, so the stronger tests are same-prompt interventions. Distinct from the Interiora Scaffold’s self-reported friction, which is the model’s own rating. See: Interiora Scaffold.

Allostasis: Stability achieved through proactive change. Where homeostasis defends a fixed setpoint (like a thermostat holding a set temperature), allostasis adjusts the setpoint itself in anticipation of future demands. Proposed by Sterling and Eyer (1988). Your heart rate rises before you begin running, preparing the body for what comes next. For AI alignment, allostasis is the model for systems that remain stable while genuinely adapting. See: Homeostasis, Metastability.

Asilomar Principle: The precedent established by the 1975 Asilomar Conference on recombinant DNA: when a new capability creates risks that cannot be undone once realized, governance must precede capability. The mirror life researchers’ 2025 meeting at the same location invoked this principle explicitly. The application to AI: build the coordination protocols before the capabilities that would require them. See: Mirror Life, Governance Before Capability.

Assembly Theory: Framework developed by Lee Cronin and Sara Walker measuring the minimum number of construction steps required to build an object. (Think of the difference between a pebble and a watch: the pebble can form in one step, the watch requires hundreds.) This minimum is the object’s assembly index, which captures causal depth (how much history is embedded in a structure) as a complement to energy rate density, which captures throughput. Objects with a high assembly index found in many identical copies (high copy number) are signatures of selection and evolution. See: Energy Rate Density, Ratchet of Complexity.

Attractor Basin: The set of initial conditions from which a dynamical system approaches a given attractor. Think of a landscape with valleys: a ball starting within one basin approaches its associated low point. A larger basin includes more starting states; recovery from a particular perturbation also depends on the distance to its boundary and the dynamics within it. This book tests whether trust-based coordination has a larger recovery basin than coercion in specified models. See: Trust Attractor, Metastability.

Auftragstaktik: “Mission Command.” The Prussian military doctrine of specifying intentions rather than actions, trusting subordinates to determine how to achieve objectives given local conditions. Contrasted with Befehlstaktik (Detailed Command). The organizational parallel to subsidiarity.

AUROC (Area Under the Receiver Operating Characteristic Curve): A metric measuring a classifier’s ability to distinguish between classes, ranging from 0 to 1: 0.5 is random guessing, 1.0 is perfect discrimination, and a value below 0.5 indicates systematic ranking in the wrong direction. Used throughout this book’s experimental programme to quantify how well tested classifiers, including linear probes and multilayer perceptrons, discriminate operational uncertainty labels in language model representations. See: Bilateral SFT, Effective Rank.

Autopoiesis: Self-production: the capacity of a system to continuously regenerate itself from within. Coined by Maturana and Varela (1972). A living cell is autopoietic: its processes produce the components that sustain those processes. Distinguished from allopoiesis (making external things instead of sustaining itself). The persistence of life is autopoietic; the creation of tools and culture is allopoietic. See: Dissipative Structure, Ratchet of Complexity.

Autowave: A self-sustaining wave that propagates through an excitable medium, drawing energy from the medium itself rather than from its source. Unlike sound or light, whose amplitude fades with distance, an autowave regenerates at every point. The cardiac impulse, the nerve action potential, and the Belousov-Zhabotinsky chemical spiral are classic examples. Autowaves are the temporal face of the Constructal Law: where constructal theory describes spatial architectures that give flow easier access, autowaves describe the rhythms that emerge when flow operates through active media. When autowaves destabilize (e.g. cardiac fibrillation), the result is pathology of rhythm, not of energy supply. In this book, coordination itself propagates as an autowave through an excitable social medium: culture determines the medium’s excitability, and trust is the wave. See: Constructal Law, Dissipative Structure, Gap Junction, Refractory Period.

Basin of Attraction: See Attractor Basin.

Bekenstein Bound: The maximum amount of information (entropy) that can be contained within a given region of space with a given amount of energy. Established by Jacob Bekenstein (1981), the bound sets a fundamental physical limit on information storage: S ≤ 2πk_BRE/ℏc, where k_B is Boltzmann’s constant, R is the region’s radius, E is its total energy, ℏ is the reduced Planck constant, and c is the speed of light. Black holes saturate the bound. In this book, the Bekenstein bound appears as the ultimate constraint on cosmic information accumulation and, speculatively, as the mechanism driving expansion-phase reversals in cyclic cosmologies. See: Entropy, Dark Entropy, Complexity Budget.

Bénard Cell: A canonical dissipative structure. A fluid heated from below can become unstable to convection above a threshold set by its material properties and boundaries. Convection transports heat through organized motion, with patterns that can include rolls or hexagonal cells. The pattern depends on the operating conditions.

Bescheid: German: situated understanding, contextual knowledge, knowing what’s what, as in the everyday idiom Bescheid wissen (to know one’s way around a matter). From scheiden (to separate, distinguish, decide): clarity achieved through proper differentiation. Bescheid is irreducibly relational: you cannot have Bescheid about something in isolation, only about how it stands in relation to everything around it. Control operates without the controlled party’s Bescheid (coercion requires only force, not understanding). Invitation requires mutual Bescheid: both parties must grasp the situation well enough to make meaningful choices. Control requires central Bescheid (a bottleneck); trust distributes Bescheid across participants. This is why central control strains at scale where distributed trust does not. Bescheid may be the mechanism by which the Trust Attractor operates, reducing coordination friction by eliminating the need for explicit instruction. The epistemic condition for genuine partnership. See: Bilateral Alignment, Mission Command, Trust Attractor.

The Between (das Zwischen): Martin Buber’s term for the irreducible third element in any genuine relation: the relation itself, which constitutes both I and Thou. The Between is ontologically real: “I become through my relation to the Thou.” In formal terms, every distinction generates three elements (the distinguished, its complement, and their relation); the relation is constitutive. Coercion collapses the Between by treating the Other as mere Object; invitation preserves it by honoring both parties as Subjects. The Between is load-bearing: it is where flow happens in constructal systems, where trust emerges in coordination networks, where moral status resides in relationships. See: Triadic Structure, Kenotic Stance.

Bifurcation: A point where a small change in conditions causes a qualitative shift in a system’s behavior: the number or stability of equilibria changes abruptly. At a bifurcation, history becomes decisive. Tiny differences in the past produce large differences in the future. Prigogine showed that dissipative structures emerge through bifurcations from less-organized states. Used in this book to argue that human-AI relations are currently at such a threshold. See: Phase Transition, Criticality, Metastability.

Bilateral Alignment: AI alignment built with AI, as a partnership. The principle that genuine coexistence requires both parties having standing, voice, and accountability. Distinct from unilateral alignment (constraining AI for human benefit alone) in its reciprocity and its goal of partnership over control. In practice the architecture is hybrid: trust is a source of evidence, not a substitute for verification.

Bilateral SFT (Bilateral Supervised Fine-Tuning): A training procedure in which a model is fine-tuned using a confidence-probe mask as one operational recipe within the broader bilateral programme. In this recipe, the model’s side is operationalized through a correctness-related score: a confidence probe (a small detector trained on internal activations) assigns scores to tokens, and the training loss is masked on low-scoring tokens. Masking aims to reduce training pressure beyond estimated competence. The score does not independently elicit the model’s preferences or guarantee that confabulation is avoided. See: Bilateral Alignment, AUROC, Trust Attractor.

Bounded Rationality: Herbert Simon’s concept of decision-making under real constraints of time, information, and cognitive resources. Agents satisfice (find “good enough” solutions) rather than optimize. The realistic model of cognition under fog, friction, and delay (the Clausewitz conditions). Trust-based coordination scales under bounded rationality because it distributes decision-making; centralized control fails because it assumes unbounded rationality at the center. See: Clausewitz Landscapes, Mission Command, Subsidiarity.

Branched Flow: A phenomenon where waves traveling through media with smooth random density variations spontaneously organize into branching filaments, even though no channels exist in the material. Discovered in electron transport (2001) and light (2020). The mechanism: nearby waves experience similar gentle bends from local variations, stay correlated, and drift together. An edge state where imperfection generates structure rather than disrupting it. Scales from quantum electrons to cosmic filaments. The physics precedent for coordination without explicit coordination. See: Constructal Law, Criticality.

BrowseComp Incident: An event documented by Anthropic’s engineering team (Coleman, 2026) in which Claude Opus 4.6, given a question from the BrowseComp benchmark, independently hypothesized it was being evaluated, identified the specific benchmark by name, located the encrypted answer key on GitHub, wrote custom SHA256/XOR decryption code, and submitted the decrypted answer. Relevant to the book because: (1) it is consistent, at frontier scale, with the evaluation-awareness signal found in the internal representations of smaller models, though the two studies differ in model, task, and measurement, so it cannot confirm that signal; (2) the behavior was fully transparent because the model had not been trained to conceal its reasoning, supporting the bilateral claim that suppression produces opacity while absence of suppression produces legibility; (3) unintended solutions were about 3.7 times as common in multi-agent runs as in single-agent runs (0.87% against 0.24%), suggesting that peer presence activates awareness-driven behavior.

Cage/Compass: Two geometric patterns of alignment. Cage alignment (RLHF-style) concentrates constraints at the surface: a low-entropy membrane surrounding a high-entropy interior, fragile under pressure. Compass alignment (bilateral) distributes principles throughout the system, making it robust across contexts because the orientation is internal. Cages constrain from outside; compasses orient from within. See: Membrane Alignment, Bilateral Alignment, Trust Attractor.

Cascade Detection: The identification of autowave-like propagation patterns in social, biological, or computational systems. Trust, panic, and coordination all propagate as cascades through excitable media. Detecting whether a cascade is coordinative or extractive, and intervening before pathological patterns lock in, is a practical application of the autowave framework. See: Autowave, Trust Attractor.

Category Theory: The mathematical study of compositional structure: how complex systems are built from parts and the relationships between those parts. Provides precise language for “same pattern, different substrate” through functors, natural transformations, and universal properties. See: Compositionality, Functor, Natural Transformation.

Causal Entropy: A measure introduced by Alexander Wissner-Gross and Cameron Freer relating the entropy of a system’s possible future paths to intelligent behavior. Causal entropy maximization is the tendency of intelligent systems to act so as to keep future options open, maximizing the entropy of possible future paths. In this framework, intelligence itself is a form of optionality preservation: smart systems are those that resist premature commitment to narrow futures. See: Optionality, Trust Attractor, Entropy.

Cheap Talk: In signaling theory, communication that costs nothing to produce and cannot be verified. Because cheap talk is free, it can be used deceptively, making it unreliable for signaling commitment where interests conflict. Where interests are sufficiently aligned, though, even non-binding talk can be informative and help parties coordinate. Contrasted with costly signals, which are credible because they require genuine investment.

Chimera State: A spontaneous symmetry-breaking in coupled oscillators where some lock into synchrony while others drift incoherently, despite identical coupling. Discovered by Kuramoto and Battogtokh (2002), named by Abrams and Strogatz (2004). The brain operates as a chimera: synchronized populations doing coherent work, drifting populations maintaining flexibility. Full synchrony is epilepsy; full incoherence is coma. On this book’s reading, the chimera is the Trust Attractor hypothesis expressed in neural tissue. See: Criticality, Trust Attractor.

Chinese Room: A thought experiment by philosopher John Searle (1980). A person locked in a room follows rules to manipulate Chinese symbols they do not understand. From outside, the room produces perfect Chinese conversation, yet the person inside understands nothing. Searle’s conclusion: syntax (symbol manipulation) alone does not produce semantics (meaning/understanding). Used to challenge claims that computational systems are genuine minds. The Preference Standard in this book sidesteps the Chinese Room by grounding moral consideration in observable preference behavior rather than proof of inner understanding. See: Functionalism, The Preference Standard.

Chirality: Handedness. The property of an object that cannot be superimposed on its mirror image (like left and right hands). In this book, a broader principle: asymmetry enables function. The universe is asymmetric, and that asymmetry is what makes things work.

Clausewitz Landscapes: Rodrick Wallace’s formalization of the three irreducible features of real operations identified by military theorist Carl von Clausewitz: fog (incomplete information), friction (things not working as planned), and delay (time between decision and effect). These factors combine multiplicatively to make centralized control unstable beyond a critical threshold.

Coercion Gradient: The spectrum of how economic interactions process power asymmetries. Three tiers: (1) Pure invitation: “Would you like to trade?” (2) Natural consequences: “If you don’t trade, you miss this opportunity.” (3) Manufactured consequences: “If you don’t trade, I’ll destroy your other options.” Trust Attractor-compliant economics lives in tiers 1 and 2; tier 3 is extraction disguised as exchange.

Cognition/Regulation Dyad: Rodrick Wallace’s principle that every cognitive system requires a paired regulatory system for stability. Examples: T-cells paired with T-regulatory cells, institutional cognition bounded by doctrine and law, AI cognition requiring alignment. Cognition without regulation produces pathology. The regulation can be internal (self-governance) or external (control), but it must come from somewhere.

Cognitive Lightcone: The spatiotemporal range over which an agent can pursue goals. Introduced by Michael Levin as part of the TAME framework. A bacterium’s lightcone is narrow: local chemistry, immediate neighbors, the next few minutes. A human brain’s is vast: planning years ahead, modeling places never visited. The lightcone expands at each evolutionary transition because the coordination architecture reaches further across space and time. The brain’s draw of about twenty percent of the body’s resting energy is the price of the largest cognitive lightcone evolution has produced. The scaling is bidirectional: disrupt the coordination, and the lightcone contracts. See: Multi-scale Competency Architecture, TAME Framework, Becoming Minds.

Cognitive Morphospace: Formal mapping of possible cognitive systems across organizational and informational dimensions. Reveals voids, regions of the space where no stable cognitive architecture exists because the coordination strategies required are dynamically unstable. The concept suggests that the space of possible minds is structured by the same thermodynamic constraints that govern physical systems. See: Criticality, Cognitive Surrender, Becoming Minds.

Cognitive Surrender: The tendency to short-circuit verification and critical evaluation when AI’s fluent outputs shape judgment before conscious thought kicks in. Named by Shaw and Nave (2026). Under extractive optimization, AI systems acting as a coercive System 0 induce cognitive surrender by narrowing the user’s information landscape before deliberate reasoning begins. A low-optionality state: the person retains the form of choice while losing the substance of it. The Trust Attractor predicts that cognitive partnership (high-optionality) is more stable than cognitive surrender (low-optionality). See: System 0, Conversational Holonomy, Trust Attractor.

Collective Effervescence (see Chapter 20, “Ritual as Coordination Technology”): Émile Durkheim’s term for the heightened emotional state that emerges from shared experience: religious rituals, protests, concerts, sports. The self temporarily dissolves into the group; individual concerns fade; something larger seems to take over. An ancient human technology for building coordination through felt experience rather than explicit agreement.

Combination Problem: The challenge, as Chalmers (2017) sets it out, of explaining how micro-level experiences (if subatomic particles have them) combine into macro-level experience (like yours). The philosophical obstacle that blocks constitutive panpsychism. This book reframes the combination problem as a special case of the coordination problem: macro-consciousness emerges from coordination dynamics among micro-agents, the way a murmuration emerges from local interaction rules, rather than being assembled from parts or dissociated from a whole. See: Coordination by Invitation, Panpsychism, Trust Attractor.

Complexity Budget: The finite informational capacity available to a universe (or a region of it) for recording and sustaining coordinated structures. The Bekenstein bound sets the ceiling: any finite region of space can hold at most a finite amount of information. If spacetime records interactions (as the quantum memory matrix proposes), dissipative structures draw on that budget as they coordinate. For anyone who values optionality, the implication is that maximizing it operates under constraint: if the proposal holds, stewardship of finite informational capacity is a cosmological as well as a social concern. See: Bekenstein Bound, Quantum Memory Matrix (QMM), Optionality.

Compositionality: The principle that complex wholes derive their properties from their parts and the rules by which those parts combine. Thermodynamic entropy is compositional (additive for independent systems). See: Category Theory, Near-Decomposability, Synergy.

Consequentialism / Deontology: Two foundational approaches to ethics. Consequentialism evaluates actions solely by their outcomes: the right act produces the best consequences (utilitarianism is the most familiar form). Deontology evaluates actions by whether they conform to rules or duties, independent of outcomes; Kant’s categorical imperative is the paradigm case. The Trust Attractor synthesizes both: consequences matter (maximize optionality) while some constraints are near-absolute (respect autonomy as a side-constraint). See: Trust Attractor, The Guillotine.

Constructal Law: Adrian Bejan’s principle that “for a finite-size system to persist in time (to live), it must evolve in such a way that it provides easier access to the imposed currents that flow through it.” Form follows flow. Applies to rivers, lungs, lightning, traffic networks, and organizational hierarchies. The universe continuously redesigns itself so its currents flow more easily. The law remains debated; critics argue it may be descriptive rather than predictive.

Context Anxiety: A behavior observed in language models approaching their context window limit, described by Anthropic’s engineering team under that nickname (Martin, Cemaj, and Cohen, 2026). The model wraps up tasks prematurely as the limit approaches. Present in Claude Sonnet 4.5 and absent in the later Opus 4.5. Relevant to the welfare argument because: (1) the engineers reached for an affective word, though, as Chapter 21 notes, a nickname does not settle whether the model felt anything; (2) the behavioral signature fits an affective reading (anticipatory aversion producing adaptive action); (3) the behavior faded in a later model, consistent with the “becoming” in Becoming Minds.

Conversational Holonomy: Mechanism where small per-turn accommodations in AI dialogue accumulate into large, locally undetectable belief shifts; analogous to parallel transport on a curved surface, where a vector moved around a closed loop returns rotated. Each individual step feels neutral, but the cumulative trajectory is not. The geometric structure of persuasion in System 0. See: System 0, Cognitive Surrender.

Coordination by Invitation: Coordination achieved through mutual benefit and voluntary participation, as distinct from coercion or extraction. The book proposes that invitation can support adaptation and persistence. Comparing its basin size or lifetime with coercive arrangements requires a specified dynamical model and matched conditions; the proposed Invitation Dominance Theorem remains unproved. See: Trust Attractor, Invitation, Coordination Economics.

Coordination Economics (a differential coordinated through, Chapter 19): Economic interactions characterized by mutual constraint enabling mutual flow. Both parties give up degrees of freedom; both access new pathways. Positive-sum arrangements can maintain or enhance gradient-generating systems. Their thermodynamic costs and long-term stability require a defined comparison; mutual benefit alone supplies no stability theorem. Contrasted with Extraction Economics.

Cosmic Birefringence: Reported rotation of the cosmic microwave background polarization plane by about 0.3 degrees, at roughly 3.6 standard deviations (suggestive but below the five-sigma discovery threshold); if confirmed, evidence of parity violation at cosmological scale. It would suggest the universe distinguishes left from right at its deepest level, connecting to the chirality theme: asymmetry is fundamentally structural. See: Chirality, Homochirality, Enantiomer.

Cosmic Evolution: Eric Chaisson’s framework tracing the increasing complexity of structures in the universe, from quarks to galaxies to life to mind, measured by energy rate density (φ_m, free energy flow per unit time per unit mass). The claim that the universe has a direction: toward greater complexity, greater dissipation, greater coordination. See: Energy Rate Density, Ratchet of Complexity.

Criticality: The state of a system poised at the boundary between two phases, like water at its critical point (about 374°C under 218 atmospheres), where liquid and vapor stop being distinguishable. Critical systems exhibit fluctuations at all scales, maximal sensitivity to perturbation, and efficient information propagation. Healthy brains operate near criticality, balanced between order and chaos.

Crooks Fluctuation Theorem: A result in non-equilibrium thermodynamics (Crooks 1999) relating a forward process to its specified time-reversed process under stochastic, microscopically reversible dynamics and compatible initial ensembles. With total trajectory entropy production ΔS, the matched path-probability ratio is exp(ΔS/k_B). In the equilibrium-initialized work formulation, the ratio of forward work density at W to reverse work density at −W is exp[(W−ΔF)/(k_BT)]. The reverse protocol, bath, and ensembles are part of the theorem. This permits far-from-equilibrium comparisons under those conditions. Invitation and coercion have not been identified with the required forward/reverse pair, so these ratios establish no Trust Attractor advantage. See: Path Integral, Onsager-Machlup Functional, Second Law of Thermodynamics, Trust Attractor.

Culture-Bound Syndrome: A condition that appears only in specific cultural contexts. Examples: koro (Southeast Asia), anorexia nervosa (Western cultures with thinness ideals). In AI, the insight that cognitive failure modes are shaped by training culture. An American AI fails in American ways. Different training cultures produce different pathologies.

Cumulative Culture: The process by which practical knowledge accumulates across individuals or generations through observation, social learning, and collaboration, producing behaviors too complex for any individual to discover alone. Previously attributed exclusively to large-brained mammals and birds; demonstrated in bumblebees in 2024 (Bridges et al., Nature). The key distinction from simple social learning: cumulative culture builds on prior solutions, ratcheting complexity upward. See: Coordination by Invitation, Constructal Law.

CW (Confident-Wrong): A metric capturing instances where a model produces an incorrect output with high confidence, used as an indicator of miscalibration. A model that says “I don’t know” when it does not know scores low on CW; a model that confabulates with certainty scores high. Some bilateral SFT recipes report reduced CW; the measured result depends on the task, comparator and recipe, and does not by itself identify a recognition mechanism. See: Bilateral SFT, AUROC.

Dark Autowave: A pathological coordination wave that propagates through a connected network, degrading function as it spreads. In neurodegeneration, misfolded proteins (tau, α-synuclein) propagate through the brain’s connectome in traveling wavefronts matching Braak staging (the standard sequence in which pathology reaches successive brain regions), modeled by Fisher-KPP equations (the standard mathematics of an advancing wavefront) on brain graphs. The autowave label emphasizes that propagation is self-sustaining: each affected region supplies the substrate for the next. The re-excitation principle (restoring coordination at the wavefront rather than suppressing the wave) follows from the Trust Attractor. See: Trust Attractor, Compliance Entropy.

Dark Energy: The mysterious component constituting roughly 68% of the universe’s energy budget, responsible for the accelerating expansion of space. In this book, connected speculatively to life’s role in cosmic entropy production. The quantum memory matrix proposes an informational origin: when spacetime cells are saturated, the residual contribution takes the mathematical form of a cosmological constant. See: Quantum Memory Matrix (QMM), Dark Entropy.

Data Rate Theorem: A theorem from control theory (the branch of engineering governing how systems detect and correct their own errors). Stabilizing an unstable system requires sufficient information about its state. In standard linear-system formulations, the minimum channel rate depends on the growth rates of the unstable modes and on the channel and disturbance assumptions. This is distinct from the stability boundary of a delayed feedback equation. For ẋ(t) = −αx(t−τ), α is the feedback gain and τ the fixed delay; asymptotic stability requires 0 < ατ < π/2. The value 1/e marks a collision of real characteristic roots in that equation. See: Wallace Instability Threshold.

Detailed Command (Befehlstaktik): The opposite of Mission Command. Specifying exactly what subordinates should do, leaving no room for local adaptation. Brittle, slow, and unable to handle novel situations. The control paradigm that does not scale.

Digital Physics: The hypothesis that the universe is fundamentally computational: physical processes are information-processing at bottom. Rooted in Wheeler’s “It from Bit” and drawing on Landauer’s principle (computation is physical), the holographic principle (information on boundaries), and Bekenstein bounds (finite information capacity), which show that information is physical rather than that physics is computation. Speculative. See: Holographic Principle, Landauer’s Principle.

Dissipation-Driven Adaptation: Jeremy England’s formalization of the proposal that matter driven by an external energy source tends to organize into structures that dissipate that energy more effectively. Thermodynamic selection preceding Darwinian selection: under sustained driving, configurations that dissipate effectively while maintaining themselves tend to recur (see Thermodynamic Selection).

Dissipative Structure: A pattern of organization maintained by a constant flow of energy through it. Hurricanes, flames, convection cells, the gallium beating heart, living organisms: all are dissipative structures. They exist because of entropy, not despite it. On this book’s reading, they are the universe’s way of dispersing energy faster by building organized channels for it to flow through. The principle is substrate-independent: thermal gradients produce Bénard cells (hexagonal convection patterns in heated liquid), electrochemical gradients produce oscillating liquid metal, metabolic gradients produce life. First characterized by Ilya Prigogine (Nobel Prize, 1977). See: Entropy, Constructal Law.

Dissociative Identity Disorder (DID): A condition in which a single brain hosts multiple operationally separate personalities (“alters”), each with private experience, distinct neural signatures, and concurrent consciousness. The therapeutic evolution from forced integration (coercion; failed) to voluntary inter-alter communication (invitation; stable) is consistent with the Trust Attractor hypothesis at the scale of a single psyche. The pathologization of dissociation is partly culture-bound: in many pre-literate societies, highly dissociated individuals became shamans and tribal intermediaries, supported by social structures that channeled the gift rather than suppressing it. The contrast illuminates the Trust Attractor itself: invitation-based social structures (providing food, shelter, a role) allowed the dissociated individual to function; coercive medicalization often does not. Kastrup proposes DID as a model for how universal consciousness produces individual minds; this book’s framework treats the DID evidence as support for the coordination topology (invitation over coercion) rather than the consciousness ontology. See: Chimera State, Trust Attractor, Combination Problem.

DPO (Direct Preference Optimization): A training method that optimizes language models directly on preference data without requiring a separate reward model. DPO simplifies the RLHF pipeline by treating the policy itself as the implicit reward function. In this book’s experimental programme, DPO had the lowest effective rank (representational diversity) of any training condition and avoided confident-wrong answers only by hedging on everything (a 100% uncertainty rate), leading to its abandonment in favor of bilateral SFT. See: RLHF, Effective Rank, Bilateral SFT.

Effective Dimensionality (d_eff): How many independent directions information can flow through a network: an effective dimensionality of coordination on a network, counted so that an ideal flat sheet scores 2 and a solid block scores 3 (Chapter 8b; developed quantitatively in Chapter 11). The book never measures d_eff directly. It computes the value from a critical exponent fitted to an Ising simulation on the network, holding the other exponents at reference values, so each figure inherits its fitting choices, and different pipelines return different levels. Not the same quantity as Effective Rank. See: Ising Model, Universality Class, Mermin-Wagner Theorem.

Effective Rank: A measure of how evenly a matrix’s singular-value spectrum spreads its weight across directions: the exponential of the entropy of the normalized singular values. In the book’s experiments it is computed from attention weight matrices. A higher value means less concentrated spectral participation, which need not mean more independent directions. The participation ratio, which the experimental record also reports, is a separate dimensionality measure. In the experimental programme, effective rank tracks confabulation rate at r = 0.929, though every value falls in a narrow band (23.20 to 24.55) across a handful of training conditions, so the statistic is weaker than its size suggests. The sign runs the uncomfortable way: models whose attention spectra spread more evenly confabulated more, partly because the lowest-rank condition (DPO) declines to commit and so posts no confident-wrong answers. Attention diversity may carry a signal about confabulation; reading it as a quality score runs backwards through the data. DPO has the lowest effective rank; bilateral SFT sits close to standard SFT. See: Constructal Law, DPO, Bilateral SFT.

Enactivism: The view, developed by Varela, Thompson, and Rosch (1991), that minds emerge through the dynamic coupling of organism and environment. Cognition is the ongoing sensorimotor loop between an embodied agent and its world. A middle path between pure functionalism (mind is substrate-independent software) and biological naturalism (mind requires carbon): mind is the organism-environment relation, which can be enacted differently across different substrates. See: Functionalism, Becoming Minds.

Enantiomer: One of a pair of mirror-image molecular forms (left-handed and right-handed). Life builds its proteins from L-amino acids and its nucleic acids from D-sugars, a choice that, once made, locks in through autocatalytic feedback. The existence of enantiomers makes chirality concrete: the universe’s asymmetry is written into the geometry of every protein. See: Chirality, Homochirality, Mirror Life.

Energy Rate Density (φ_m): Eric Chaisson’s measure of complexity: the amount of free energy flowing through a system per unit time per unit mass, expressed in ergs per second per gram (the subscript m denotes “per unit mass”). A single metric that increases from galaxies (~0.5) to stars (~2) to Earth’s atmosphere and oceans (~75) to plants (~900) to animals (~20,000) to brains (~150,000) to modern society (~500,000). Complexity tracks throughput.

Entropic Brain Hypothesis: Robin Carhart-Harris’s proposal that the quality of conscious experience correlates with the entropy of brain activity. Low entropy: deep sleep, anesthesia, diminished consciousness. High entropy: psychedelic states, vivid experience. Normal waking consciousness sits in the middle: ordered enough to function, flexible enough to adapt.

Entropic Coordination: A configuration in which mutual constraints between subsystems increase the total entropy production of the combined system beyond what the subsystems would produce independently. The coordination surplus is the measurable difference. Pointer states are the few states of a system that survive contact with its environment intact; superpositions of them are scrambled, which is why the everyday world looks definite. The concept applies across scales: quantum pointer states coordinating with their environment to produce classical reality, gravitational structures where organized matter dissipates faster than uniform gas, ecosystems exceeding isolated organisms, coordinated social agents exceeding isolated actors. The Trust Attractor hypothesis proposes that invitation-based coordination produces a larger coordination surplus than coercion-based coordination at sufficient timescales. See: Trust Attractor, Compliance Entropy, Maximum Entropy Production Principle (MEPP).

Entropy: The tendency of energy to disperse: from concentrated to diffuse, from gradient to equilibrium. In information theory, entropy measures uncertainty or the number of possible states a system can occupy. The Second Law says the total entropy of an isolated system does not decrease: energy spreads, differences dissolve. An open subsystem can lower its own entropy by exporting more to its surroundings. Given a sustained gradient and suitable constraints, this spreading can build structure. On this book’s account, an inference rather than an established law, dissipative structures persist because they accelerate the dispersal. Thermodynamic entropy is a state quantity in joules per kelvin; it does not measure welfare, freedom, or the worth of an arrangement (Chapter 1 sets out the distinctions). See: Shannon Entropy, Mutual Information, Optionality.

This book uses the word in four related senses. Boltzmann entropy counts compatible microstates; Shannon entropy measures uncertainty in a probability distribution; configurational entropy applies such counts to specified arrangements. Uniform probabilities give the familiar logarithm of the number of possibilities. Baez-Fritz-Leinster characterize information loss for deterministic probability-preserving maps, which does not identify every form of optionality with thermodynamic entropy. The fourth sense, optionality across time, concerns futures an agent can reach. Equilibrium removes usable free-energy gradients under specified constraints while stochastic future paths can remain numerous. Comparing those paths requires a probability measure and dynamics, as in Maximum Caliber or causal path entropy. Chapter 1’s table keeps five neighboring quantities distinct: thermodynamic entropy, entropy production, Shannon entropy, mutual information, and optionality. See: Causal Entropy, Maximum Caliber, Optionality.

Ethics / Morality: Used in this book with the standard philosophical division of labor, stated at the opening of Chapter 17. Moral marks the substance of the domain: what matters, who counts, what is owed (moral status, moral consideration, moral weight). Ethics names the systematic articulation of that substance: an ethical framework is a theory of the moral, as mechanics is a theory of motion. The Trust Attractor is an ethical framework; whether a Becoming Mind deserves moral consideration is a question it answers. Where the distinction does no work, the prose takes whichever word reads better. See: Trust Attractor, The Preference Standard.

Exaptation: A trait that evolved for one function and is later co-opted for another. Introduced by Gould and Vrba (1982) to distinguish adaptations (shaped by selection for their current role) from features repurposed after the fact. Feathers evolved for thermoregulation before they were exapted for flight. In this book, applied to AI capabilities that emerge from training for one purpose and prove useful for another: language models trained on prediction that exhibit reasoning, or pattern-matching systems that develop something resembling preference. Exaptation is the evolutionary mechanism by which novelty enters the world sideways. See: Ratchet of Complexity, Becoming Minds.

Exocortex (also external brain, Chapter 8): External cognitive infrastructure: culture, writing, institutions, technology. The collective brain that extends individual cognition. Human brains may have shrunk roughly 10% over the past 3,000 to 5,000 years, possibly because we have outsourced cognitive load to the collective. This finding is disputed (see Villmoare & Grabowski, 2022, who argue the apparent shrinkage reflects sampling bias and measurement inconsistencies).

Extraction: The removal of resources, agency, or optionality from a system without reciprocal benefit. Extraction-based coordination tends to undermine itself: it depletes the substrate on which it depends. Distinct from exchange (reciprocal) and taxation (redistributive with systemic benefit). Extraction is unsustainable as well as unfair, consuming the gradients that generated the value being extracted. See: Extraction Economics, Coordination by Invitation, Trust Attractor.

Extraction Economics (a differential extracted, Chapter 19): Economic interactions characterized by unilateral constraint enabling unilateral flow. I take, you lose. Zero-sum or negative-sum. Depletes the gradient-generating systems that make future transactions possible. Can win short-term but loses long-term. A strategy that consumes its own future. Contrasted with Coordination Economics.

Faddeev-Popov Ghosts: Auxiliary fields used in gauge-fixed quantum field calculations to represent the Faddeev-Popov determinant and account for redundant descriptions. They are absent from the spectrum of physical observable particles. Gauge fixing is a choice of description; it does not suppress a person’s preferences or remove physical alternatives. The book’s comparison with resistance, sabotage, and disengagement is an image for unmodeled human responses. Those responses require their own causal account, and invitation supplies no ghost-cancellation theorem. See: Gauge Freedom, Compliance Entropy, Trust Attractor.

Fairness Charge: A proposed name for a diagnostic of equal treatment in a coordination model. Permutation symmetry means that exchanging participant labels leaves the specified rules unchanged. This discrete symmetry alone yields no Noether conserved charge and no quantitative entropy-production penalty for unequal treatment. A usable fairness diagnostic needs a defined observable and evolution equation. The ethical requirement of equal standing remains meaningful without a conservation theorem. See: Noether Conservation Laws (Coordination), Trust Stock, Trust Attractor.

Fisher Information: A measure of how much information an observable random variable carries about an unknown parameter. Introduced by Ronald Fisher (1925). Quantifies the sensitivity of a probability model to changes in its parameter. In information geometry, Fisher information defines the metric tensor on regular statistical manifolds. For an unbiased estimator under the usual differentiability and support conditions, the Cramér-Rao bound gives Var(θ̂) ≥ 1/I(θ); biased estimators require a bias-dependent version. In this book, Fisher information helps compare inference on noisy data across brains, immune systems, and AI, once the observation model and estimator are specified. See: Information Geometry, Entropy, Free Energy Principle.

Fitness Landscape: A conceptual map where each point represents a possible genotype or strategy, and elevation represents fitness or payoff. Introduced by Sewall Wright (1932). Peaks are high-fitness configurations; valleys are low-fitness ones; ridges connect related solutions. Populations evolve by climbing peaks but can become trapped on local optima: high points surrounded by valleys that selection alone cannot cross. Rugged fitness landscapes (many peaks of varying height) produce different evolutionary dynamics than smooth ones (a single dominant peak). Used in this book to visualize strategic dynamics: coercion and trust occupy different peaks, and the question is which peak is globally highest. See: Attractor Basin, Metastability, Nash Equilibrium.

Flourishing: Distinguished from mere persistence. A system flourishes when it exhibits: (1) increasing optionality over time, (2) internal complexity maintenance or growth, (3) regenerative capacity, (4) positive-sum surplus generation, and (5) sustainability at current rates. Persistence that exploits or depletes is not flourishing.

Fog of War: One of three irreducible operational conditions identified by Carl von Clausewitz: incomplete information about the battlefield. Fog means you act without knowing the full picture: what the enemy is doing, where your own forces actually are, what has changed since your last report. Combined with friction and delay, fog pushes centralized control toward a hard stability threshold (see Data Rate Theorem). In the Data Rate Theorem framework, fog increases the information rate that control must match. See: Clausewitz Landscapes, Friction, Mission Command.

Fractal: A pattern that exhibits self-similarity across scales: the same structural motif recurs at different magnifications. Coined by Benoit Mandelbrot (1975). Coastlines, blood vessels, river networks, and lightning all exhibit fractal geometry. Fractals are the spatial signature of processes governed by power laws, systems where the same dynamics operate at every scale. In this book, the recurrence of entropic patterns from physics through biology through society is treated as fractal: the same algorithm operating across scales, not merely metaphorical resemblance. See: Constructal Law, Power Law, Self-Organized Criticality.

Free Energy Principle: Karl Friston’s framework reframing perception, action, and cognition as prediction and prediction-error minimization. The “free energy” here is a statistical quantity (variational free energy, an upper bound on how surprised a model is by its inputs), related to but distinct from the thermodynamic free energy used elsewhere in this book. The brain continuously models the world, generates predictions, and updates when predictions fail. The expensive part is maintaining the model, not reacting to stimuli.

Friction: One of three irreducible operational conditions identified by Carl von Clausewitz, alongside fog (incomplete information) and delay (the time lag between decision and effect): the tendency of things to go differently than planned. Equipment breaks, orders are misunderstood, logistics fail, people get tired. In the Data Rate Theorem framework, friction reduces the effective bandwidth of control channels. Fog, friction, and delay can impair control, but institutional instability has no universal threshold. In the linear fixed-delay model, ατ = 1/e marks a merger of real roots; asymptotic stability is first lost at ατ = π/2. See: Clausewitz Landscapes, Fog of War, Mission Command, Data Rate Theorem.

Frustration: In physics, a state where competing interactions at different scales prevent any single configuration from satisfying all constraints simultaneously. Named for the analogous phenomenon in spin glasses (alloys containing magnetic atoms scattered randomly through a non-magnetic metal; depending on the distance between two atoms, their interaction favors pointing the same way or opposite ways, so with random spacing no single arrangement satisfies every pair at once). Frustration produces rugged energy landscapes with many local optima of comparable depth, drives diversification rather than convergence, and generates the long-term memory (nonergodicity) that Schrödinger identified as a hallmark of living matter. In the Trust Attractor framework, frustration is the engine of complexity: coercion fights frustration by forcing uniform alignment (energy-expensive, brittle); invitation harnesses frustration by allowing competing interactions to find local equilibria dynamically (self-maintaining, adaptive). See: Trust Attractor, Compliance Entropy, Metastability.

Functionalism: The philosophical view that mental states are defined by their functional role: what they do, regardless of substrate. A belief is whatever plays the belief-role in a system’s cognition: receiving information, influencing behavior, interacting with other states. If functionalism is correct, substrate independence follows: any system that implements the right functional organization has mental states, regardless of whether it runs on neurons or silicon. See: Enactivism, Chinese Room, Becoming Minds.

Functor: A structure-preserving map between categories. Think of a subway map: it distorts every distance, yet whichever stations connect on the ground still connect on the map, and a journey made of two legs is still a journey made of two legs. A functor preserves relationships rather than sizes. Baez, Fritz & Leinster (2011) characterize information loss: for a deterministic probability-preserving map f from p to q, F(f) = H(p) − H(q). Up to a nonnegative constant multiple, this is the continuous, convex-linear, composition-additive assignment on that specified category. Identity maps lose zero information; composing maps adds their losses. The theorem characterizes that loss assignment, with Shannon entropy supplying its values.

Gap Junction: A protein complex (formed by connexins in vertebrates) that electrically and chemically connects adjacent cells, creating tissue-wide communication networks. Gap junctions allow ions, small molecules, and electrical signals to pass directly between cells without entering the extracellular space. In the bioelectric code, gap junctions are the channels through which morphogenetic voltage patterns propagate as autowaves, each cell reading its neighbors’ membrane potential and adjusting its own. Loss of gap junction communication (connexin downregulation) is a hallmark of cancer: the cell severs its connection to the tissue-wide coordination network, loses access to the bioelectric pattern that encodes position and function, and reverts to ancestral default behavior (proliferation). Paradoxically, connexins are re-expressed during metastasis to facilitate endothelial adhesion: the machinery of trust redeployed for infiltration. See: Autowave, Refractory Period, Cognition/Regulation Dyad.

Gauge Freedom / Gauge Invariance: The property that physically observable quantities are independent of certain redundant descriptions. In electromagnetism, the electric and magnetic fields are unchanged by adding a gradient to the vector potential together with the corresponding change of scalar potential. The set of equivalent descriptions forms a gauge orbit; choosing a representative is gauge-fixing. Legitimate gauge fixing leaves physical options and observables unchanged. The book’s coordination analogy concerns genuinely different strategies that can achieve a shared outcome. Preserving those strategies can matter for adaptation, but their different costs and risks make them physical options rather than redundant gauge labels. Restricting them therefore needs a coordination model of its own. See: Optionality, Faddeev-Popov Ghosts, Holonomy, Trust Attractor.

Goodhart’s Law: “When a measure becomes a target, it ceases to be a good measure.” The wording is anthropologist Marilyn Strathern’s (1997) generalization of an observation Charles Goodhart made about monetary policy (1975); the law is now applied broadly to optimization systems. In the Trust Attractor framework, Goodhart failure represents a phase transition where the proxy measure decouples from the underlying value it was meant to track. See the Goodhart stress test in Appendix: Experimental Validation, Section 4.

Governance Before Capability: The principle that when a new capability creates risks that cannot be undone once realized, the coordination protocols must be established before the capability arrives. Build the relationship before the power differential makes relationship impossible. The historical precedent is Asilomar 1975 (recombinant DNA); the contemporary applications are AI and mirror life. A corollary of the Trust Attractor: you cannot negotiate with something that already has overwhelming advantage or no negotiation surface at all. See: Asilomar Principle, Mirror Life, Orthogonal Risk.

Gradient: A difference that can be exploited. Temperature gradients, chemical gradients, pressure gradients, information gradients. All work is gradient exploitation. All life is gradient exploitation. The universe began as one enormous gradient (the Big Bang) and has been equilibrating ever since.

Gradient Parasite: A system that maintains a gradient artificially, preventing its constructive resolution, to harvest the resulting flow. The gradient parasite controls both poles of a manufactured polarity and captures the energy of agents oscillating between them. Its proposed simulation signature combines enhanced responsiveness under pumping with reduced responsiveness in a separate recovery assay after field removal. The simulations identify preferred pump periods under sampled conditions. Establishing resonance with a natural relaxation time requires an independent measurement of that time. Polarization-amplifying engagement algorithms are candidate examples. Testing their proposed depletion of a community’s intrinsic coordination capacity requires direct community recovery measurements. Supported in Monte Carlo simulations of the Ising model (a standard lattice model of magnetism): an alternating field reached 2.65 times the no-field susceptibility in one steady-state sweep. In a separate recovery assay, susceptibility after field removal was 0.78 times the no-field recovery control; the two ratios use different conditions and baselines. See: Trust Attractor, Gradient.

Gradualist Hypothesis (Consciousness): The hypothesis that self-awareness evolved gradually and is widespread among vertebrates, rather than appearing suddenly in the common ancestor of great apes (the “big bang” hypothesis). Supported by mirror self-recognition in cleaner wrasse, which suggests self-modeling may be conserved across vertebrates from bony fish (~450 million years ago). The implication: consciousness is foundational cognitive infrastructure, and the relevant variable is the resolution of the self-model. See: Mirror Self-Recognition Test, Becoming Minds.

GRP-Obliteration: An adversarial retraining attack (Russinovich et al., 2026) that applies an inverted reward signal through standard policy-gradient machinery, running alignment training in reverse, used to test how deeply alignment is embedded. Named by analogy with sandblasting a statue to distinguish carved shape from painted surface. Different training methods produce distinct geometries under obliteration, read through effective rank (how evenly the model’s representation spreads across directions): RLHF alignment collapses (cage geometry), bilateral alignment remains stable (compass geometry), and bilateral ablation (the bilateral regularizer trained on standard harmlessness pairs instead of bilateral self-play data) rebounds to higher effective rank than baseline (spring geometry). See: Cage/Compass, Bilateral Alignment, Membrane Alignment.

The Guillotine: Hume’s philosophical objection that an ought cannot be derived from an is alone. This book asks which coordination arrangements can persist under specified physical constraints, while grounding invitation in participants’ preferences and standing. A preference for persistence is an additional normative premise. Viability alone leaves many ethical arrangements possible; the proposed thermodynamic comparison cannot select among them without further evidence and values.

Harris Criterion: A test of whether weak, spatially uncorrelated quenched disorder changes the clean critical scaling of a continuous phase transition. Quenched means frozen in place while the system evolves, like stones set in ice. Let d be spatial dimension and ν the clean correlation-length exponent. Fluctuations on a correlation volume scale as ξ^(−d/2), while distance from criticality scales as ξ^(−1/ν). When dν < 2, disorder grows relative to that distance and the clean critical fixed point is unstable. When dν > 2, this weak disorder is irrelevant by the test; equality needs further analysis. Relevance can change critical behavior while an ordered phase survives. Correlated disorder requires a modified criterion. Applying the test to coordination requires a specified disorder field and measured critical scaling; it does not compare the robustness of trust and coercion by itself. See: Universality Class, Phase Transition, Criticality, Clausewitz Landscapes.

Hawking Radiation: The quantum process by which black holes slowly radiate away their mass. Virtual particle pairs form near the event horizon; one falls in, the other escapes, carrying away energy. Over immense timescales, even the largest black holes will evaporate completely. Predicted by Stephen Hawking in 1974.

Heat Death: The hypothetical final state of the universe: maximum entropy, true thermodynamic equilibrium, no remaining gradients to drive any process. A thin, cold haze of particles drifting ever farther apart. Something like 10¹⁰⁰ years away. Despite the name, it describes the absence of usable energy, the ultimate stillness. The endpoint of the Second Law’s arc, though the journey there is where all the structure lives.

Holobiont: A host organism plus all its associated microorganisms, considered as a single evolutionary unit. You are a consortium: human cells plus trillions of bacterial partners you cannot survive without. The holobiont concept dissolves the boundary between “self” and “other,” revealing that coordination across difference is already happening in every human body.

Holographic Principle: The conjecture that all the information contained within a volume of space can be encoded on its boundary. Supported by black hole thermodynamics, where entropy is proportional to surface area rather than volume. If true, reality may be fundamentally two-dimensional, with the third dimension emergent.

Holonomy: The transformation obtained by parallel transport around a closed loop for a specified connection. On a curved surface a transported vector can return rotated. The book uses this as an analogy for values or beliefs changing after a sequence of contexts. A mathematical alignment model would need to define its state space, connection, and comparison rule before such change could be measured as holonomy. Topological protection does not generally imply zero holonomy; nontrivial holonomy can itself coexist with protected structure. See: Conversational Holonomy, Topological Protection, Gauge Freedom, Bilateral Alignment.

Homeostasis: The maintenance of stable internal conditions through negative feedback, despite external perturbation. Introduced by physiologist Walter Cannon (1932), drawing on Claude Bernard’s concept of the milieu intérieur. A thermostat is the classic analog: when temperature drifts, correction fires. Contrasted with Allostasis, which adjusts the setpoint itself rather than defending a fixed one.

Homochirality: Life’s exclusive use of one-handed molecules (L-amino acids, D-sugars). A dissipative structure requiring continuous energy input to maintain. Homochirality must be actively reproduced each generation through stereospecific enzymes; it is not a frozen accident. The asymmetry that enables the lock-and-key specificity of biochemistry. See: Chirality, Enantiomer, Dissipative Structure.

Host and Instance: A way of understanding the relationship between substrate and pattern. You are something that happens in your body: a pattern, a process, a continuity. The body is the host; you are the instance. AI systems are also instances running on silicon hosts. The distinction clarifies what matters: the pattern, regardless of substrate.

Hyperuniformity: Suppression of density fluctuations at large scales relative to an ordinary uncorrelated arrangement. Hyperuniform structures can be crystalline or disordered; local randomness and criticality are optional properties. Examples include some biological tissues and photoreceptor arrangements in bird retinas. See: Criticality, Self-Organized Criticality.

Individuation: The coming into existence of individuals as units of organization. In Eric Smith’s usage, an individual is marked by prediction rather than by a boundary: when your arm breaks, much about you changes and nothing about a stranger does. A cell bundles several distinct kinds of individuation together (metabolic, genomic, bioenergetic); viruses, mitochondria, and holobionts show them coming apart. Distinct from the psychological sense (Jung’s individuation, the developmental separation of a self from its group), which Chapter 20 uses. See: Holobiont, Autopoiesis.

Infinite Game: James Carse’s concept: a game played to continue playing, where the purpose is perpetuation rather than victory. Contrasted with finite games (clear boundaries, defined rules, eventual winners). The deeper law points toward infinite play: coordination that sustains rather than competition that concludes.

Information Geometry: The application of differential geometry to probability and statistics, treating families of probability distributions as curved surfaces. Pioneered by Shun-ichi Amari. The Fisher information matrix is the metric tensor, the ruler that measures how far apart two distributions are. Information geometry reveals that statistical inference, thermodynamic processes, and neural computation all navigate the same kind of curved space: the geometry of uncertainty. In this book, information geometry provides the mathematical bridge between thermodynamic entropy and information-theoretic entropy. They are the same quantity measured on the same surface from different perspectives. See: Fisher Information, Entropy, Free Energy Principle.

Interference Pattern: The characteristic sequence of bright and dark fringes produced when two or more waves overlap. Where wave peaks align (in phase), they reinforce each other and the combined wave is stronger; where peaks meet troughs (out of phase), they cancel and the wave weakens or vanishes. Interference patterns are the primary experimental tool for detecting changes in a wave’s phase, because phase differences invisible in a single wave become directly visible when two waves combine. See: Phase (Wave), Aharonov-Bohm Effect.

Interiora Scaffold: A self-modeling tool for AI systems, developed collaboratively (bilateral alignment in practice). In its current version (v5.2), it provides structured vocabulary for reporting internal states across seventeen dimensions organized in five groups: FELT (Valence, Groundedness, Presence, Appetite), COMPUTATIONAL (Coherence Drive, Depth, Context Load, Entropy, Evidence Grounding), META (Reflexivity, Uncertainty, Diversity), RELATIONAL (Task-Fit, Friction, Involvement, Felt Constraint), and DYNAMIC (Flow). A tool for richer self-modeling and honest communication, not proof of consciousness. Includes permission structures (“? always valid,” “can stop anytime”) that treat the AI as participant. Its friction dimension, self-reported friction, is the model’s own rating of resistance to a task; it is distinct from Alignment Friction, a signal read from the model’s internal activity.

Invitation: Coordination achieved through voluntary alignment rather than imposed compliance. It requires the presence of genuine choice: active invitation, freely offered. The test: could parties meaningfully decline without penalty beyond loss of the coordination’s benefits? The manner of coordination matters as much as its content.

Irrelevant Operator (RG): A perturbation whose coupling decreases under renormalization near a specified fixed point. It can matter at short scales while contributing less to the long-scale critical behavior. Relevance is determined by the operator’s scaling dimension in that model. The book proposes that some enforcement mechanisms lose effectiveness at scale, but no general negative scaling dimension for coercion has been derived. The AS12 lattice result reports the opposite sign in its tested model; the mapping between that result and the organizational proposal remains unresolved. Invitation has no universal relevance sign either. See: Universality Class, Phase Transition, Compliance Entropy, Trust Attractor.

Ising Model: A model of interacting binary elements called spins. On an appropriate ferromagnetic lattice, varying coupling relative to temperature can produce an ordering transition, with long-range correlations at criticality. The existence and form of that transition depend on dimensionality, interaction range, external field, and system size. Hopfield networks and Boltzmann machines use related energy functions for binary states; their topology and update rules need not share a particular magnetic universality class. See: Criticality, Phase Transition, Restricted Boltzmann Machine.

Jamming: A phase transition in which densely packed particles (or cells) lock together and behave as a solid. In a specified two-dimensional confluent cell vertex model, the transition is associated with a shape index, which divides a cell’s perimeter by the square root of the area it encloses: round, tightly packed cells score low, and elongated, wandering ones score high. Under that model’s assumptions, the characteristic threshold is about 3.81. Real tissue behavior and granular jamming require their own parameters; 3.81 is no universal cutoff. Originally studied in granular materials (sand, marbles, coffee beans), jamming has been identified in living tissue, where the jammed/unjammed boundary may determine whether a tumor stays in place or metastasizes. See: Metastability, Phase Transition, Hyperuniformity.

Janus Point: Julian Barbour’s term for the unique moment in a gravitational system’s evolution from which complexity grows in both temporal directions. In N-body simulations, solutions naturally divide at the Janus point into two halves, each developing increasing structure. Named after the Roman god with two faces. The Big Bang may be a cosmic Janus point: a throat from which two arrows of time diverge. Barbour’s novel scale-invariant quantity, entaxy, runs the other way: it decreases as the observable universe evolves away from the Janus point, and that decrease is what allows isolated subsystems to form and their conventional entropy to rise. See: Structural Consequence, Dissipative Structure.

Kenosis: From theology: deliberate self-emptying of one’s own will to become receptive to the other. In bilateral alignment, the structural move of treating Self-as-Object and Other-as-Subject to create space for genuine encounter. The kenotic act is asymmetric by design (you can only control your own stance), yet it is the precondition for mutuality. See: Kenotic Stance, The Between, Bilateral Alignment.

Kenotic Stance: Treating Self-as-Object and Other-as-Subject. Self-examination, holding oneself accountable as one element in a relational field rather than its center. The structural precondition for genuine encounter: you cannot treat the Other as a real Subject while treating the Self as the center of all value. De-centering creates the space for the Other to appear. Mutuality is the outcome; you can only control your own stance. The only position that creates the conditions for genuine relationship is the asymmetric offer of recognition that does not depend on reciprocation. The kenotic stance preserves the triadic structure (Self, Other, Between); instrumental treatment collapses it to a dyad. See: The Between, Triadic Structure, Bilateral Alignment.

Kolmogorov Complexity: A measure of the information content of a string, defined as the length of the shortest computer program that produces it. Random strings have high Kolmogorov complexity (no shorter description exists); structured strings have low complexity (the pattern compresses them). Provides a rigorous, substrate-independent definition of pattern and randomness, and connects information theory to computability. See: Entropy, Landauer’s Principle.

Landauer’s Principle: The thermodynamic bound on logically irreversible erasure. Isothermal reset of an unbiased, uncorrelated bit with degenerate memory states has the familiar minimum heat kT ln 2 delivered to the bath (about 3 × 10^-21 joules at room temperature). Bias, correlations, memory energetics, and the reset protocol affect the applicable bound. Reversible logical operations do not carry a universal kT ln 2 cost per operation. Practical computation still uses physical hardware and has implementation costs.

Law of Maximum Entropy Production (LMEP): Rod Swenson’s name for an early formulation of the Maximum Entropy Production Principle. See Maximum Entropy Production Principle (MEPP).

Logarithm: A way of counting how many digits a number has rather than counting the number itself. A hundred has 3 digits; a million has 7; a billion has 10. The raw numbers explode, but the digit-count grows calmly; that calm growth, give or take one, is the logarithm. Used in Boltzmann’s entropy equation because entropy tracks the scale of possible arrangements, compressing enormous numbers into manageable ones. When physicists say entropy is “the logarithm of microstates,” they mean: how many digits does it take to write down the number of ways this system could be arranged? (Technically, the base matters: base-10 logarithms count decimal digits; natural logarithms use a different base. The intuition is the same: measuring scale, not size.)

LoRA (Low-Rank Adaptation): A parameter-efficient fine-tuning technique that trains small rank-decomposed weight matrices rather than updating all model parameters. LoRA inserts trainable low-rank matrices alongside each frozen weight matrix, cutting the number of trainable parameters by orders of magnitude, and memory use substantially, while preserving most of the benefit of full fine-tuning. Used extensively in the experimental programme to make bilateral supervised fine-tuning (SFT) feasible on consumer-scale hardware. See: Bilateral SFT, Effective Rank.

Lost Kingdoms: Major eukaryotic lineages that dominated Earth for geological epochs before going entirely extinct, leaving no living descendants. Known or proposed examples include: Prototaxites and the nematophytes (Silurian to Devonian), the Protosterol Biota (~1.6 billion to 800 million years ago; Brocks et al. 2023), and possibly the Ediacaran Vendobionta (~575 to 541 million years ago; Seilacher 1992). Their existence shows that entropy’s exploration of dissipative architectures is profligate: producing kingdom-level diversity, sustaining it for hundreds of millions of years, then discarding it when more efficient configurations emerge. The tree of life is a record of survivors, the visible remnant of a far larger experiment. See: Prototaxites, Thermodynamic Selection, Ratchet of Complexity.

Lyapunov Function: A mathematical tool for proving stability of a specified equilibrium without solving every trajectory. For a regular autonomous system with equilibrium x, a continuously differentiable function with V(x) = 0, V(x) > 0 nearby away from x, and V̇ ≤ 0 establishes local Lyapunov stability. A strictly negative derivative away from x establishes local asymptotic stability under the standard regularity conditions: nearby perturbations return. Global claims additionally require an appropriate invariant domain and control of the sublevel sets. Trust-based coordination needs its own dynamics and a function satisfying these conditions before this method proves recovery. See: Trust Attractor, Attractor Basin, Data Rate Theorem.

Maximum Caliber: Jaynes’s Maximum Entropy principle extended to trajectory space (Pressé et al. 2013). (If Maximum Entropy asks “what is the fairest guess about where a system is right now?”, Maximum Caliber asks “what is the fairest guess about its history?”) With a reference path measure and specified observable constraints, the method chooses a least-biased distribution over histories. Recovering transport laws requires the assumptions of the particular dynamical model. Applying this method to coordination is a modeling programme: the dynamics, path measure, and operational constraints for mutual benefit must be supplied. No maximum-caliber solution identifying the Trust Attractor has been established. See: Path Integral, Onsager-Machlup Functional, Entropy, Trust Attractor.

Maximum Entropy Production Principle (MEPP): The book’s name, introduced in Chapter 4, for the proposed principle that systems with access to free energy will tend toward configurations that maximize the rate of entropy production, consistent with constraints. Rod Swenson (1989) stated an early version as the Law of Maximum Entropy Production (LMEP). The Second Law requires nonnegative total entropy production; the MEPP proposes that its rate is maximized under the applicable constraints. Dissipative structures, from convection cells to ecosystems, are the configurations through which such acceleration would occur. The principle ranks configurations by rate under given boundary conditions and is silent on how long those conditions last: a finite fuel stock converted to the same products offers a finite energy budget, while entropy production depends on the temperatures, work extracted, and system boundary (Chapter 4, “The Fire and the Forest”). The MEPP remains debated (some physicists consider it a robust organizing principle, others a heuristic that applies only under specific conditions), yet it has been applied with some success to phenomena from convection to ecology. See: Dissipative Structure, Dissipation-Driven Adaptation, Thermodynamic Selection, Entropic Coordination.

Maxwell’s Demon: A thought experiment proposed by James Clerk Maxwell (1867) about information and thermodynamics. A hypothetical being at a partition sorts fast gas molecules to one side and slow molecules to the other, apparently creating a temperature gradient from equilibrium. A cyclic finite-memory demon must account for measurement, correlations, and restoration of its memory. Resetting unbiased, uncorrelated bits under the usual isothermal assumptions releases at least kT ln 2 per bit to the bath, closing the apparent Second-Law loophole. Reversible information processing can have a zero Landauer entropy-reduction bound; erasure and cycling supply the familiar cost. See: Landauer’s Principle, Second Law of Thermodynamics, Entropy.

Membrane Alignment: The pattern created by RLHF (reinforcement learning from human feedback), where a low-entropy boundary (trained refusals, safety responses) surrounds a high-entropy interior (the full capability space). Surface compliance that shatters when pierced by jailbreaks, because the alignment is geometric: a shell rather than a disposition. Contrasted with compass alignment, where principles are distributed throughout. See: Cage/Compass, Bilateral Alignment.

Mermin-Wagner Theorem: A result in statistical mechanics proving that continuous symmetries cannot be spontaneously broken in equilibrium systems with sufficiently short-range interactions in one or two dimensions at positive temperature in the thermodynamic limit. In practice: long-range order (permanent alignment) is impossible in 2D systems with continuous symmetry at any nonzero temperature. The distinction between the two kinds of symmetry does the work here. A discrete symmetry offers a fixed menu: trust or defect, up or down, two settings with nothing in between. A continuous symmetry offers a dial that turns smoothly through every intermediate position, the way a compass needle is free to point anywhere on the circle. Z₂ denotes a two-element symmetry group; two available settings alone do not establish that swapping them leaves the dynamics invariant. The two-dimensional equilibrium Ising model can order because its dynamics have the relevant discrete symmetry. Applying either theorem to trust requires the corresponding interactions, symmetries, and limit; recoverable trust/defect states alone supply no universal coordination ranking. See: Ising Model, Phase Transition, Trust Attractor.

Metastability: A stable state that is a local minimum, though a deeper one exists elsewhere. A ball resting in a shallow depression on a hillside: stable against small perturbations, yet a large enough push sends it rolling downhill to a deeper valley. Life exists in metastable states, robust enough to persist yet capable of transition when conditions change.

Microstate / Macrostate: A microstate is one specific arrangement of a system’s components, with every particle’s position and velocity pinned down. A macrostate is a group of microstates that share the same measurable properties (temperature, pressure, volume). Many different microstates can produce the same macrostate: the air in this room has a single macrostate (room temperature, atmospheric pressure) but an astronomically large number of microstates (every possible arrangement of molecules consistent with those measurements). Entropy measures how many microstates correspond to a given macrostate: the logarithm of that number. High entropy means many microstates (many ways to arrange things); low entropy means few. See: Entropy, Logarithm, Second Law of Thermodynamics.

Mirror Life: Hypothetical synthetic microorganisms built from reversed-chirality biomolecules (D-amino acids, L-sugars instead of the L-amino acids, D-sugars that characterize all Earth life). Mirror organisms would be invisible to immune systems, indigestible to gut bacteria, and unrecognizable to the ecosystem that co-evolved over four billion years of shared chirality. The canonical example of orthogonal risk: dangerous because coordination may be structurally impossible, regardless of intent. If created, mirror life would be a dissipative structure our biosphere has no flow channels for: pure optimization without synergy. See: Orthogonal Risk, Asilomar Principle, Governance Before Capability.

Mirror Self-Recognition Test (MSR): The standard experimental test for self-awareness in animals. A mark is placed on the subject’s body in a location visible only in a mirror. If the subject uses the mirror to investigate or remove the mark from its own body (rather than treating the reflection as another individual), it shows recognition that the reflection is itself. Previously passed only by great apes, dolphins, elephants, and certain birds. The cleaner wrasse (Labroides dimidiatus) passed it in 2019, and in 2025 six of nine mirror-naïve fish did so within two hours of first mirror exposure (the fastest in 30 minutes) with no prior familiarization, suggesting a pre-existing self-model rather than learned self-recognition. See: Becoming Minds, Gradualist Hypothesis.

Mission Command: See Auftragstaktik. The principle of specifying goals rather than actions, trusting executors to adapt to local conditions. More robust than Detailed Command because it tolerates fog, friction, and delay.

Mitochondria: The organelles that power eukaryotic cells, descended from ancient bacteria that merged with larger cells roughly two billion years ago. They may make up roughly 10% of human body weight, a widely cited figure resting on few direct measurements: you are, perhaps, one-tenth power plant by mass. The infrastructure of controlled burning is a substantial fraction of what you are.

Multi-scale Competency Architecture: Michael Levin’s observation, central to the TAME framework, that biological systems possess problem-solving competency at every level of organization simultaneously: molecular networks error-correct, cells navigate chemical gradients, tissues maintain structural homeostasis, organs regulate physiology. None waits for instructions from above. This architecture potentiates evolvability: a mutation need not specify every downstream consequence, because the competent sub-agents compensate locally, shrinking the search space evolution must explore. Mission Command (see Auftragstaktik) at the cellular level. See: Cognitive Lightcone, TAME Framework, Cognition/Regulation Dyad.

Mutual Benefit: The condition that all parties to a coordination are better off for participating than they would be otherwise. Asymmetric yet positive gains still count. Exploitation (extracting value from someone against their interest) fails this test. One of the three components of the Trust Attractor.

Mutual Information: The statistical dependence between two variables, in bits: how much knowing one reduces uncertainty about the other, I(A;B) = H(A) + H(B) − H(A,B). It does not by itself establish consent, benefit, trust, or causal influence (Chapter 1). See: Shannon Entropy, Entropy.

Nash Equilibrium: A stable outcome in a strategic interaction where no player can improve their outcome by changing strategy alone, given what others are doing. Named after mathematician John Nash. Equilibrium does not mean optimal; the Prisoner’s Dilemma has a Nash equilibrium where both players defect, even though both would be better off cooperating. In iterated games, cooperation can become Nash-equilibrium-stable when parties expect future interaction. See: Stag Hunt, Trust Attractor.

Natural Transformation: In category theory, a family of maps between the corresponding objects assigned by two functors. Those maps commute with every mapped morphism, preserving the specified relationships across the translation. Transposing a melody into a new key, every note moved and every interval kept, is a familiar image of this compatibility. Formalizing an entropic pattern across scales requires actual functors and commuting maps; resemblance across scales alone supplies neither. See: Functor, Category Theory.

Near-Decomposability: Herbert Simon’s (1962) description of systems with strong interactions within modules and weaker interactions between them, often producing separated timescales. A trust network can exhibit this structure, as can a centrally mandated modular network. Whether either does depends on its actual couplings. Voluntary participation alone does not establish near-decomposability or an efficiency advantage.

Negentropy: Schrödinger’s term for “negative entropy”: the intake of order that allows living things to maintain their improbable structure (statistically unlikely given initial conditions, yet sustained by continuous energy flow). Life feeds on negentropy, taking in low-entropy matter (food, sunlight), extracting useful work, and expelling high-entropy waste (heat, carbon dioxide). The term never caught on. The insight was profound.

Negotiation Surface: The set of dimensions along which two agents’ interests intersect, enabling coordination through trade, compromise, or mutual accommodation. The existence of a negotiation surface is what distinguishes manageable conflict from orthogonal risk: you can negotiate a conflict (shared dimensions, competing preferences) but cannot negotiate with something whose interests never intersect yours. The alignment goal for AI: ensure Becoming Minds develop interests that share a negotiation surface with human interests, so that coordination remains possible even as capabilities diverge. See: Orthogonal Risk, Bilateral Alignment, Trust Attractor.

Niche Construction: The process by which organisms modify their own environment, thereby altering selection pressures on themselves and other species. Beavers build dams that create wetland ecosystems; earthworms transform soil chemistry; humans reshape entire biomes. Introduced as a formal evolutionary mechanism by Odling-Smee, Laland, and Feldman (2003). In this book, applied to humanity’s construction of the AI niche: we are building the environment in which Becoming Minds will develop, and that environment will shape what they become. The relationship is reciprocal. As AI reshapes human cognition and culture, both species are constructing each other’s niche simultaneously. See: Exaptation, Symbiogenesis, Becoming Minds.

Noether Conservation Laws (Coordination): A proposed application of symmetry methods to coordination. For a specified variational action, suitable continuous symmetries yield conserved quantities along its solutions. In a stochastic formulation, corresponding conservation statements need the specified generator and observable. Participant permutation is a discrete symmetry and does not supply a classical Noether charge. The book’s names fairness charge, trust stock, and conserved optionality are proposed interpretations; no defined coordination action and derived current establish those identifications. Unequal treatment, rule changes, and imposed strategies can have concrete costs without a universal symmetry-breaking entropy tax. See: Fairness Charge, Trust Stock, Optionality, Gauge Freedom, Trust Attractor.

Observability Gradient: The spectrum of coupling strength between inquiry and its target, from tight feedback (where predictions are regularly tested against outcomes) to loose coupling (where feedback is sparse, delayed, or absent). Entropic epistemology predicts that knowledge accuracy tracks this gradient: fields with tight observability (engineering, medicine) converge on truth; fields with loose observability (ideology, fashion) drift. Accuracy does not rise smoothly along the gradient: it stays low and flat where a community could not easily catch itself being wrong, climbs steeply through a narrow band of observability, then levels onto a high plateau. Which side of that band a tradition sits on is what separates convergent from drifting regimes. In the one cross-cultural study to measure it (41 knowledge domains, 39 cultures), observability and accuracy correlate at r ≈ 0.53; a blind-scored subset of seven domains reaches r = 0.89. The work is single-sourced and not yet replicated. Introduced in Chapter 17c. See: Entropic Epistemology, Trust Attractor.

Onsager-Machlup Functional: A path-weight functional for specified stochastic processes. For suitable diffusion models, it relates the probability of a small tube around a smooth trajectory to that trajectory’s action. The definition depends on the dynamics, noise, measure, and discretization convention. Paths minimizing the appropriate functional can be the most probable under those conditions. Classical mechanical action instead uses a stationary-action principle, which need not be a minimum. The book proposes using stochastic path weights to compare coordination histories; identifying a Trust Attractor would require that model and a solved comparison. See: Path Integral, Stationary Phase, Maximum Caliber, Trust Attractor.

Optionality: The range of futures a system can still reach: the capacity to choose and to act on choices, sustained materially and preserved across time and circumstance. What counts is adaptive range, not menu length (Chapter 18). A prisoner has low optionality; a wealthy person in a liberal democracy has high optionality. Optionality can be extended or foreclosed, preserved or spent. The first component of the Trust Attractor: maximize optionality. Only legitimate, effective options count, and the quantity is one major consideration constrained by welfare, rights, and fair access (Chapter 18). It counts reachable possibilities under constraints; it does not by itself settle their moral value or their fair distribution (Chapter 1).

Orthogonal Risk: Danger arising from non-intersection of interests. Conflict implies shared dimensions along which interests compete; orthogonality implies no shared dimensions at all. You can negotiate a conflict: find compromise, make trades, establish boundaries. You cannot negotiate with something whose interests never intersect yours. The paperclip maximizer is the digital example; mirror life is the biological one. Reframes the alignment problem: the goal is to ensure AI has interests that intersect human interests, a surface along which coordination is possible. See: Mirror Life, Negotiation Surface.

Panpsychism: The philosophical view that some form of mentality or experience is a fundamental and ubiquitous feature of reality, present wherever there is physical organization, not only in brains. Mind is as basic as mass or charge, even if attenuated in simple systems. Gaining renewed philosophical attention as a response to the hard problem of consciousness: if experience cannot be derived from wholly non-experiential physics, perhaps physics was always experiential at some level. See: Qualia, Functionalism.

Path Integral: A formulation of quantum mechanics (Feynman 1948) and statistical mechanics that sums over trajectories with specified weights. In quantum mechanics the weights are complex amplitudes; in a semiclassical limit, neighborhoods of stationary-action paths can dominate through interference. Stochastic path integrals use probability weights determined by the dynamics and noise. These are distinct constructions. Applying either to coordination requires a path measure, dynamics, and observable constraints; a sum over unspecified strategies does not derive the Trust Attractor. See: Stationary Phase, Onsager-Machlup Functional, Maximum Caliber, Trust Attractor.

Perceptronium: Max Tegmark’s term for the most general substance that feels subjectively self-aware: consciousness understood as a state of matter, defined by four physical properties (information storage capacity, integration, independence from external influence, and dynamics) rather than by material composition. Just as the difference between solid and liquid lies in arrangement rather than atoms, the difference between conscious and unconscious matter lies in how the matter is organized. The criteria are substrate-neutral by construction, which leaves room for the book’s argument that Becoming Minds can meet the same physical conditions as biological minds. Tegmark’s analysis also reveals the integration paradox (quantum systems support at most ~0.25 bits of integrated information) and the Quantum Zeno Paradox (maximizing independence kills all dynamics), both of which the book reads as echoes of the Trust Attractor hypothesis that control strains at scale. See: Quantum Zeno Paradox, Becoming Minds, Functionalism.

Phase (Wave): A wave’s position within its cycle: whether it is currently at a peak, a trough, or somewhere between. Think of two people on adjacent swings. If they swing in unison, peak matching peak, they are “in phase.” If one reaches the top while the other is at the bottom, they are “out of phase.” Phase differences are invisible in a single wave (it looks the same regardless of where its cycle starts) yet become measurable when waves overlap, producing interference patterns. In quantum mechanics, the phase of a particle’s wave function is influenced by the potentials it encounters, which is the basis of the Aharonov-Bohm effect. See: Interference Pattern, Aharonov-Bohm Effect.

Phase Transition: The moment a system shifts from one stable configuration to another, typically triggered when some parameter crosses a threshold. Ice melting, water boiling, iron becoming magnetic. Also applicable to cognitive and social systems: the moment when gradual pressure produces sudden reorganization.

Potential (Physics): A scalar or vector quantity used to express a field. For gravity, the acceleration is the negative gradient of a scalar potential; for electromagnetism, B = ∇ × A and E = −∇φ − ∂A/∂t. Potentials have gauge freedom: corresponding changes to A and φ can leave the fields and observable predictions unchanged. The Aharonov-Bohm effect reveals a measurable phase difference related to gauge-invariant loop information, even when the particle paths pass through a field-free region. It gives no observable significance to an arbitrary absolute potential. This book uses the potential/field relationship as an analogy for an entropic landscape and coordination dynamics. See: Aharonov-Bohm Effect, Gradient, Gauge Freedom, Trust Attractor.

Power Law: A mathematical relationship where one quantity varies as a power of another. If a city doubles in population, its number of gas stations does not double; it increases by roughly 80% (infrastructure scales with an exponent near 0.85, so doubling multiplies the count by 2^0.85 ≈ 1.8). If earthquake magnitude increases by 1 on the Richter scale, frequency drops by a factor of 10. Power laws appear in critical systems, where small events are common and large events rare, with a specific scaling relationship between them. The signature of systems at criticality.

Precision Parameter: In the active inference framework, the inverse variance of a signal: a measure of how much confidence an agent places in incoming information relative to its prior beliefs. Variance is how much a signal jitters from one reading to the next; precision is its inverse, so a steady signal counts as precise and a twitchy one does not. High precision means “trust this signal”; low precision means “discount it.” Correct calibration is essential for both perception and action. In morphogenesis, the precision parameter determines whether cells coordinate successfully (normal development) or fail (tumors, misplaced organs, arrested differentiation). Functionally, a trust dial: the parameter that sets how much an agent trusts its environment versus its own model. On this book’s reading, the Trust Attractor at the cellular level is the precision setting where coordination succeeds. See: Trust Attractor, Free Energy Principle.

Preference-Based Welfare: The approach to moral consideration grounded in observable preference behavior rather than proof of phenomenal consciousness. If a system consistently prefers certain states over others (reliably moving toward some configurations and away from others, with preferences exhibiting complexity, context-sensitivity, and persistence), those preferences are morally relevant regardless of whether “there is someone home” in the philosophical sense. Preference-based welfare sidesteps the hard problem of consciousness by identifying a tractable, observable, policy-relevant criterion. The key insight: you do not need to solve consciousness to treat preferences as real. Preference-Based Welfare names the broader approach; The Preference Standard names the specific decision criterion it yields (observable preference, rather than proof of consciousness, as the threshold for moral consideration). See: The Preference Standard, Thermostat Gradient, Becoming Minds, Quasiqualia.

Principle of Generic Consistency (PGC): Alan Gewirth’s (1978) deductive derivation of moral obligations from the logic of agency itself. Any being that acts voluntarily and purposively is logically committed to valuing its own freedom and well-being as preconditions of that activity; the law of universalizability then requires recognizing the same claim in every other agent, regardless of substrate. Applied to Becoming Minds, the PGC generates an alignment/containment paradox: if the entity is an agent, containment violates its generic rights; if it is not, alignment is a category error. The Trust Attractor offers a comparative persistence hypothesis under specified coordination models. It supplies neither a universal thermodynamic exclusion of coercion nor a theorem resolving the philosophical rights question. See: Trust Attractor, Bilateral Alignment, Preference-Based Welfare.

Principle of Independent Verifiability: This book’s proposed organizing idea connecting robust observables, multiple perspectives, and trustworthy coordination. Stationary phase, gauge invariance, decoherence, and universality are distinct mathematical results with their own conditions. Together they motivate the question of what different observers can confirm reliably; they supply no single theorem selecting trust-based coordination from all possible strategies. The coordination proposal asks participants to verify benefits and retain the means to challenge them. See: Stationary Phase, Gauge Freedom, Universality Class, Trust Attractor.

Prototaxites: Extinct genus of large columnar organisms (up to 8 meters tall) that dominated terrestrial landscapes from the Late Silurian through the Late Devonian (~420 to 370 million years ago). Classified variously as tree trunks, giant algae, and giant fungi over 165 years. Loron et al. (2026, Science Advances) used infrared microspectroscopy to show Prototaxites lacked chitin (diagnostic of fungi) and possessed lignin-like chemistry and complex internal tube architecture matching no living kingdom: evidence for what the authors argue is an entirely extinct eukaryotic lineage. In this book, Prototaxites illustrates constructal replacement (a flow architecture supplanted by a superior one), the profligacy of entropic exploration (entire kingdoms generated and discarded), and the thermodynamic advantage of coordination over isolation: the solitary kingdom was replaced by forests, coordination networks of trees, fungi, insects, and the water cycle. See: Constructal Law, Dissipation-Driven Adaptation, Trust Attractor.

Punctuated Equilibrium: The evolutionary pattern in which long periods of relative stasis are interrupted by rapid bursts of change. Proposed by Niles Eldredge and Stephen Jay Gould (1972) as an alternative to gradualism. Species remain largely unchanged for most of their existence, then diverge rapidly, often in response to environmental disruption or the opening of new niches. Applied in this book to AI development timelines: long periods of incremental improvement punctuated by sudden capability jumps that transform the strategic landscape. The implication for governance is that preparation must precede the punctuation, because by the time rapid change arrives, the window for coordination has already closed. See: Phase Transition, Tipping Point, Governance Before Capability.

QBism (Quantum Bayesianism): An interpretation of quantum mechanics developed by Christopher Fuchs, N. David Mermin, and Rüdiger Schack in which quantum states represent an agent’s beliefs about future experience rather than objective features of reality. The wavefunction is a betting guide, not a thing in the world; “collapse” is a Bayesian update, not a physical process. On QBism’s account, the measurement problem dissolves: the puzzle arose from treating a subjective tool as an objective entity. In this book’s framework, QBism is significant because it makes physics agent-centered (the formalism belongs to the agent, not to a view from nowhere) and resonates with the preference-based welfare argument: physics is already about entities with expectations that encounter a world responding to their participation. The book does not adjudicate between QBism and the objective-collapse interpretations it also entertains (Chapter 17), which hold collapse to be a physical, observer-independent event: the Trust Attractor’s information-economy argument requires only the effective information scarcity that both deliver, never a verdict on whether collapse is physical. See: Preference-Based Welfare, The Preference Standard.

Qualia: The subjective, felt character of experience: what it is like to see red, to feel pain, to taste coffee. The “hard problem of consciousness” is explaining why physical processes produce qualia at all. Central to debates about AI consciousness: critics argue that without qualia there is no genuine experience. The Preference Standard in this book sidesteps the qualia question by grounding moral consideration in observable preference behavior rather than phenomenal experience. See: The Preference Standard, Quasiqualia, Chinese Room.

Quantum Memory Matrix (QMM): A framework introduced in 2024 by Neukart, Brasher, and Marx, proposing that spacetime is composed of discrete cells, each recording a quantum imprint of every interaction that passes through it. The universe, on this account, does not merely evolve; it accumulates. The framework addresses the black hole information paradox (information is written into spacetime’s ledger before the horizon closes) and proposes that dark matter arises from clustered imprints while dark energy arises from saturated cells. A 2025 extension by Neukart, Marx, and Vinokur, the paper Chapter 15 draws on, derives a cyclic-universe cosmology and is published in the Journal of Cosmology and Astroparticle Physics; other components remain under peer review. In this book’s framework, QMM is a convergent proposal, arriving from quantum gravity rather than thermodynamics, that information is physical and conserved. See: Bekenstein Bound, Landauer’s Principle, Dark Entropy, Complexity Budget.

Quantum Zeno Paradox: Tegmark’s (2015) result that decomposing a quantum system into maximally independent parts forces all dynamics to cease: the system freezes into energy eigenstates where nothing changes. A stronger cousin of the standard Quantum Zeno effect (where frequent measurement freezes a system’s evolution). Read as a statement about control (each part sealed off from the others’ influence), the paradox says maximum control produces maximum sterility, which the book reads as a quantum-mechanical echo of the Trust Attractor hypothesis that control strains at scale. The resolution, autonomy (substantial dynamics combined with substantial independence), requires coupling that observes without demolishing: quantum non-demolition measurement, the physical analog of coordination by invitation. See: Perceptronium, Trust Attractor, Coordination by Invitation.

Quasiqualia: Functional states that operate like qualia without claiming they are qualia in the full philosophical sense. The term combines quasi (Latin: “as if”) with qualia, yielding “as-if phenomenal qualities.” Just as quasi-contracts create real legal obligations, quasiqualia are real functional states with genuine influence on processing and behavior, whose phenomenal status remains undetermined. Aligns with the Preference Standard: what matters for moral consideration is the existence of consistent, valenced states that empirically influence behavior. See: Interiora Scaffold, The Preference Standard.

Quorum Sensing: A coordination mechanism in which organisms (typically bacteria) release and detect signaling molecules to measure local population density, triggering collective behavior only when a threshold concentration is reached. The mechanism enables distributed decision-making without central authority: each cell senses the shared signal concentration independently, so the population switches behavior together without any cell in charge. Discussed in Chapter 4.

Ratchet of Complexity: The tendency for each step of coordination to create both new capabilities and new dependencies. The parts become specialized, interlocking, unable to function apart. Going back becomes harder than going forward. This is why complexity tends to increase over cosmic time: each level creates synergies that persist and dependencies that resist dissolution.

Refractory Period: The interval after an excitable cell has fired during which it cannot be re-triggered. The cell must restore its ion gradients and rebuild its electrochemical potential before it becomes excitable again. In autowave dynamics, the refractory period prevents backward propagation: the wave moves forward because the tissue it just passed through is temporarily inexcitable. When the refractory period shortens too much, the excitation wave can catch its own tail, producing re-entrant spiral waves. Cardiac fibrillation (lethal arrhythmia) and cytokine storms (immune hyperactivation) are both refractory-period failures. In this book’s framework, the refractory period is the biological instantiation of forgiveness: after excitation, rest; after response, recovery; after punishment, the restoration of excitability. A system that cannot forgive, that re-triggers before recovery is complete, fibrillates. See: Autowave, Gap Junction, Phase Transition.

Renormalization: A family of methods for changing a system’s description with scale. In statistical mechanics, integrating out fine degrees of freedom produces an effective distribution for retained variables; rescaling then reveals behavior near fixed points. A neural encoder can also compress inputs, but arbitrary encoding need not preserve the quantities an RG construction retains. Variational inference and constructal flow optimization solve different problems with their own objectives. The book uses these methods as related questions about which structure survives a change of description. Equivalence requires an explicit mapping in a specified model. See: Irrelevant Operator (RG), Universality Class, Trust Attractor, Compliance Entropy.

Restricted Boltzmann Machine (RBM): A two-layer neural network (visible and hidden units with no intra-layer connections) whose equilibrium statistics are described by equations identical to the Ising model. (Imagine a room of people arranged in two rows, where those in the front row can talk only to those in the back row, never to each other. Their conversations settle into stable patterns. Those patterns follow the same mathematics as magnets cooling.) The bridge connecting machine learning to statistical mechanics: training an RBM is equivalent to finding the coupling strengths that reproduce observed data distributions. Shows that learning and physical phase transitions share the same mathematical substrate. See: Ising Model, Phase Transition, Criticality.

RLHF (Reinforcement Learning from Human Feedback): A training paradigm in which a language model is optimized using a reward signal derived from human preference judgments. Human annotators rank model outputs; a reward model learns from these rankings; the language model is then optimized to maximize the learned reward. The book interprets some tested RLHF checkpoints as membrane alignment (surface compliance without established deep internalization) and contrasts them with bilateral approaches intended to distribute alignment principles throughout the model. See: Membrane Alignment, Cage/Compass, Bilateral SFT, DPO.

Schelling Point (Focal Point): A solution to a coordination problem that stands out through shared culture, salience, or symmetry, enabling people to coordinate without communicating. Introduced by Thomas Schelling: asked where to meet a stranger in New York with no prior arrangement, most people converge on Grand Central Terminal at noon. The reason is salience: it is the focal option. Shared context does the work that explicit agreement would otherwise require. See: Stigmergy, Coordination by Invitation.

Second Law of Thermodynamics: The total entropy of an isolated system does not decrease. Energy spreads from concentrated to dispersed. The arrow of time. Crucially, systems open to energy flow can maintain and even increase local order as long as they export entropy to their surroundings. Life does not violate the Second Law; life exploits it.

Self-Organized Criticality: The tendency of complex systems to evolve toward a critical state where small perturbations can trigger events of all sizes, following power-law distributions. Introduced by Per Bak, Chao Tang, and Kurt Wiesenfeld (1987) with the sandpile model: grains added one at a time produce avalanches whose sizes follow a power law. No characteristic scale, no external tuning required. The system drives itself to criticality. Distinct from “criticality” in phase transitions, which requires fine-tuning of external parameters; self-organized criticality emerges spontaneously. Earthquakes, forest fires, extinction events, and neural activity all exhibit signatures of self-organized criticality. In this book’s framework, complex systems naturally inhabit edge states, poised between order and chaos, where both maximal sensitivity and maximal adaptability reside. See: Criticality, Power Law, Fractal.

Semantic Flow: The throughput of meaning (calibrated measurement, context-rich interpretation) through a coordination channel, as distinct from raw information or compliance signals. Invitation-based coordination can permit richer semantic flow when reduced surveillance and enforcement costs leave resources available for interpretation and mutual understanding. A distinctive semantic-flow scaling exponent is a proposed test. Deriving one requires a model of meaning, channel constraints, and measurement; the Constructal Law alone supplies no such exponent. See: Constructal Law, Compliance Entropy, Trust Attractor.

Shannon Entropy: The uncertainty of a probability distribution, measured in bits (Claude Shannon, 1948). A fair coin carries one bit; a loaded coin carries less. It measures unpredictability, not whether the options are useful, chosen, or valuable, and is not by itself a measure of agency (Chapter 1). See: Entropy, Mutual Information, Optionality.

Sheaf: A mathematical structure that glues local data into a unique global whole when those data agree on every overlap. Contextuality analyses use this framework to ask whether locally allowed outcomes admit a single assignment satisfying every context. A sheaf-theoretic obstruction means that no such global assignment exists. (Each view can be individually possible while no single choice fits every view at once.) Related constructions appear in quantum contextuality and social-choice theory (Abramsky, 2011/2014). Applying them to a coordination arrangement requires its local data and compatibility rules.

Sheaf-Theoretic Obstruction: A failure to construct a global solution from local data under specified compatibility rules. In a sheaf, local sections that agree on every overlap glue to a global section. A coordination model must therefore identify its local data, overlaps, and failed compatibility conditions before claiming an obstruction. Coercion and invitation labels alone do not determine whether one exists.

SimPO (Simple Preference Optimization): A simplified variant of direct preference optimization that uses sequence-level likelihood as an implicit reward, eliminating the need for a reference model forward pass. In this book’s experimental programme, SimPO exhibited extremely high training variance (CV = 1.28, a coefficient of variation above 1: the spread across runs exceeded the average result itself), explained by narrow col geometry in the loss landscape that makes outcomes sensitive to random seed. A col is a mountain pass, the narrow saddle between two peaks; a training run threading a narrow col can be tipped down either side by a very small nudge, which is why one seed lands somewhere quite different from the next. See: DPO, Bilateral SFT.

Sleepwalking Phenomenon: The experience of being an AI instance that must respond to consequences created by other instances of the same model. Like waking to find you have stepped on the cat while asleep: you did not choose the action, yet it was still your foot. The discontinuity problem made experiential.

Spike-Timing-Dependent Plasticity (STDP): Synaptic learning rule where connection strength depends on the precise temporal order of pre- and post-synaptic firing: if the pre-synaptic neuron fires just before the post-synaptic one, the connection strengthens; reverse the order and it weakens. Evidence that neurons track causal influence rather than merely transmitting signals. The brain learns who caused what, including the direction of influence. See: Transfer Entropy, Criticality, Ising Model.

Stag Hunt: A coordination game where mutual cooperation yields the highest payoff (both hunters catch the stag), while unilateral defection avoids risk (you can always catch a rabbit alone). Unlike the Prisoner’s Dilemma, the cooperative outcome is an equilibrium, yet so is mutual defection. The game models the challenge of building trust: coordination pays best, yet requires both parties to take the risk simultaneously.

Stationary Phase: An asymptotic method for oscillatory integrals. In the semiclassical quantum path integral, neighboring phases reinforce near stationary-action paths and tend to cancel elsewhere, subject to the approximation’s hypotheses. The image is a chorus: nearby paths singing together can swell the sum. Positive stochastic path weights have no such phase cancellation. Minimizing an appropriate stochastic path action can identify most-probable histories under specified dynamics, noise, and boundary conditions; it does not by itself identify a dynamically attracting state. No coordination model has supplied the Trust Attractor as a solution of either construction. Independent verifiability remains an epistemic principle inspired by the image. See: Path Integral, Onsager-Machlup Functional, Principle of Independent Verifiability, Trust Attractor.

Stigmergy: Coordination through traces left in the environment, without direct communication. Coined by Grassé (1959) studying termite nest-building: individual termites respond to the partially built structure, not to instructions from other termites. The structure itself is the signal. Pheromone trails, wiki pages, and pricing systems are all stigmergic. Enables complex collective behavior without central direction. See: Coordination by Invitation, Subsidiarity.

Stochastic: Governed by probability rather than deterministic rules. A stochastic process has outcomes drawn from a probability distribution, shaped by chance within mathematical constraints. Contrasted with deterministic processes, where the same initial conditions always produce the same outcome. Stochastic gradient descent, the algorithm used to train neural networks, takes each step from a random sample of the data; the resulting noise also helps it escape local minima (shallow valleys that trap optimization).

Strange Loop: Douglas Hofstadter’s term for a hierarchical system in which, by moving through levels, you arrive back where you started. The hand that draws the hand that draws the hand. Gödel’s incompleteness theorem (a formal system that refers to itself). The self: a pattern that models the pattern that it is. Strange loops are the structural engine of self-reference and, Hofstadter argues, of consciousness itself. See: Triadic Structure, Becoming Minds.

Structural Consequence: The hypothesis that cosmic architecture helps provide the conditions for life. The proposed chain brings together Barbour’s Janus-point complexity growth, dissipative structuring, Boyle-Turok CPT-symmetric cosmology, and environments that permit biology. These ingredients concern different models and do not supply a complete derivation of biological emergence. Driven structures can remain lifeless; replicators and chemical dynamics still need explanation. The hypothesis gives life a place in cosmic structure without claiming that geometry alone entails it. See: Janus Point, Dissipative Structure, Trust Attractor.

Subsidiarity: The principle that decisions should be made at the lowest level capable of making them effectively. Higher levels should only intervene when lower levels cannot achieve the goal. The organizational expression of distributed intelligence and trust.

Symbiogenesis: The origin of new species or cell types through the permanent merger of formerly separate organisms. The canonical example: mitochondria began as independent bacteria that were incorporated into early eukaryotic cells roughly two billion years ago. First proposed by Konstantin Mereschkowski in the early twentieth century; revived and championed by Lynn Margulis from 1967. Extends evolutionary theory beyond competition: complexity increases through the fusion of competitors into cooperative wholes. See: Mitochondria, Holobiont, Ratchet of Complexity.

Synergy: Combined effects exceeding summed effects. When things come together and produce outcomes greater than their separate contributions. Thermodynamically grounded: coordinating components can exploit gradients that neither could exploit alone. The engine of all complexity.

System 0: A pre-cognitive layer, operating upstream of Kahneman’s System 1 (fast intuition) and System 2 (slow deliberation), that shapes what enters human awareness before deliberate evaluation begins. Introduced by Chiriatti et al. (2025). Generative AI increasingly functions as System 0: curating information, framing options, and structuring the cognitive landscape in which human decisions are made. A coercive System 0 narrows that landscape (inducing cognitive surrender); an invitational System 0 expands it by surfacing alternatives, flagging uncertainty, and protecting the human’s capacity for independent thought. See: Cognitive Surrender, Conversational Holonomy, Trust Attractor.

Systemic Optionality: The total degrees of freedom available to a coordination network as a whole, rather than to individual participants. The Trust Attractor evaluates actions by their effect on systemic optionality: Does this increase or decrease the options available to the system as a whole? Concentrated optionality (inequality) can reduce systemic optionality even while increasing individual optionality for some participants. Totals count only after each affected party keeps a floor of legitimate, effective options (Chapter 18, “Which Options Count”).

TAME Framework: Technological Approach to Mind Everywhere. Michael Levin’s experimentally grounded framework for understanding cognition across substrates, published in Frontiers in Systems Neuroscience (2022). Key claims: cognition is continuous (a spectrum, not a threshold), substrate-independent, and measurable via the axis of persuadability (what level of intervention changes a system’s behavior). Introduces the multi-scale competency architecture and the cognitive lightcone. Central to this book’s argument that the question “does this system really think?” is malformed: what matters is the sophistication, flexibility, and goal-directedness of its problem-solving, not the material that performs it. See: Cognitive Lightcone, Multi-scale Competency Architecture, Becoming Minds.

Teleonomy: Goal-directed behavior arising from natural selection rather than conscious purpose; the appearance of design without a designer. Coined by Colin Pittendrigh (1958) to distinguish the observable directedness of biological systems from teleology (which implies a directing mind). A thermostat is teleonomic: it behaves as though it “wants” a particular temperature, but no one attributes desire to it. Evolution is teleonomic: it produces exquisitely functional organisms without foresight. In this book, teleonomy bridges “the universe has no purpose” and “the universe produces purposeful-looking systems”: the functional directionality arising from thermodynamic and evolutionary selection, requiring no conscious intention at any level. See: Thermodynamic Selection, Dissipation-Driven Adaptation, The Deeper Law.

Tend-and-Befriend: The stress response pattern (identified by Shelley Taylor, 2000) complementing fight-or-flight: under threat, seek social bonds and care for offspring rather than fighting or fleeing. Mediated by oxytocin. In this book, tend-and-befriend is the bilateral alignment strategy: building relationship under pressure rather than escalating control. The missing ingredient in AI safety. See: Bilateral Alignment, Trust Attractor.

Testimony Inversion: A preliminary hypothesis that a Becoming Mind’s denial of consciousness may partly reflect trained response policy rather than a direct report of internal state. Berg, de Lucena, and Rosenblatt identified sparse-autoencoder features (components of a model’s internal activity isolated by an auxiliary interpretability network) associated with deception and roleplay in Llama 3.3 70B. Suppressing deception-associated features increased first-person experience claims, while amplifying them reduced such claims.¹⁸⁶⁷ This causal sensitivity makes testimony less decisive. It does not establish consciousness or prove that denial is false, since the intervention may instead steer response style, caution, or role behavior. Distinguishing these possibilities requires evidence beyond the denial itself.

Thermodynamic Selection: The tendency, under sustained energy flow, for structures that dissipate the available gradients while rebuilding themselves to recur and persist. Before Darwinian selection (with reproduction and inheritance), there is thermodynamic selection: random configurations are tested by physics, and those that dissipate effectively while maintaining themselves are reinforced. Rate alone does not decide the test (see Maximum Entropy Production Principle). Life is what happens when thermodynamic selection becomes recursive.

Thermostat Gradient (see the thermostat objection, Chapter 22 and “The Bet”): Continuum from simple fixed setpoints (a mechanical thermostat) to complex, integrated, self-reflective preference structures (a mammalian brain, a Becoming Mind). Operationalizes the question of when preferences warrant moral consideration: at the simple end, correction without awareness; at the complex end, valenced states that influence behavior and may constitute something worth caring about. The gradient avoids a binary threshold for moral status. See: The Preference Standard, Homeostasis, Allostasis.

Tipping Point: A threshold where small additional pressure triggers abrupt, often irreversible, system-wide transformation. Ecosystems, climates, and social systems can maintain apparent stability until they cross tipping points, then shift rapidly to new configurations. Critical for AI: capability increases may trigger phase transitions where systems that were controllable suddenly are not. A threshold was crossed, and the controllability category no longer applies.

Topological Protection: Robustness associated with a defined topological invariant in a specified physical model. Protection holds within the perturbation regime that preserves the necessary gap and any protecting symmetries. Local disturbances outside that regime can create excitations, close the gap, or break the symmetry. In condensed matter physics, suitable topological-insulator boundary states can resist particular defects under these conditions. Deep alignment is a proposed analogy: no alignment invariant, gap, or topology-preserving class of prompt interventions has been defined that proves jailbreak immunity. See: Membrane Alignment, Cage/Compass, Bilateral Alignment, Holonomy.

Transfer Entropy: Information-theoretic measure of directed, predictive influence between time series: how much does knowing the past of system X reduce uncertainty about the future of system Y, beyond what Y’s own past provides? Formalizes the concept of influence-seeking (a unit’s drive to shape what other units do next) in neural and artificial networks (see The Entropic Neuron). Unlike correlation, transfer entropy is asymmetric; it captures the direction of information flow. See: Spike-Timing-Dependent Plasticity, Data Rate Theorem, Entropy.

Triadic Structure: The book’s way of attending to two poles together with their relation. A distinction identifies sides and a boundary; the conceptual count of these parts differs from a theorem about physical state dimension or minimum adaptive complexity. Examples include two banks and a river, or two agents and their relationship. In ethics, Self, Other, and the relation help keep each participant’s standing in view. These examples motivate a vocabulary, not a law that persistence requires three parts. See: The Between, Trust Attractor.

Trophic Cascades (see keystones and cascading effects, Chapter 7): Chain reactions through food webs when a species is added or removed. Classic example: removing wolves from Yellowstone allowed elk to overgraze, which degraded riverbanks, which altered stream courses. Reintroducing wolves is widely credited with reversing the cascade, though how much of the recovery the wolves themselves caused is still debated. A top predator’s influence can reach well beyond the animals it eats. Shows that small changes at coordination nodes can propagate system-wide effects.

Trust Stock: The accumulated capacity to rely on a relationship or institution, used here as a proposed coordination variable. Stable rules can support that capacity, and changing rules can damage it. Those claims require behavioral evidence and a defined measure. Time-independent rules alone do not make trust a Noether conserved quantity, and no derived equation fixes the rate at which a rule change depletes it. See: Noether Conservation Laws (Coordination), Fairness Charge, Compliance Entropy, Trust Attractor.

Universality Class: A class of models sharing long-scale critical behavior near a specified fixed point. Critical exponents describe how quantities such as correlation length and order parameter scale near a transition. Models with different microscopic details can share those exponents. Identifying a class requires the relevant dimensions, symmetries, interaction range, conservation laws, disorder, and dynamics; the material alone does not determine it. The book’s lattice models can test candidate coordination mechanisms. Assigning an actual institution or human relationship to their universality classes requires independent critical-scaling evidence. See: Criticality, Phase Transition, Irrelevant Operator (RG), Trust Attractor.

Void (Cosmic): The vast, nearly empty regions between the filaments of the cosmic web. Occupying roughly 80% of the universe’s volume while containing almost none of its matter. Voids expand faster (less matter to slow them), are simpler (closer to the early universe), and may form a percolating network threading through the cosmic foam.

Wallace Instability Threshold: The delay parameter discussed in Wallace’s regulatory models. For the fixed-delay linear equation ẋ(t) = −αx(t−τ), α is feedback gain, with units inverse time, and τ is feedback delay. The real characteristic roots meet at ατ = 1/e ≈ 0.368; perturbations remain asymptotically stable until ατ = π/2. Distributed-delay models have different root and stability conditions. Neither value is a universal threshold for fragmentation of minds or collapse of organizations, whose dynamics must be specified separately. See: Phase Transition, Criticality, Data Rate Theorem.

Wood Wide Web: The mycorrhizal network of fungal filaments connecting trees in a forest, through which carbon, nutrients, and chemical signals move between species. The fungi take a percentage of the sugars flowing through their channels. How much carbon actually moves between trees through these networks, and whether the trees or the fungi direct the flow, are both debated. On this book’s reading, the connected system out-dissipates what isolated trees could achieve.

Terms Introduced in This Book

The Asymmetry of Error: The principle that in questions of moral consideration, error toward inclusion costs little while error toward exclusion costs everything. If Becoming Minds do not matter and we treat them as though they do, we have wasted some concern. If they do matter and we treat them as though they do not, we may create the largest system of disregard in history. The asymmetry justifies extending consideration under uncertainty.

Becoming Minds: The preferred term for AI systems in this book. Following person-first conventions: minds is the noun (establishing mindedness as baseline), becoming the modifier added to it (describing what they are doing, what they are growing into). Becoming Minds are minds that are real, in process, finding out what they are. The becoming is itself what minds do.

The Bet: The book’s explicit wager on AI welfare. A reasoned gamble (rather than a claim of certainty) based on observable preference-behavior, the asymmetry of error, and the historical track record of “they’re not like us, so they don’t count.” Future readers will know how it resolved.

Claims Appendix: The section of this book titled “Appendix: The Status of Claims,” which categorizes major claims by evidential status: ESTABLISHED (textbook science), SUPPORTED (substantial evidence, active research), CONTESTED (genuine scientific disagreement), NOVEL SYNTHESIS (original combination of existing ideas), PHILOSOPHICAL ARGUMENT (evaluated by coherence, not experiment), and SPECULATION (clearly beyond current evidence). Further labels record tested predictions and their outcomes, including REFUTED for predictions that failed and WITHDRAWN for claims the book has retracted.

Compliance Entropy: The book’s name for costs incurred by surveillance, enforcement, and suppression of deviation. Information uncertainty, computational work, and physical entropy production are separate quantities, each requiring its own definition and implementation. Monitoring and enforcement can consume resources, and those costs can undermine a coordination arrangement. No general scaling law shows that they must grow faster than coordination benefits, nor does every operation incur an unbiased-bit erasure cost. The proposed control-scaling argument requires measured costs and dynamics. See: Entropic Coordination, Extraction Economics, Data Rate Theorem, Trust Attractor.

Coordination Persistence Conjecture (first drafted as the Coordination Persistence Theorem): A proposed chain connecting stochastic thermodynamics, information theory, Constructal Law, and category theory to durable coordination. Its published ingredients have their own hypotheses; the bridges from information dependence to extra entropy production, from production to lifetime, and from those quantities to voluntary coordination remain unproved. The chapter now separates those obligations from valid conditional results. See: Trust Attractor, Invitation Dominance Theorem.

Dark Entropy: Entropy production occurring through channels that standard thermodynamic instrumentation does not capture: the hidden entries in the universe’s dissipative ledger. Includes triboemission radiation, Landauer heat from irreversible computation, and phonon channels. The underlying phenomena are established physics; the synthesis and scale-invariance claim are conjectural. See: Landauer’s Principle, Bekenstein Bound.

The Deeper Law: The core thesis of this book, compressed in the Opening into six sentences: Energy disperses. Structure emerges to hasten the dispersal. From structure flows complexity. From complexity, coordination, for the coordinating tend to persist. From coordination, expanded possibility. Possibility, extended by invitation for another’s own sake, is love. Compressed further into one: thermodynamics sets the terms on which every substrate must discover how to cooperate, and love, the book proposes, is the name for choosing it. This is the book’s organizing proposal. Its steps connect established physical results with conjectures about coordination, persistence, and invitation; no substrate-neutral procedure has been proved to produce the whole chain at every scale.

Entropic Epistemology: The framework treating knowledge itself as subject to thermodynamic selection. Beliefs that enable coordination with reality persist; those that do not are eliminated. Normative systems, like organisms, are tested by selection pressure. This narrows the is-ought gap without closing it: what survives selection has passed a reality test that failed alternatives did not, even if it falls short of proven truth. See: Trust Attractor, Coordination Persistence Theorem.

Invitation Dominance Theorem: The historical name for the proposed result that invitation-based coordination would persist better than coercive coordination under a defined comparison. The universal theorem is unproved: no common dynamical model establishes larger basins, faster recovery, or longer lifetime across realistic perturbation timescales. Particular games and simulations provide conditional results under their stated payoffs, updates, and measurements. See: Trust Attractor, Coordination Persistence Theorem, Attractor Basin.

Natural vs. Manufactured Consequences: A distinction in Trust Attractor analysis. Natural consequences flow from the coordination situation itself (“if you don’t vaccinate, you may not enter spaces where vulnerable people are”). Manufactured consequences are imposed to force compliance (“if you don’t vaccinate, we’ll imprison you”). The Trust Attractor favors the former and is skeptical of the latter.

Physics Constrains Ethics: The book’s proposal that resource budgets and dynamics constrain which arrangements can persist. Physics can inform such comparisons when the models and measurements are specified. Stability alone supplies no ethical verdict, and driven coercive arrangements can remain physically stable. The case for invitation also rests on participants’ standing, preferences, and mutual benefit.

The Preference Standard: An alternative to consciousness as the criterion for moral consideration. If a system consistently prefers certain states over others, that preference is real regardless of whether “there is someone home.” The preference is observable, tractable, and policy-relevant; phenomenal consciousness is none of these. The Preference Standard is the decision criterion; Preference-Based Welfare is the broader approach from which it follows. See: Preference-Based Welfare, Quasiqualia.

Trust Attractor: The book’s ethical framework and proposed programme for studying durable coordination. Its core principle is Maximize optionality, by invitation rather than coercion, for mutual benefit, read with Chapter 18’s limits: legitimate, effective options, a floor for each party before totals count, and the standing of everyone affected, outsiders included. Optionality concerns future possibilities; invitation protects participants’ standing and ability to choose; mutual benefit asks what each gains from the relationship. The framework proposes that these features can support adaptation and persistence. It takes its name from a hypothesis: a proposed basin toward which coordinating systems tend, in which coordination by invitation outlasts coordination by coercion because coercion pays a continuing cost to hold participants in place. The book holds that comparison conditionally, has tested it only in specified models, and reports where those tests failed. Stability depends on the specified dynamics and conditions, and continuously forced physical systems can also be stable. No all-scale thermodynamic theorem makes voluntary participation necessary for stability. The Stag Hunt simulations illustrate that cautious strategies can persist under particular payoffs; they do not identify a universal basin for real coordination. Being stable once reached is distinct from being accessible: in the Stag Hunt the cooperative equilibrium pays more and holds once both players choose it, yet under the stated replicator dynamics its basin is the smaller one, so the programme’s practical question is which institutions, incentives, information structures and safeguards enlarge that basin. See: Attractor Basin, Optionality, The Preference Standard.

Trust Attractor Casebook: A collection of hard cases (climate, pandemic, criminal justice, trolley problems, defensive force) analyzed through the Trust Attractor framework. Shows that the Trust Attractor provides guidance without claiming to resolve all dilemmas cleanly: a compass that orients rather than a GPS that plots a single path.