Algebraic models of conceptual metaphor: contributions to the understanding of mathematics learning processes

Abstract

This thesis studies a human cognitive phenomenon called Conceptual Metaphor in the context of mathematics learning and reasoning. Metaphor enables the understanding of an abstract concept called target, e.g. numbers, in terms of a more concrete concept called source, e.g. piles of can-dies. Often, inferences from the source are carried to the target and applied there yielding some conclusions about the target. This is known as reasoning by analogy. Empirical evidence indicates that metaphor enhances learning. Converging evidence is pro-vided by working scientists who report the use of analogies while developing their theories. On the other hand, some people advise against its usage in education. They argue that politicians and communicators often lead people into erroneous conclusions by using metaphor, and then, analo-gies undermine objective reasoning. This discussion highlights the need for research to shed light into the learning mechanics underlying metaphor in order to understand its scope and limitations. This work presents a formal model of metaphor which can be used as a framework to study learning by analogy. Since the model is abstract, we use Chapter 1 to make ideas more concrete: we use our formalism for analize deeply a well known example. Along these lines, Chapter 5 presents formalizations of other metaphors frequently encountered in mathematics teaching. The model is built in Chapter 4 where the source and the target of a metaphor are formalized by a key concept named domain. Some results of this chapter are accompanied by cognitive in-terpretations, as for example, Theorems 40, 41, 42, and Proposition 26 can be seen as descriptions of how an analogy carries reasonings from its source to its target. Also, Theorems 30 and 31 sug-gest models for the process of learning by analogy. Finally, Chapter 4 presents some theoretical constructions such as products and coproducts of domains. Our metaphor model relates two domains, each one defined as a mixture of language and semantics. Most results of Chapter 4 need the premise that the two involved languages are compatible . Mathematically, they need a map able to preserve the structure determined by a syntactical operation called substitution. This compatibility notion is characterized for the case of language terms in Chapter 2 by applying unification theory and graph theory. And in Chapter 3, this compatibilitynotion is characterized for the case of the language formulas by adapting the methods of Chapter 2. Finally, one Appendix (Relational Spaces) presents another approach to study metaphor. There, domains are defined with semantics only, leaving language aside. Most of the results emphasized above are lost or at least weakened suggesting that the abstract information provided by symbols and the recursion provided by the grammar of the language are necessary to mimic metaphor s behavior. This observation, together with other results of this thesis, might point to a relation between the recursion property of human languages1 and the ability of learning by analogy. 1 The linguist Noam Chomsky claims that recursion is the only human component of the faculty of language [49].