Economics · AI / Data Science · Philosophy
I'm Isaac, an economist by training, deeply interested in causal inference, AI/machine learning, ethics, epistemology, and human cognition. I recently completed an MSc in Social Data Science with Distinction at the University of Oxford, and previously worked as a predoctoral research fellow at the Manos Institute for Cognitive Economics.
My path runs from philosophy and ethics through First Class Honours in economics to applied ML. I'm interested in high-impact careers that involve understanding, predicting, or shaping complex social phenomena.
I designed and ran a funded, representative online RCT (n = 400) testing whether telling people that economists broadly agree on a policy question shifts their own policy beliefs. The design adapts consensus-messaging experiments from the public-health literature to economics. Supervised by Prof. Greg Taylor (Oxford Internet Institute).
Fielded via Qualtrics and Prolific, with the analysis in R and Python, including multiple-hypothesis adjustments and attrition diagnostics.
Why did American economic sentiment stay grim while the fundamentals recovered? I replicated the most prominent academic explanation using a conceptually identical procedure and found that the headline result failed to replicate. I then developed a partisan-asymmetry explanation that accounted for a larger share of the sentiment–reality gap than any competing account in the literature at the time of writing. Supervised by Prof. Richard Holden (UNSW).
Along the way I built a novel Google Trends–based real-time sentiment index that mechanically rules out salient forms of partisan signalling, and built on this work as a research assistant at the University of New South Wales (UNSW).
An economic-history paper used HSV colour histograms from European oil paintings to build high-frequency historical indices of economic growth. I replicated its core methodology (including PCA-based feature extraction) from the written description alone, then extended it with features from a pre-trained neural network to test whether composition, texture, and style encode growth signal beyond colour.
Taught myself database theory and finite model theory from a graduate textbook, well outside the lab's existing expertise, and developed original proposals for applying it to a model of perception in a working paper. I led weekly research sessions for a team of economics professors, presenting technical results and defending the proposed applications under sustained questioning, including live proofs and worked examples. I also read widely across computer science, mathematics, philosophy, and psychology.
Research assistant to Professor Richard Holden. Worked on building a real-time, end-to-end Australian economic sentiment index from Google Trends data. The work involved scraping, cleaning, and time-series construction in R, Python, and Stata, including modifying open-source packages to fit the specifications of several research projects.
Applied ML, deep learning & LLM architectures, algorithmic fairness, quantitative research design.
PhD-level coursework in econometrics, causal inference, micro, macro, and health economics.
Metaethics, applied ethics, metaphysics, epistemology, philosophy of language, and formal logic.
Award for Academic Excellence and Scholars Award; distinguished achiever.
A bot that posts every weekday at 11am with the lunch menus from the cafés around the Oxford Internet Institute and the day's events in nearby buildings. Built with Caleb Agoha, it combines scraped café and event listings with Claude's vision model, which reads the cafés' weekly menu images.
We handed it over to the OII, which integrated it into an internal Microsoft Teams channel and forked it into the institute's GitHub organisation so future cohorts can maintain it.
Essays coming soon.