Philosophy of Computing

What it asks

Computation is defined mathematically and performed physically, and the relation between those is not obvious.

Say what makes a physical system compute something. If the answer is that its states can be mapped onto the steps of a computation, then almost anything computes almost anything, since mappings are cheap. That result is easy to prove and clearly wrong, and blocking it has proved surprisingly hard.

The second question concerns what a program IS. It is written like a text, copyrighted like a text, and it also does things, like a machine. Neither category fits, which is why software has never sat comfortably in law and why arguments about whether it can be proved correct became philosophical arguments.

The recurring problems

Implementation. What makes this physical process an instance of that computation? Any answer must exclude trivial mappings without excluding real computers, and the obvious criteria do not.

The status of proof. In the 1970s and 1980s a serious dispute broke out over whether programs can be proved correct. The objection was that a proof concerns an abstract specification while the program runs on physical hardware in a physical world, so the proof cannot cover what actually matters. It was fought bitterly, it was never resolved, and it comes back with every safety-critical system.

Abstraction. Computing works by layering abstractions, each treating the one below as reliable. When a lower layer leaks, the abstraction fails in ways that are invisible from above. That structure has no real parallel in other engineering.

Whether the mind computes. Everything above bears directly on whether calling the brain a computer says anything, since the answer depends on what computing is.

How it developed

The mathematical foundations came first. Turing's analysis in 1936 of what a person following rules could do, and the thesis that this captures the effectively calculable, defined the subject before any machine existed.

As computing became an industry, its philosophy grew from practical disputes rather than from academic philosophy. The 1979 attack on program verification came from working computer scientists, and the sharpest reply in 1988 came from a philosopher.

The ontology of software developed in parallel, driven partly by law: courts had to decide whether a program was a text, a machine, an invention or a method, and the categories genuinely did not fit.

The field is now recognisably its own, with work on the nature of computation, on abstraction, on the epistemology of simulation, and on what it means to trust a system nobody can inspect.

The positions within it

Semantic and mechanistic accounts of computation divide over whether representation is essential to it. Pancomputationalism holds that everything computes, and is usually treated as a reductio. Formalist and empiricist views divide over whether programs are mathematical objects or experimental ones. All are in the Schools section.

Where it stands now

Simulation has become the central topic. Large parts of science now run on models nobody can survey in full, which raises a genuine epistemological question about what a simulation result is evidence for.

Machine learning has sharpened it. Systems whose behaviour is not specified but trained cannot be verified in the old sense at all, and the questions the verification dispute raised forty years ago are now being asked about systems where nobody even has a specification to prove things against.