Overview
My current research lies at the intersection of the philosophy of science, philosophy of AI, and data- and computational philosophy. I focus on the epistemology of machine learning: how learning from data functions, how stability and simplicity relate to generalization, and when model outputs count as scientific knowledge.
Dissertation
Working title: Toward an Epistemology of Machine Learning.
The project investigates whether and how machine learning contributes to scientific knowledge. It analyzes the roles of data and phenomena, the relation between models and theories, and how explanation and understanding interact with predictive success. On the technical side, it connects philosophical analyses of simplicity with deep‑learning practice by studying pruning and training dynamics. I introduce a Noise‑Cancelling Razor that captures dynamic simplicity: subnetworks that remain stable under controlled perturbations (e.g., noisy validation signals used to guide early stopping) tend to generalize better. This bridges conceptual debates with empirically testable criteria.
Current Working on...
- Dynamic Simplicity & Pruning: From the Lottery Ticket Hypothesis to the Noise‑Cancelling Razor; sparsity, early‑stopping stability, and noise‑cancelling indices.
- Reverse Prompt Engineering (assistant) : Mining and classifying elementary Python questions across forums; mapping to curricula; evaluating LLMs (ChatGPT, Copilot, Gemini, Grok) on foundational skills.
Areas
- AOS: Philosophy of Science; Epistemology; Philosophy of AI.
- AOC: Logic; Philosophy of Language; Philosophy of Mind; Metaphysics.