>> research directions <<

I study how neural networks represent and reason about the world, drawing on ideas from cognitive science. Current threads include:

• Symbols in neural networks. How do abstract concepts // symbolic structures emerge in DNNs? What mechanisms allow systems to move from continuous sensory input → discrete representations? Does a lack of symbolic structure explain persistent limitations of these models?
• Thinking across modalities. How do different modalities (e.g. vision, language) afford distinct kinds of reasoning? Is natural language really the best general-purpose medium for thought?
• Cognitively-inspired interpretability. How can we repurpose methodologies & frameworks from cognitive science & computational neuroscience to understand the internal workings of modern models? Conversely, what can probing them reveal about human intelligence?
• Representational alignment & universality. To what extent do intelligent systems converge on a shared representational space? What are the key representational differences between humans and machines, and how can we bridge them? How can we design benchmarks and evaluation protocols that enable meaningful human–machine comparisons?

:: contact ::

I’m always excited to collaborate or exchange ideas. If you'd like to talk about my work, your work, or anything really, please reach out!