Slashing Metaphor with Occam’s Razor

We argue that, from the perspective of a language-endowed social intelligent agent, processing metaphor – and figurative language in general – is an epiphenomenon of an agent’s ability to learn lexical senses on the fly and to postpone or disregard analysis of parts of input without compromising its functioning. This work is a step toward developing an explanatory theory of why people often operate successfully without fully understanding language input and sometimes without even attempting to process parts of it. The basis for understanding novel metaphors is modeling the lifelong, “diachronic” functioning of an agent. This paper provides a descriptive account of this theory and discusses how several existing components of the OntoAgent environment are used or are being extended to support implementation of the theory in a proof-of-concept system.