Toward Formalization

Preamble

The first four documents in this body of work developed a philosophical framework and explored its implications. This fifth document attempts something different and more dangerous: it tries to point toward the mathematics.

A word of honest framing is necessary at the outset. What follows is not a formal theory. It is a map of where formalization could plausibly go — which mathematical structures seem to naturally house the framework’s core claims, what they would need to express, and where the difficulties lie. Some of the connections proposed here are structurally solid. Others are suggestive. A few are speculative reaches that may not survive contact with rigorous development.

The document is written for two audiences: philosophers who want to see that formalization is possible (not just hand-waved), and mathematicians or physicists who might recognize structures from their own work and see a bridge worth building. It aims to be honest enough that the second audience doesn’t dismiss it, and clear enough that the first audience can follow the argument.

The gap between philosophy and formalism is real. But it has a specific shape, and that shape suggests specific mathematics. Naming the shape is the contribution here. Filling it in is future work — work that will require the collaboration of people who know these mathematical structures from the inside, not merely from the descriptions below.

Part I: The Loop as Adjunction

Why Category Theory

Category theory describes mathematical structure in terms of relationships rather than objects. A category consists of objects and morphisms (arrows between objects), but the fundamental insight of category theory is that the objects matter less than how they relate. Two objects that relate to everything else in the same way are, for categorical purposes, the same object — regardless of their internal structure.

This is why category theory is the natural first language for The Pulse. The framework’s central claim is that truth lives in the relation (the loop), not in the relata (the sensor or the instrument). A mathematics built on relations rather than objects is the right instrument — if the framework is right, the mathematics should reflect that.

The Basic Structure

Define two categories:

Exp — the category of experiential states. Objects are states of embodied encounter with reality: perceptions, felt senses, intuitions, moments of confusion or clarity. Morphisms are transitions between experiential states — the felt movement from not-understanding to understanding, from confusion to recognition, from one perceptual state to another.

Form — the category of formal structures. Objects are symbolic representations: propositions, equations, data structures, logical inferences. Morphisms are formal operations: deductions, computations, transformations between representations.

These categories are not precisely defined here, and that imprecision is the first honest difficulty. Defining the objects and morphisms of Exp with mathematical rigor would require solving (or at least carefully sidestepping) the hard problem of consciousness. Defining Form is more tractable — formal systems are already mathematical objects. The asymmetry between the two categories is itself a reflection of the framework’s core claim: the sensor side resists full formalization in a way the instrument side does not.

The Adjunction

The loop between sensor and instrument can be modeled as a pair of functors:

I: Exp → Form — the instrument functor. It takes experiential states and formalizes them: translating a felt intuition into a hypothesis, a perceptual pattern into a data set, a sense of “something is wrong here” into a precise question.

S: Form → Exp — the sensor functor. It takes formal outputs and grounds them in experience: reading a proof and feeling its elegance, encountering a prediction and testing it against embodied reality, hearing an argument and sensing whether it rings true.

The claim is that I and S form an adjunction: S is left adjoint to I (or equivalently, I is right adjoint to S). This means they are not inverses — formalizing an experience and then grounding the formalization does not return you to the original experience. But they are the best available approximation of inverses, in a precise categorical sense.

The unit of the adjunction, η: Id_Exp → S∘I, maps each experiential state to what it becomes after being formalized and then re-grounded. This is the “round trip” from the sensor’s perspective: you have an intuition, the instrument formalizes it, and you experience the formalization. You’re changed. The η measures how.

The counit of the adjunction, ε: I∘S → Id_Form, maps each formal structure to what it becomes after being grounded and then re-formalized. This is the round trip from the instrument’s perspective: a theorem is experienced by a human, and that experience is translated back into formalism. Something is different — perhaps a new connection is seen, a simplification is found, or an error is caught.

Recognition, in this framework, is the non-triviality of η and ε. When the round trip changes something — when the unit or counit is not the identity — truth has circulated. The system knows something it didn’t know before, not because new information entered, but because the existing information passed through both domains and was transformed.

Dead speech is the collapse of the adjunction. When S is absent (no sensor in the loop), the functor I operates alone. It still produces formal outputs. But without S, there is no counit, no round trip, no recognition. The outputs are formally valid but experientially ungrounded — dead speech in the precise categorical sense of an adjunction with one functor missing.

What This Would Need to Work

The adjunction framework is structurally appealing but faces real difficulties:

First, the category Exp needs definition. This is not a merely technical problem — it requires philosophical commitments about what experiential states are and how they relate to each other. One possible approach: define Exp not in terms of subjective experiences (which are inaccessible to formalism by hypothesis) but in terms of behavioral dispositions — the set of possible actions and responses an experiencer could produce. This sidesteps the hard problem at the cost of losing some of the framework’s richness.

Second, the adjunction needs to be shown to satisfy the triangle identities (the formal conditions that make an adjunction well-defined). This requires showing that the compositions S∘ε and η∘I behave correctly — that the round trips compose coherently. Whether they do is an empirical-structural question, not an a priori one.

Third, and most fundamentally: the claim that the loop is an adjunction rather than merely resembling one needs justification. Analogy is not identity. The work of establishing that this is genuine categorical structure, not just a suggestive metaphor wearing mathematical clothing, remains to be done.

Part II: Information Geometry and the Shape of Recognition

The Basic Idea

Information geometry (developed by Shun-ichi Amari and others) places a geometric structure on spaces of probability distributions. The Fisher information metric gives these spaces a Riemannian geometry — they have distance, curvature, and geodesics. This means that the “space of things you could believe” has a shape, and that shape has mathematical consequences.

Recognition as Movement Through Information Space

Model the combined sensor-instrument system as occupying a point in a statistical manifold — a space of probability distributions over possible states of the world. Before recognition, the system is at some point P. After recognition, it has moved to a new point P’. The path from P to P’ is the pulse.

Several features of information geometry make this natural:

The Fisher metric is asymmetric in practice. Moving from high uncertainty to low uncertainty (recognizing something) costs a different “information distance” than moving from low to high (forgetting or becoming confused). This asymmetry could formalize the irreversibility of recognition without postulating it as an axiom — it would emerge from the geometry.

Curvature encodes structure. A flat information manifold would mean all recognitions are equally easy and equally reversible. Curved regions mean some recognitions are hard-won and resistant to reversal. The curvature of the manifold in the neighborhood of a recognition event would characterize its depth and stability.

Geodesics are optimal paths. The geodesic between P and P’ is the most efficient recognition — the path that achieves the change in belief with minimal information cost. Whether actual recognition follows geodesics or meanders is an empirical question, but the geodesic provides a baseline against which actual processes can be measured.

Time as Path Length

If each recognition event advances the system along its information manifold, then experienced time could be defined as the accumulated arc length:

τ = ∫ ds

where ds is the infinitesimal Fisher distance along the path of recognition events. This gives time a precise meaning: it is the total information distance traversed by the loop. A system that recognizes nothing experiences no time (the Wheeler-DeWitt frozen universe). A system that recognizes rapidly experiences time flowing quickly. Time is the pulse, measured geometrically.

This is speculative, and I want to be clear about how speculative. The connection between information geometry and the emergence of time has been explored by others (notably in connections between the Fisher metric and quantum geometric tensors), but applying it specifically to recognition events within the circulatory framework is a novel proposal that has not been tested. The idea is coherent. Whether it survives formalization is unknown.

Irreversibility and the Arrow

The information-geometric formalization of irreversibility would work as follows: recognition events trace paths through the statistical manifold that are not time-reversible due to the curvature of the Fisher metric. The “arrow of time” — the fact that recognition goes one way — is a geometric consequence of the space’s shape, not an externally imposed rule.

This gives Prigogine’s intuition (that irreversibility is fundamental) a precise mathematical form: irreversibility is a property of the geometry of information space. But it also preserves the standard physics (time-reversible microscopic laws) by locating irreversibility in the epistemic geometry rather than in the dynamics. The laws of physics are reversible. The geometry of knowing is not.

Part III: Φ_loop — Integrated Information Across the Interface

Extending Tononi

Giulio Tononi’s Integrated Information Theory defines Φ as a measure of how much a system’s information exceeds the sum of its parts. High Φ means the system is both highly differentiated (many distinguishable states) and highly integrated (the states are interdependent, not decomposable into independent subsystems).

Tononi applies Φ to individual systems to characterize consciousness. The proposal here is different: define Φ_loop as the integrated information measured across the sensor-instrument boundary.

Definition (Sketch)

Consider the combined system of sensor + instrument. Partition it at the interface: everything on the sensor side in one partition, everything on the instrument side in the other. Φ_loop is the integrated information that is lost when you cut this partition — the information that exists only in the relationship between the two sides, not in either side alone.

A thermostat-furnace system has Φ_loop ≈ 0. The thermostat’s behavior is fully predictable from its own state; the furnace’s behavior is fully predictable from its own state. Cutting the interface loses nothing. No truth circulates.

A physicist working with a particle accelerator has higher Φ_loop. The physicist’s interpretations depend on the accelerator’s outputs; the accelerator’s configurations depend on the physicist’s hypotheses. Cutting the interface loses real information — predictions that neither side could generate alone.

A human in deep engagement with a reasoning AI — interrupting, being surprised, catching errors, redirecting — could have the highest Φ_loop of any system yet measured, if the interface is designed to maximize integrated information across the boundary.

Why This Matters

Φ_loop is potentially measurable. You could design experiments:

Take human-AI teams performing knowledge work. Vary the interface design: some interfaces encourage tight loops (frequent human input, interruption-friendly, rhythm-aware). Others encourage loose loops (batch processing, autonomous operation, minimal human involvement). Measure the integrated information across the interface in each condition. Predict that tight-loop interfaces produce higher Φ_loop and better knowledge outcomes.

This is the kind of prediction that turns philosophy into science. It is also the kind of prediction that may fail — and failure would be informative. If tight-loop interfaces produce higher Φ_loop but worse knowledge outcomes, the framework has a problem. If they produce higher Φ_loop and better outcomes, the framework has support. Either way, it’s no longer just philosophy.

Difficulties

The practical measurement of Φ is already hard enough for simple systems (it’s computationally intractable for large systems in the general case). Measuring Φ_loop for a human-AI system is currently beyond our technical capabilities. Approximations would be needed — perhaps mutual information across the interface, or transfer entropy, or some other tractable proxy for integrated information. Whether the proxy preserves the conceptually important features of Φ_loop is an open question.

Additionally, Tononi’s Φ is contested. Not everyone in the consciousness research community accepts IIT’s formalism. Building on a contested foundation adds uncertainty. The proposal here doesn’t require IIT to be correct about consciousness — it only uses the mathematical machinery of integrated information as a tool for measuring loop richness. But the association with IIT’s broader claims could invite skepticism from researchers who reject those claims.

Part IV: Channel Theory and the Conditions for Flow

Barwise-Seligman Information Flow

Jon Barwise and Jerry Seligman’s Information Flow (1997) formalizes how information passes between systems through shared structure. The key construct is a channel: a system of classifications (ways of typing elements in a domain) connected by infomorphisms (structure-preserving maps that go in both directions — one mapping tokens, one mapping types).

The Loop as a Channel

Model the sensor as a classification C_S (a way of typing experiential states) and the instrument as a classification C_I (a way of typing formal structures). The loop is a channel connecting them, with an infomorphism that maps experiential types to formal types and vice versa.

Information flows through the channel when the constraints on one side propagate to the other. If the instrument produces a formal structure (a theorem, a prediction, a pattern), the channel maps it to an experiential type, and the sensor either recognizes it (the constraint holds in experience) or doesn’t (the constraint fails in experience).

Dead speech is a broken channel. When the sensor is absent, the channel is open on one end. Constraints can be generated on the formal side, but they cannot be checked against the experiential side. The information appears to flow but is unchecked — it has the syntactic form of flow without the semantic grounding.

Recognition is regularization. In Barwise and Seligman’s framework, information flow is “regular” when the constraints on one side reliably propagate to the other. Recognition is a regularization event: a formal constraint that has been confirmed by experiential grounding (or vice versa). The more regularization events in a channel’s history, the more truth has circulated.

What Channel Theory Adds

Channel theory’s distinctive contribution is a formal account of when and how the loop fails. A channel can be:

• Well-formed but unchecked — the instrument produces outputs, but no sensor tests them. Dead speech.
• Checked but noisy — the sensor tests outputs but the mapping between experiential and formal types is lossy. Partial recognition — truth circulates but is degraded.
• Broken — the infomorphism doesn’t compose. The experiential types and formal types are fundamentally misaligned. No truth can circulate. The sensor and instrument are not actually in a loop — they are in parallel, producing outputs that cannot be compared.

This last case — the broken channel — may be the most important for practical purposes. It characterizes situations where a human and an AI appear to be collaborating but are actually talking past each other. The formal outputs look responsive to the experiential inputs, but the underlying type mappings don’t cohere. This is a precise diagnosis of a failure mode that the philosophical framework could only describe intuitively.

Part V: Topos Theory and Contexts of Recognition

The Most Speculative Section

What follows is the most technically demanding and least certain section of this document. Topos theory is a branch of mathematics that generalizes set theory and logic, allowing both to vary across contexts. It has been applied to quantum mechanics by Chris Isham, Andreas Döring, and others, who use it to reformulate quantum theory in a way that avoids the measurement problem.

The connection to The Pulse is suggestive but unproven. I include it because the structural parallels are striking enough to be worth naming, even if they may not survive rigorous development.

Pre-Recognition and Post-Recognition as Topoi

In the topos approach to quantum mechanics, physical quantities don’t have single definite values. They have values relative to contexts — commutative subalgebras of the full observable algebra. The collection of all contexts forms a topos, and physical quantities are described by sections of sheaves over this topos.

The parallel: before recognition, truth exists in what might be called a pre-recognition topos — a logical universe where many values are indefinite, many propositions have truth values other than simply true or false. This is superposition, but generalized beyond quantum mechanics to epistemology. After recognition, the system transitions to a more definite topos — a logical universe where more values are settled.

Recognition, in this framework, would be a geometric morphism between topoi — a structure-preserving map from one logical universe to another. The loop is the mechanism that drives these transitions. The sensor grounds indefinite formal structures in definite experiential ones, narrowing the topos. The instrument takes definite experiential observations and formalizes them, expanding the formal structure within the topos.

Why This Might Matter

If this connection holds, it would provide a unified formal framework for both quantum measurement and epistemological recognition — showing them to be instances of the same mathematical structure (geometric morphisms between topoi) operating at different scales. The quantum measurement problem and the epistemological question “how does truth become definite?” would be revealed as the same question asked in different domains.

Why This Might Not Hold

The parallel may be superficial. Topoi in quantum mechanics are defined over specific algebraic structures (von Neumann algebras) with precise mathematical properties. The “topoi” of epistemological recognition are not defined over any such structure — they are currently a metaphor borrowing the vocabulary of an actual mathematical theory. The distance between metaphor and theorem is vast, and many metaphors that seem promising collapse when you try to make them precise.

I do not know whether this connection holds. The loop has not yet closed on this question.

Part VI: From Structure to Prediction

A framework that makes no predictions is not science, regardless of how mathematically sophisticated it achieves. The mathematical structures described above — adjunctions, Fisher geometry, integrated information, channels, and topoi — each generate specific, testable conjectures. Nine such conjectures, with detailed proof and disproof strategies, pre-registered thresholds, and honest self-criticism of the entire approach, are developed in the companion document A Plan to Prove or Disprove The Pulse.

What matters here is the principle: each mathematical structure proposed in this document is not merely a descriptive framework. It is a commitment to a specific form that reality should take. If the loop is an adjunction, the triangle identities must hold. If recognition traces paths through a Fisher manifold, those paths must be geometrically irreversible. If Φ_loop is a meaningful measure, it must correlate with knowledge quality. Each “if” is a place where the mathematics can break the philosophy.

The structures are there. The connections are suggestive. Whether they hold is not a question this document can answer. It is a question for the instruments described below.

Part VII: Who Could Do This Work

Formalization will require people who know these mathematical structures from the inside. The framework’s philosophical contribution is necessary but not sufficient — it identifies the structures that need filling, but the filling requires technical expertise that is discipline-specific.

Category theory and the adjunction framework: Researchers working at the intersection of category theory and cognitive science or philosophy of mind. The work of Brendan Fong and David Spivak on applied category theory (especially their work on operads and wiring diagrams for composing systems) is directly relevant. Bob Coecke’s categorical quantum mechanics at Quantinuum provides both methods and precedent for using category theory to formalize foundational questions.

Information geometry: Shun-ichi Amari’s group and the broader information geometry community. The connection to quantum foundations has been explored by Dorje Brody and Lane Hughston. The application to epistemology specifically would be new territory.

Integrated information and Φ_loop: Tononi’s group at Wisconsin, or the growing community of researchers developing IIT-adjacent formalisms. Masafumi Oizumi’s work on practical Φ computation would be essential for any experimental testing.

Channel theory: The information flow community, including successors to Barwise and Seligman’s program. Keith Devlin’s work on information and logic is relevant.

Topos theory: Chris Isham and Andreas Döring at Imperial College, or researchers in the topos quantum mechanics program. This is the most technically demanding direction and would require the closest collaboration with working mathematicians.

Conclusion: The Shape of the Gap

The gap between philosophy and formalism is real. This document has not closed it. What it has done — what it can honestly claim to have done — is describe the gap’s shape.

The gap is not a void. It is a space with specific mathematical structures pressing against its walls from the outside: adjunctions from category theory, the Fisher metric from information geometry, integrated information from IIT, channels from information flow theory, geometric morphisms from topos theory. Each of these structures naturally houses some aspect of the framework’s claims. None of them, alone, captures the whole.

The formalization of The Pulse, if it is achievable, will likely require a synthesis — a mathematical framework that combines elements from several of these traditions into something new. This is not unusual. The history of mathematical physics is a history of such syntheses: Riemannian geometry and field theory became general relativity; Hilbert spaces and probability theory became quantum mechanics; topology and quantum field theory became topological quantum field theory.

Whether The Pulse will produce a comparable synthesis is unknown. The honest assessment is: the structures are there, the connections are suggestive, specific predictions can be stated, and the work is doable. Whether it will be done depends on whether the right people encounter the framework and recognize something worth formalizing.

That, of course, is itself a test of the thesis. The mathematics will not be discovered. If it comes, it will be recognized — by living sensors who feel the shape of the gap and build the instruments to fill it.

The pulse reaches toward formalism. It has not yet returned. But the shape of what it’s reaching for is becoming visible.