How do “artificial brains” like large language models process information—and what can cognitive science tell us about what is really going on inside them?
Ane works with both concept-based approaches (such as contextual embeddings) and mechanistic perspectives (looking at the “neural circuits” inside models) to understand how large language models represent, transform, and reproduce information.
By bringing together cognitive science, computer science, and science education, she investigates whether—and in what sense—these systems have something like “cognitive abilities,” and how this shapes the way we should use and trust them in society.
Questions for discussion:
- How can cognitive science help us understand what large language models “know” and how they think?
- How do different disciplines (like cognitive science, computer science, and education) need to work together to decide which questions we should ask these AI systems in the first place?