undecidability

Undecidability is a concept in computability theory describing decision problems for which no algorithm can determine the correct answer for every possible input. A decision problem asks a yes-or-no question about a finite input; a problem is decidable if there exists a Turing machine (or equivalent procedure) that halts on all inputs and always answers correctly. If no such algorithm exists, the problem is undecidable.

Undecidable problems are closely related to uncomputability, but the terms are used in slightly different senses. A problem may be uncomputable in the sense that no algorithm can compute a function that solves it for all inputs; many undecidable problems are also uncomputable, though the distinction is subtle and technical.

A classic example is the Halting Problem: given a program and its input, determine whether the program will eventually halt. There is a procedure that can confirm halting when it occurs, but no general algorithm can decide halting for all program-input pairs. Consequently, the Halting Problem is undecidable. Related undecidable questions include program equivalence (whether two programs compute the same function) and many problems about the behavior of programs and formal systems.

Theoretical results illuminate the scope of undecidability. The Entscheidungsproblem, the challenge of deciding the truth of first-order logic sentences, was shown undecidable by works of Church and Turing. Rice's theorem states that every nontrivial semantic property of the computed functions is undecidable. Other notable undecidable problems include the Post correspondence problem and the word problem for finitely presented groups.

Undecidability has practical implications: it sets fundamental limits on what can be automated in mathematics, verification, and software analysis. In response, researchers study decidable fragments, semi-decision procedures, and heuristic or approximate methods.