Abstract
Dependent types are increasingly used in functional programming languages. The surface syntax of dependent types, as seen by a programmer, is elaborated by a compiler into an internal, type-theoretic representation. In order to perform this step, the compiler needs to infer a nontrivial amount of information to successfully type-check the internal representation. This process—type refinement—is complex, implementation dependent, and very few formal developments currently exist. We discuss a novel and simpler formalisation of type refinement in first order type theory with dependent types. We propose a translation of type-refinement problems to Horn-Clause logic with explicit proof-terms, using proof-relevant resolution as the type inference mechanism.