Formal Sciences: Mathematics, Logic, and Computer Science

The formal sciences occupy a peculiar and foundational position in the scientific landscape — disciplines that produce rigorous, verifiable knowledge without relying on physical experiments or empirical observation. Mathematics, logic, and computer science share a methodology built on proof, abstraction, and deductive reasoning. Understanding where the formal sciences begin and end helps clarify how knowledge itself gets structured — and why disciplines from physics to economics depend on formal methods as their backbone.

Definition and scope

A formal science is a discipline that studies formal systems: abstract structures defined by explicit rules, symbols, and inference procedures. Unlike the natural sciences — which test hypotheses against physical reality — formal sciences derive their conclusions from within a system of axioms and definitions. That distinction matters enormously. A mathematical theorem isn't "probably true" pending further data; once proven, it holds without exception within its axiom set.

The three primary formal sciences are:

1. Mathematics — the study of quantity, structure, space, and change through abstract reasoning. Mathematics encompasses subfields including number theory, algebra, geometry, calculus, topology, and combinatorics.
2. Logic — the study of valid inference and formal argumentation. Mathematical logic, propositional calculus, and predicate logic establish the rules by which conclusions legally follow from premises.
3. Computer science — the study of computation, algorithms, data structures, and formal languages. Although it has a rich applied side, its theoretical core (complexity theory, automata theory, computability) is as abstract as any branch of mathematics.

Statistics is sometimes classified as a formal or semi-formal science depending on the context — its methods are formal, but its application is empirically oriented. The Stanford Encyclopedia of Philosophy treats the boundary between formal and empirical sciences as one of the more contested classification questions in philosophy of science.

The formal sciences sit at the structural foundation of the broader scientific enterprise described at The Science Authority — supplying the language, proof standards, and modeling tools that empirical sciences depend on.

How it works

Formal sciences operate through a specific epistemic cycle that looks nothing like laboratory science. It runs something like this:

• Axioms are accepted as starting assumptions — not proven, but chosen for coherence and utility (e.g., Euclid's five postulates, or the Zermelo-Fraenkel axioms for set theory).
• Definitions introduce precise concepts with no ambiguity tolerated.
• Theorems are derived from axioms and prior theorems through deductive proof.
• Proof verification — either by peer review or, increasingly, by formal proof assistants like Lean or Coq — confirms that no inferential step was illicit.

The power of this approach is cumulative and compounding. The Pythagorean theorem, proven roughly 2,500 years ago, has not been revised. It has been generalized, contextualized, and applied in ways its originators never imagined — but the original proof still stands. Compare that with the natural sciences, where even well-established theories get refined when new empirical evidence emerges. For a deeper look at how proof and methodology differ across science types, the page on how science works unpacks those structural contrasts.

Computer science adds a third validation mode: computation itself. An algorithm's correctness can sometimes be formally proven, and its performance measured empirically by running it. This hybrid character — formal in theory, empirical in implementation — is part of what makes computer science epistemologically interesting.

Common scenarios

Formal sciences show up, often invisibly, across nearly every field of inquiry:

• Cryptography depends on number theory — specifically the computational difficulty of factoring large integers. The RSA algorithm, used in most secure web communication, rests on a mathematical conjecture (integer factorization hardness) that has resisted disproof for over 40 years.
• Statistics and probability underpin clinical trial design, economic modeling, and machine learning — all formal constructs applied to real-world data.
• Logic in law and philosophy structures valid argumentation, contract interpretation, and legislative drafting. The concept of "if-then" conditions embedded in legal language maps directly onto propositional logic.
• Theoretical physics uses differential geometry and tensor calculus — branches of mathematics — to express general relativity. Einstein didn't invent that mathematics; he adopted a framework Bernhard Riemann had developed decades earlier for purely abstract reasons.
• Algorithm design in computer science produces formal guarantees: a sorting algorithm that runs in O(n log n) time will scale predictably, regardless of which machine runs it.

Decision boundaries

Knowing what the formal sciences can't do is as important as knowing what they can.

Formal vs. empirical sciences: Mathematics can prove that a particular encryption scheme is theoretically unbreakable under specific assumptions — but it cannot determine whether a given key was actually compromised in a real network intrusion. That requires forensic investigation. Formal tools model reality; they do not directly observe it.

Proof vs. evidence: In formal sciences, a single valid counterexample demolishes a conjecture entirely. In empirical science, a single anomalous result rarely overturns a well-supported theory — it prompts investigation. These are genuinely different epistemic standards, not one being more rigorous than the other.

Completeness limits: Kurt Gödel's incompleteness theorems (1931) established that any sufficiently powerful formal system contains true statements that cannot be proven within that system. This is not a flaw in mathematics — it is one of the most precise and astonishing results mathematics has ever produced about itself. The formal sciences are not omnipotent; they are, however, extraordinarily honest about their own limits.

Computability limits: Alan Turing's halting problem proof (1936) demonstrated that no algorithm can determine, for all possible programs, whether a given program will halt or run forever. Some problems are formally undecidable — not merely unsolved, but provably beyond algorithmic resolution.