Formal Sciences: Mathematics, Logic, and Computer Science

The formal sciences occupy a distinctive position in the landscape of human knowledge — they generate truths not by observing the physical world but by reasoning within precisely defined systems of rules and symbols. Mathematics, logic, and computer science share this character, and together they supply the structural backbone that makes empirical science computable, testable, and communicable. What follows examines how these disciplines are defined, how they actually produce knowledge, where they appear in practice, and how researchers decide which formal tools are appropriate for a given problem.

Definition and scope

A formal science is a discipline whose objects of study are abstract structures — numbers, sets, propositions, algorithms — and whose methods of establishing truth are proof and logical deduction rather than experiment and observation. The distinction matters enormously. When a biologist claims a drug reduces tumor volume by 40%, that claim depends on measurement and could in principle be falsified by new data. When a mathematician claims that there are infinitely many prime numbers, the proof — Euclid's, established roughly 300 BCE — is not subject to revision by experiment. It is simply true within the axioms.

The three pillars of formal science each have a distinct domain:

1. Mathematics — the study of quantity, structure, space, and change. It branches into pure mathematics (pursued for internal coherence and beauty) and applied mathematics (pursued for modeling real-world phenomena). The American Mathematical Society (AMS) classifies active mathematical research into more than 60 subdisciplines under its Mathematics Subject Classification system.
2. Logic — the study of valid inference. Formal logic, as codified in the 19th and early 20th centuries by figures including Gottlob Frege and Bertrand Russell, provides the rules that govern what follows from what. It is the grammar of all formal reasoning.
3. Computer science — the study of computation: what can be computed, how efficiently, and through what structures. The field inherits its theoretical foundations from Alan Turing's 1936 paper introducing the Turing machine, a formal model that still defines the limits of what any algorithm can accomplish (Association for Computing Machinery, ACM Digital Library).

The boundary between formal and natural science is not always clean — computational biology and mathematical physics operate in both worlds simultaneously — but the defining feature remains: formal sciences validate claims through proof, not experiment.

How it works

Formal science advances through a specific epistemic machinery. A mathematician or logician begins with axioms — statements accepted without proof as the foundation of a system. From these, theorems are derived using rules of inference, producing results that are necessarily true within the system. This is deductive reasoning in its purest form, the same structure Euclid used in the Elements and that Whitehead and Russell attempted to extend to all of mathematics in Principia Mathematica (1910–1913).

Computer science adds a layer: it asks not just whether something is provable but whether it is computable — and if so, at what cost. Computational complexity theory measures cost in terms of time and space as a function of input size, captured in classes like P (problems solvable in polynomial time) and NP (problems whose solutions can be verified in polynomial time). Whether P equals NP is one of the seven Millennium Prize Problems identified by the Clay Mathematics Institute (Clay Mathematics Institute), each carrying a $1,000,000 prize for solution.

The broader scientific method explained on this network applies differently here — formal sciences do not generate hypotheses to be falsified by data. They generate conjectures to be proven or disproven by argument. A conjecture that resists proof for decades (Fermat's Last Theorem, unproven for 358 years until Andrew Wiles's 1995 proof) is not "probably true" — it is simply unresolved.

Common scenarios

Formal sciences appear wherever precision and structure are non-negotiable. Some of the most consequential applications:

• Cryptography — the security of internet transactions depends on number-theoretic results, particularly the computational difficulty of factoring large integers. The RSA algorithm, developed in 1977, rests entirely on mathematical structure, not physical experimentation.
• Statistical modeling — the mathematics underlying regression, hypothesis testing, and Bayesian inference are prerequisites for any quantitative research. Researchers working through statistical analysis in research are applying formal mathematical tools whether or not they think of it that way.
• Algorithm design — sorting 1 billion records efficiently is a formal problem with measurable answers. The difference between an O(n log n) sort and an O(n²) sort is not academic at that scale; it translates to hours versus milliseconds.
• Proof assistants — software systems like Lean and Coq allow mathematicians to encode proofs in a formal language that a computer can verify. The 2023 formalization of the proof of the Liquid Tensor Experiment conjecture in Lean demonstrated that industrial-scale mathematics can be machine-checked (Lean Theorem Prover community, leanprover.github.io).

Decision boundaries

Choosing among formal tools requires clarity about what kind of question is being asked. A useful frame:

• Proof vs. approximation — some problems demand exact answers (cryptographic keys must be correct, not close). Others tolerate approximation, where heuristic algorithms outperform exact methods in practice.
• Discrete vs. continuous mathematics — computer science and combinatorics operate largely in discrete domains (graphs, integers, binary strings). Physics and engineering lean on continuous mathematics (calculus, differential equations). Many modern problems — machine learning, for instance — sit at the intersection.
• Decidable vs. undecidable — Gödel's incompleteness theorems (1931) proved that any sufficiently powerful formal system contains true statements that cannot be proven within that system. Turing extended this to show that certain computational problems, like the halting problem, have no algorithmic solution at all. These are hard limits, not engineering challenges waiting for a faster processor.

The national science authority home provides broader context for how formal science sits within the wider ecosystem of scientific disciplines and research infrastructure. For researchers navigating where formal methods intersect with computational and data-driven research, the distinctions above — provable versus approximable, decidable versus undecidable — are the ones that actually determine what methods are available.