The Formal System as a Directed Graph
A formal axiomatic system defines a directed graph:
- Vertices: all well-formed formulas constructible from the system's symbols
- Edges: inference steps — one formula follows from others by a rule of inference
- Axioms: distinguished source vertices with no incoming edges
- Theorems: all vertices reachable from the axiom set
A proof of theorem T: a directed path from the axiom set to T. The proof is a sequence of vertices A₁, A₂, ..., Aₙ = T where each step follows by an inference rule.
Two fundamental properties of a formal system, expressed geometrically:
Consistency: no formula F and its negation ¬F are both reachable from the axioms. Geometrically: the theorem vertex F and the theorem vertex ¬F are not both reachable. No 'explosion' path exists.
Completeness: every formula F or ¬F is a theorem (reachable). Geometrically: the graph is strongly connected in the sense that for every vertex F, at least one of F or ¬F has a path from the axioms.
Gödel Incompleteness as a Geometric Property
Kurt Gödel proved in 1931 that any consistent formal system powerful enough to express basic arithmetic is incomplete: there exist formulas G such that neither G nor ¬G is provable.
Geometrically: in any sufficiently rich consistent formal system, there are vertices in the formula graph that are not reachable from the axioms — neither the vertex G nor the vertex ¬G has a path from the axiom set.
Gödel's construction: he encoded a formula G that says, in effect, 'I am not provable.' If G were provable, the system would be inconsistent (a true statement says it is not provable). If ¬G were provable, the system would be inconsistent (G would be false but the system proves it). So neither G nor ¬G is provable — G is an unreachable vertex in a consistent system.
The incompleteness is not a defect of the axioms chosen: Gödel showed it is a structural property of any consistent system expressive enough to handle arithmetic. The unreachable vertices cannot be removed by adding more axioms without generating new ones.
Mathematical Objects as Points in a Space
The Platonic view of mathematics can be formalized geometrically: mathematical objects inhabit an abstract space whose points are the objects themselves and whose structure is given by the relations among them.
Consider the natural numbers ℕ = {0, 1, 2, 3, ...}. The divisibility relation defines a partial order: m divides n iff m | n. This partial order defines a geometry — the Hasse diagram of the divisibility lattice.
Every prime number sits at a minimal position above 1. Every composite number sits above its prime factors. The structure of this space encodes all of number theory.
What makes this Platonic: the structure exists whether or not any mind studies it. The fact that 7 is prime — that 7 has no divisors between 1 and 7 — is a fact about the position of 7 in the divisibility lattice, independent of notation, culture, or civilization.
Any civilization that investigates counting and divisibility will discover the same structure. The geometry of the number system is universal.
Navigating the Divisibility Lattice
In the divisibility lattice, the least common multiple (lcm) of two numbers corresponds to their join (lowest upper bound) and the greatest common divisor (gcd) corresponds to their meet (greatest lower bound).
The gcd can be computed via the Euclidean algorithm: gcd(a, b) = gcd(b, a mod b), terminating when b = 0.
What Abstraction Strips Away
Geometric abstraction: projecting a high-dimensional object onto a lower-dimensional subspace. The projection loses information (coordinates not in the subspace) but retains the subspace structure perfectly.
Mathematical abstraction works the same way. A group is a set with one binary operation satisfying four axioms. Abstracting to group structure strips away: the specific elements, the specific operation's computational rule, any additional structure (order, metric, topology). What remains: the four-axiom skeleton.
The payoff: every theorem about groups applies to ALL groups — integers under addition, rotations under composition, permutations under composition, symmetries of a molecule, Galois groups of polynomial equations — simultaneously. The abstract theorem is provable once; its instances are free.
This is why pure mathematicians resist adding domain-specific assumptions: each added assumption restricts the theorem's applicability. A theorem about fields (adding multiplicative inverse) applies to fewer structures than a theorem about rings (no multiplicative inverse required).
The Precision-Generality Tradeoff
There is a tradeoff: more abstract theorems apply more broadly but say less about specific instances. A theorem about vector spaces over a field says less about ℝ^n than a theorem specifically about ℝ^n (where distance and angle are defined).
Hamming's implied rule: abstract as far as possible while retaining the properties you need. Abstract too far and your theorems become vacuously general ('any set with any operation satisfies...'). Abstract too little and your theorems fail to transfer to new applications.