Structured Derivations as a Unified Proof Style for Teaching Mathematics

Structured derivations were introduced by Back and von Wright as an extension of the calculational proof style originally proposed by E.W. Dijkstra and his colleagues. Structured derivations added nested subderivations and inherited assumptions to the original calculational style. This paper introduces a further extension of the structured derivation format, and gives a precise syntax and semantics for the extended proof style. The extension provides a unification of the tree main proof styles used in mathematics today: Hilbert-style forward chaining proofs, Gentzen-style backward chaining proofs and algebraic derivations and calculations (in particular, Dijkstra’s calculational proof style). Each of these proof styles can be directly modelled as an extended structured derivation. Even more importantly, the three proof styles can be freely intermixed in a single structured derivation, allowing different proof styles to be used in different parts of the derivation, each time choosing the proof style that is most suitable for the (sub)problem at hand. We describe here (extended) structured derivations, feature by feature, and illustrate each feature with examples. We show how to model the three main proof styles as structured derivations. We give an exact syntax for this proof style and define the semantics of structured derivations. The latter is done by showing how each structured derivation can be automatically translated into an equivalent Gentzen-style derivation. Structured derivations have been primarily developed for teaching mathematics on secondary and tertiary education level. The syntax of structured derivations determines the general structure of the proof, but does not impose any restrictions on how the basic constructs of the underlying mathematical domain are expressed. Hence, the style can be used for any kind of mathematical proofs, calculations, derivations, and general problem solving found in mathematics education at these levels. The exact syntax makes it easy to provide computer support for structured derivations.