> results indicate that linguistic representations are neither utilized nor required for inductive or deductive logical reasoning.
It's often said that "Writing is thinking." At least verbal and written language is considered an important mode of thinking. If logical reasoning does not depend on linguistic representations, nor does it require the language brain network, it implies that natural language is a kind of byproduct, secondary to the deeper level of "thinking without words". I wonder how we can better understand, personally, what's happening at that level and consciously engage with it. It also raises the question, how this is relevant to large language models and their ability to simulate the thinking process. In a way they do not think with words directly either, but its "thoughts" emerge from the network of relationships between words. Still, it seems this research on the non-verbal nature of logical reasoning points to unexplored potential, how we can attempt to model the deeper dynamics "below" thinking.
Research summary is very good. But not seeing an abstract syntax for the distinct representational format:
Conclusions. Our results provide evidence against the hypothesis that natural language serves as the medium of abstract logical reasoning and suggest that such reasoning is underpinned by a distinct representational format.
Nevertheless, a good read that summarizes conventional learning and reasoning methods.
> results indicate that linguistic representations are neither utilized nor required for inductive or deductive logical reasoning.
It's often said that "Writing is thinking." At least verbal and written language is considered an important mode of thinking. If logical reasoning does not depend on linguistic representations, nor does it require the language brain network, it implies that natural language is a kind of byproduct, secondary to the deeper level of "thinking without words". I wonder how we can better understand, personally, what's happening at that level and consciously engage with it. It also raises the question, how this is relevant to large language models and their ability to simulate the thinking process. In a way they do not think with words directly either, but its "thoughts" emerge from the network of relationships between words. Still, it seems this research on the non-verbal nature of logical reasoning points to unexplored potential, how we can attempt to model the deeper dynamics "below" thinking.
Research summary is very good. But not seeing an abstract syntax for the distinct representational format:
Nevertheless, a good read that summarizes conventional learning and reasoning methods.