Nikky Lu
(she/they)
nikky.lu@brown.edu
I am a researcher in computer science, philosophy, and cognitive science. My work is motivated by concepts from philosophy and cognitive science to understand AI systems and human cognition.
Currently, I am pursuing a MS at Brown, working on interpretability inspired by human perception and ways of interaction. Previously, I was a researcher at the Human and Machine Learning Lab at NYU, advised by Dr. Brenden Lake. I was also a trainee in the Computational Neuroscience Training Program funded by NIH at NYU under the mentorship of Dr. Weiji Ma. I wrote my thesis in philosophy of mind and language, exploring whether language model's internal representations are meaningful.
Outside of research, I make art, climb mountains, and just enjoy being a flâneur around the world. See more in the Art Projects and Travel Log sections.
Research Interests
Broadly: Cognitive AI, Interpretability, Representations
Recents
- Spring 2026, Wednesdays 2:00 - 3:00 PMHosting Philosophy of Mind, Language, and AI Reading Group at Brown University
- Spring 2026TAing for CSCI 1460 Computational Linguistics course at Brown. Leading weekly discussion sections on socially responsible computing for AI and particularly LLMs and NLP
- May 19, 2025, 9:00 am - 5:00 pmPresented research at Annual Symposium of Computational Neuroscience at NYU
Research
Ensemble Perception in Vision Transformers (2026)
Investigating whether vision transformers encode scene-level summary statistics across spatial tokens, drawing on psychophysical theories of ensemble perception in human vision.
PI: Dr. Thomas Serre
Stylistic Accommodation in Language Models (2026)
Examining whether alignment training introduces representational priors that affect models' ability to accommodate to interlocutor style in conversation.
Optimizing Environments Through Goal Interpretation (2025)
Studied how varying levels of abstraction in goal description impact human understanding and creativity through behavioral experiments and computational models.
PI: Dr. Brenden Lake
Thesis: The Vector Grounding Problem (2025)
Whether language models can establish intrinsic meaning in their representations beyond mere ascriptions by external interpreters.