Embedded Files
I am recruiting outstanding undergrads, MSc, PhD, and postdocs.
I am recruiting outstanding undergrads, MSc, PhD, and postdocs.
Tip: Be as specific as possible regarding what you are interested in!
Tip: Be as specific as possible regarding what you are interested in!
Topics:
Topics:
I'm fascinated by human concepts and what it means for a machine to have them. These are the compact, abstract units humans use to think, learn, and communicate. My lab works at the intersection of AI/NLP/LLMs, cognitive science, and human-AI interaction, and most projects fall into one of three buckets:
I'm fascinated by human concepts and what it means for a machine to have them. These are the compact, abstract units humans use to think, learn, and communicate. My lab works at the intersection of AI/NLP/LLMs, cognitive science, and human-AI interaction, and most projects fall into one of three buckets:
1) Training concept-aware models: Can we build LLMs that represent and reason over concepts rather than just tokens?
1) Training concept-aware models: Can we build LLMs that represent and reason over concepts rather than just tokens?
2) Concept interpretability: How can we tell what concepts models actually hold, how they are being used, and where they diverge from human ones?
2) Concept interpretability: How can we tell what concepts models actually hold, how they are being used, and where they diverge from human ones?
3) Using computational methods to understand humans: Models and human data can be a lens on cognition itself. I am equally excited about projects that turn this around and use AI to learn something new about us.
3) Using computational methods to understand humans: Models and human data can be a lens on cognition itself. I am equally excited about projects that turn this around and use AI to learn something new about us.
Mentoring style:
Mentoring style:
My default is to meet you where you are and adjust from there. Early on, I tend to be more hands-on, e.g., helping scope the problem, reading drafts closely, and sitting with you through the first few analyses. As you build momentum and judgment on a project, I pull back and shift into more of a sounding board: someone who asks hard questions and helps you see around corners, rather than someone directing the next step. The goal is to build independence.
My default is to meet you where you are and adjust from there. Early on, I tend to be more hands-on, e.g., helping scope the problem, reading drafts closely, and sitting with you through the first few analyses. As you build momentum and judgment on a project, I pull back and shift into more of a sounding board: someone who asks hard questions and helps you see around corners, rather than someone directing the next step. The goal is to build independence.
I do not believe in one mentoring style fitting everyone in the lab at once. Undergrads and MSc students usually need more structure and more frequent check-ins; PhD students and postdocs get more room to set their own direction. We will talk explicitly about what kind of support works for you, and revisit it as you grow.
I do not believe in one mentoring style fitting everyone in the lab at once. Undergrads and MSc students usually need more structure and more frequent check-ins; PhD students and postdocs get more room to set their own direction. We will talk explicitly about what kind of support works for you, and revisit it as you grow.
Expectations:
Expectations:
- Curiosity and initiative: I care more about how you think through a problem than about credentials. Half-formed ideas are often where the good stuff starts.
- Rigor: Concepts and cognition are easy to speculate about and hard to measure well. I expect careful experimental design, honest reporting of negative results, and skepticism toward your own favorite hypothesis.
- LLMs and other helpers: Using LLMs is great, but I expect you to know everything you did, double-check your LLMs, and be able to explain every decision. (e.g., "The LLM said it's best, so that's what I did" is not how we do science).
- Follow-through: Research is mostly unglamorous persistence. I expect you to own your project end-to-end, including the boring parts (cleaning data, writing docs, fixing the plot for the third time).
- Tell me when things are stuck: I would rather hear "this isn't working" in week two than in week eight. Flagging a dead end early is a sign of good judgment, not failure.
In return, you can expect me to read your drafts closely, push back honestly, protect your time from unnecessary distractions, and advocate for you.
In return, you can expect me to read your drafts closely, push back honestly, protect your time from unnecessary distractions, and advocate for you.
Page updated
Report abuse