Human Mathematics: Cognitive and Structural Insights

• Human Mathematics is a field that examines the subset of formal mathematics shaped by human cognitive, linguistic, and cultural constraints.
• It utilizes models like resource-bounded derivability and compressed dependency graphs to explain the efficiency and modularity of human mathematical reasoning.
• Research integrates neurobiological findings and structural frameworks to highlight the creative, narrative, and practical dimensions in mathematical practice.

Human Mathematics (HM) designates the distinctive subset of mathematical activity, reasoning, and structure that is realizable, valued, and transmitted by human minds. The concept encompasses cognitive, linguistic, organizational, and aesthetic criteria that sharply delimit HM from the totality of formal mathematics (FM)—the full deductive closure of a given axiom system—and highlight fundamental constraints and capabilities rooted in human biology and culture. Recent literature analyzes HM from multiple perspectives: mathematics-as-compression and concept recombination, neurocognitive architecture, evolutionary and linguistic universals, formal structural and computational models, and the practical organization of mathematical knowledge and practice.

1. Definitional Boundaries and Foundational Models

Human Mathematics is formally construed as a strict subset HM ⊂ FM, where FM is the set of all theorems deducible in a formal system (e.g., ZFC) (Ruelle, 2022, Aksenov et al., 20 Mar 2026). Ruelle sharpens this by defining HM in terms of resource-bounded derivability:

$$HM(L, P) = \{\varphi \in Th(A) : |\text{statement of } \varphi| \leq L,\, \text{human-readable proof exists of length} \leq P\}$$

with parameters L (statement length) and P (proof length) reflecting cognitive and communicative constraints. The union over tolerable L, P reflects the practical reach of human mathematicians.

Aksenov et al. further characterizes HM as the region of FM that admits dramatic compression via definitions, theorems, and hierarchical concept nesting. In monoidal models, where FM is the saturation of all legal symbol strings (proofs) over a set of primitive generators, HM is the subset where macros (named definitions) allow exponential expressivity with only logarithmic-sized macro sets (in the free abelian monoid $A_n$), a property echoed by deep, compressed dependency graphs in large formal libraries such as Lean's MathLib (Aksenov et al., 20 Mar 2026).

2. Cognitive and Linguistic Foundations

The cognitive roots of HM derive from a small number of evolutionarily ancient operations: qualitative equivalence (categorization), quantitative ordering (comparison), and associative chaining. These primitive processes, formalized as equivalence relations, total orders, and Markovian association chains, underlie the construction of the natural numbers (via the Peano axioms) and counting algorithms in children (Tuncer, 2011). Human languages universally encode constants, variables, quantifiers, implication, negation, and conjunction/disjunction—Gilkey et al. term these "soft universals"—making language a sufficient vehicle for recursive mathematical reasoning, proof by induction, and model-theoretic constructions (0909.3591).

The formation and utilization of compressed, named modules—concepts, theorems, procedures—are driven by the severe working-memory limits (~7 chunks), slow serial/rate-limited computation, and the need for chunking and renaming. This shapes the historically varying but structurally similar landscape of HM: from geometric constructions to category-theoretic abstractions, all characterized by the progressive naming and organizing of concept-packages (Ruelle, 2022).

3. Neurobiological Realization

Mathematical reasoning recruits a distributed but specialized network in the brain—centered on the intraparietal sulcus (IPS; number sense, quantity, visuospatial memory), inferior temporal gyrus (visual recognition of numerals and objects), and dorsolateral prefrontal cortex (executive control, working memory, abstraction)—and minimally engages classical language areas (Davis et al., 2021). fMRI studies of expert mathematicians (Amalric & Dehaene) consistently report bilateral IPS and ITG activation during mathematical judgment, regardless of linguistic or sensory experience, indicating that HM is fundamentally non-linguistic at the core, though linguistic competence is necessary for proof transmission and education.

The formation of "repeatable mental units" (compression/chunking) leverages these brain networks, with pedagogical implications: visual-spatial tasks, manipulative representations, and minimization of external linguistic load scaffold the relevant neural systems.

4. Structural, Organizational, and Technological Aspects

Human Mathematics relies on the efficient integration of four active aspects—Inference (formal deduction and proof), Computation (symbolic/numeric evaluation), Tabulation (systematic collection/storage of objects and data), and Narration (human-centric exposition, pedagogy, and communication)—all coordinated via a central knowledge-organization layer (ontology, theory graphs, symbol indices) (Carette et al., 2019). This "tetrapod" architecture allows mathematicians and (potentially) machine partners to traverse proofs, compute examples, query large datasets, and assemble these elements into narratives.

Within this framework, Big Math projects, e.g., the classification of finite simple groups or the formalization of the Kepler conjecture, stress the “One-Brain Barrier.” Coordinated modularization, theory morphisms, and software architectures (e.g., MMT/OMDoc, Coq, Lean) are necessary to maintain human tractability within HM, preventing knowledge siloes and fostering reusability.

5. Information-Theoretic and Empirical Compression

Aksenov et al. provide a formal and empirical model of why only a vanishingly small, highly compressible island within FM becomes HM (Aksenov et al., 20 Mar 2026). In the free abelian monoid $A_n$, appropriately chosen, logarithmically sparse macro sets offer exponential expansion (number of accessible objects scales as $b^{s/(n(b-1))}$ with budget $s$), reflecting positional notation and the design of compact concept hierarchies. In the non-abelian case, linear expansion is the best achievable without exponentially many macros. Empirical evaluation of MathLib shows that unwrapped formula/proof length grows exponentially with depth (layers of nesting), while wrapped length (tokenized definitions) remains nearly constant—quantitatively consistent with the $A_n$ model and inconsistent with $F_n$.

PageRank-style metrics on dependency graphs, factoring in reductive and deductive compression, quantify mathematical "interest" and provide guidance for automated reasoning toward compressible, human-valued regions.

6. Evolutionary and Causal-Enumerative Generalization

Evolutionary perspectives identify the cognitome—a neural hypernetwork of cognitive groups (COGs) defined by maximally specific causal rules—as the organizing substrate for natural intelligence (Vityaev, 7 Dec 2025). The universal operation underpinning HM is closure under maximally specific probabilistic causal relationships: the brain enumerates all statistical regularities (causal rules), prunes to maximal specificity, and inductively closes over the resulting ruleset. This generates classification, functional behavior, prototype formation, causal modeling, and even consciousness (as the maximally integrated closure)—providing a formal, set-theoretic foundation for HM that generalizes across individual and species differences.

7. Humanization, Aesthetics, and Multicultural Perspectives

Human Mathematics is not reducible to formalism or deduction. Inoue stresses the role of individual creativity, aesthetic criteria (naturalness, elegance, "story-like" proofs), multicultural influences, and the personal "style" of mathematical thinking (Inoué, 2023). Even within a classical logic framework, he urges the deliberate inclusion of epistemic, constructive, modal, and alternative logical approaches to diversify HM and broaden its expressive range.

The ultimate aim is a mathematics that is not only correct, but deeply human—expressive, narratively rich, resonant with human intuition, and open to cultural variation. The future of HM lies in the systematic development of such alternative frameworks, deeper investigations into foundational constants ("the secret of 1"), and integration of higher-order structures, e.g., possible-world semantics, in model theory and beyond.

---

References:

(Tuncer, 2011, 0909.3591, Ruelle, 2022, Davis et al., 2021, Aksenov et al., 20 Mar 2026, Carette et al., 2019, Vityaev, 7 Dec 2025, Inoué, 2023)