Fleeting Connections: An Inquiry into Analogical Reasoning

Addressing issues often requires identifying new connections between concepts or events that initially seemed unrelated. Innovative solutions of this kind are based on analogical reasoning, which is a relationship-centered thinking process and plays a crucial role in human problem-solving and decision-making. The challenges in analogical reasoning research span from understanding the neural mechanisms involved to developing computational models that can mimic human-like reasoning processes. Bridging the gap between seemingly unrelated domains often requires identifying and representing the underlying structural similarities. In this analysis, we applied a similar procedure as described by Green et al. 2010. However, we gave the analogies of the original experiment to the Large Language Model (LLM) Llama2 instead of humans. For this reason, while the authors also measured and evaluated human reaction times in relation to the semantic distance between words, we decided to evaluate the model only on response accuracy with the purpose of comparing Llama2's performance with that achieved by humans.

The analysis was conducted in Pytho