Computational metaphor identification to foster critical thinking and creativity

Metaphor, the partial framing of a target concept in terms of a source concept, permeates human thought and action. Metaphors often manifest themselves as linguistic patterns in which language associated with a source concept is used to describe a target concept. Any given metaphor highlights some aspects of a concept while simultaneously downplaying others. Novel metaphors can provide creative reframings of familiar concepts by highlighting those aspects hidden by more common metaphors. However, due to their ubiquity, conceptual metaphors can be difficult to examine critically, if we can even notice them in the first place.;To address such difficulties, this dissertation develops computational metaphor identification (CMI), which identifies potential conceptual metaphors in written text. CMI maps selectional preferences of relatively frequent nouns in a source corpus to those in a target corpus. Such mappings indicate potential metaphors from concepts in the source corpus to those in the target. CMI can be used to draw attention to potential conceptual metaphors that might otherwise go unnoticed, making those metaphors available for critical and creative examination. For example, what might a given metaphor highlight, what might it hide, and what alternative metaphors might frame the situation differently? In this way, CMI is designed not as a type of computational reasoning, but as a means of facilitating human reasoning.;To evaluate its capacity to foster critical thinking and creativity, computational metaphor identification was incorporated into an educational module about cell biology, which was used to perform an experimental study in a 7th grade classroom. Students' answers to written questions about the cell were analyzed using CMI, and potential metaphors were presented back to students. The results demonstrate that the use of CMI effectively fostered both critical thinking about metaphors and creative generation of alternative metaphors. These results also speak to the varying roles of surface and structural similarity in metaphorical reasoning, as well as the relationship between noticing similarities and noticing differences when thinking about metaphors. This evaluation not only demonstrates CMI's usefulness in educational contexts, but it also carries broader implications for exploring the relationship between computation and human thought.