Comparing Abstraction in Humans and Large Language Models Using Multimodal Serial Reproduction
This work addresses the problem of understanding abstraction differences between humans and AI models for cognitive science and AI researchers, but it is incremental as it builds on existing serial reproduction paradigms.
The study investigated how language affects abstraction formation by comparing humans and GPT-4 in a multimodal serial reproduction task, finding that adding language had a larger impact on human reproductions than on GPT-4's, suggesting more dissociable visual and linguistic representations in humans.
Humans extract useful abstractions of the world from noisy sensory data. Serial reproduction allows us to study how people construe the world through a paradigm similar to the game of telephone, where one person observes a stimulus and reproduces it for the next to form a chain of reproductions. Past serial reproduction experiments typically employ a single sensory modality, but humans often communicate abstractions of the world to each other through language. To investigate the effect language on the formation of abstractions, we implement a novel multimodal serial reproduction framework by asking people who receive a visual stimulus to reproduce it in a linguistic format, and vice versa. We ran unimodal and multimodal chains with both humans and GPT-4 and find that adding language as a modality has a larger effect on human reproductions than GPT-4's. This suggests human visual and linguistic representations are more dissociable than those of GPT-4.