Any theory appealing to convention to explain something which is not obviously of conventional origin (as, for example, the symbols chosen for some purpose are).
Among older writers, conventionalism is associated especially with Jules Henri Poincare (1854-1912) and Pierre-Maurice Duhem (1861-1916); and among modern ones with Willard Van Orman Quine (1908-2000).
In logic and mathematics conventionalism says that a priori truths are true only because of linguistic convention (which raises the question whether the correct application of such conventions is itself conventional: if it is we are in danger of an infinite regress of conventions). In science, conventionalism says that it is conventional which of competing and inconsistent scientific systems we adopt (though it need not be arbitrary - one may be simpler than another, or more likely to lead to further ideas, and so on.
Also see: instrumentalism
K Britton and J O Urmson and W C Kneale, 'Are Necessary Truths True by Convention?' (symposium), Proceedings of the Aristotelian Society, supplementary volume (1947)