Hi — I'm Abraham Jacob Fresen (Bram).
I'm a PhD researcher at the Max Planck Institute for Psycholinguistics.
I study how language models predict language — and the brain.
Language models.
They are remarkably good at predicting language.
Surprisingly, they also predict how the human brain responds to language.
But why?
To find out, I look at which concepts a model represents.
Then how accurately it represents them — and whether that drives how well it predicts the brain.
Take syntax: the structure a model assigns to a sentence.
Do models that capture syntax better also predict the brain better?
And if I erase the model's grasp of syntax — does its prediction of the brain break too?
That's how I find which features actually matter.
Language models don't just predict language — they also predict how the human brain responds to it. Why?
That is the question my PhD asks, using tools from AI interpretability and neuroimaging. I probe which human-interpretable features (like syntax) models and brains represent, how accurately, and whether that is what drives a model's ability to predict the brain.
I can also manipulate a model's internal representations — erasing its grasp of a concept, say — to causally infer which features actually matter for brain alignment.
I do this in the Language and Predictive Computation Group at the Max Planck Institute for Psycholinguistics, supervised by Dr. Micha Heilbron.
Alongside my research, I work part-time as an AI & Innovation Consultant at ClickValue, building applied machine-learning systems and automation — from retrieval-augmented AI agents to causal-impact analytics pipelines.
- NeuroAI
- Predictive computation
- Model–brain alignment
- Mechanistic interpretability
- Computational linguistics
News
- 2026 News Started my PhD at the Language and Predictive Computation Group, Max Planck Institute for Psycholinguistics, Nijmegen, supervised by Dr. Micha Heilbron.
- 2025 News Obtained my Research MSc in Brain & Cognitive Sciences (cum laude) at the University of Amsterdam.
- 2024 Conference Presented our work on syntactic representation and brain alignment at CogSci 2024.
- 2024 Paper New paper out in the Proceedings of the Cognitive Science Society — see Publications.
Publications
AI & Innovation Consulting
I work part-time as an AI & Innovation consultant and engineer at ClickValue, helping teams scope, prototype, and ship applied machine-learning solutions.
What I help with
- AI agents & retrieval-augmented generation (RAG) — design and implementation
- Applied ML / NLP prototyping, web apps, and data pipelines
- Causal-impact analysis, A/B experiment design, and statistical consulting
- LLM evaluation, interpretability, and model-behaviour auditing
- Workflow automation, dashboards, and analytics engineering
Interested in working together? Get in touch.
CV
Education
- 2022 – 2025 Research MSc Brain & Cognitive Sciences (Cum Laude) — University of Amsterdam
- 2017 – 2022 BSc Cognition, Language and Communication, minor Artificial Intelligence — University of Amsterdam
Experience
- 2026 – present PhD Researcher — Language and Predictive Computation Group, Max Planck Institute for Psycholinguistics, Nijmegen. Supervised by Dr. Micha Heilbron. Full-time
- 2026 – present AI & Innovation Consultant — ClickValue Part-time
- 2024 – 2026 AI & Innovation Consultant — ClickValue Full-time
- 2024 – 2025 Research Intern, NeuroAI & interpretability — Amsterdam Brain & Cognition (ABC), University of Amsterdam. Causal probing of LLM representations against fMRI and concept erasure (LEACE); supervised by Dr. Micha Heilbron.
- 2023 – 2024 Research Intern, model–brain alignment — Institute for Logic, Language and Computation (ILLC), University of Amsterdam. LLM–fMRI alignment and dependency probing (DepProbe); supervised by Dr. Jelle Zuidema & Marianne de Heer Kloots.
- 2018 – 2024 Teaching Assistant (statistics, cognition, research methods) — University of Amsterdam
- 2022 Research Intern, psycholinguistics (EEG) — University of Amsterdam. Supervised by Dr. Monique Flecken.