I am an applied mathematician, quant and data scientist with almost two decades of experience building and validating models across academia and industry. My focus is measurement: defining what “correct” means for a model, building the tests that check it, and identifying sources of bias and accounting for them.
Research
I am the author of Mitigating Bias in Machine Learning, a book in progress, and co-author of “Bias in Retrieval Systems” (ACM Books, 2024) and “Bias and the Web” (Springer, 2023). In 2023 I was listed among the 100 Brilliant Women in AI Ethics™.
My research interests include:
- Individual fairness [1], [2], which I believe provides a blueprint for how we should think about the fairness of predictive algorithms. My work reconciles conflicting research on its meaning and behaviour.
- The apparent trade-off between individual and group fairness [3], [4]. I argue that individual fairness is better understood as an extension of group fairness than as acting in opposition to it, and that existing empirical evidence of a trade-off between the two [5] in fact demonstrates the well-understood trade-off between utility and fairness.
- Generalised entropy indices as utility or objective functions, and how different parameter choices might give better measures of utility across both groups and individuals.
- Mapping philosophical notions of fairness to mathematical criteria, and mathematically rigorous approaches to questions about fairness.
- Evaluation bias in search and recommender systems: sampling from heavily skewed distributions, incomplete relevance judgements and data missing not at random.
Career
In 2023 I was a Senior Research Scientist on the Responsible AI team at the Institute for Experiential AI at Northeastern University, where I reviewed a health insurer’s predictive model for bias in healthcare eligibility decisions.
Before this, I was a Principal Data Scientist at Doma (formerly States Title), working on title risk products, and an Artificial Intelligence Fellow at Insight Data Science, where I developed DebiasML, a practical and effective method for reducing bias in machine learning algorithms.
I spent over a decade as a Derivatives Quant at the Royal Bank of Scotland, managing pricing model risk as part of the independent model validation team. My work involved analysing and building pricing models for exotic instruments to quantify valuation risk and determine their approval for use, devising mitigation strategies, and designing model governance processes and standards. The role was cross-functional, negotiating agreement among traders, regulators and everyone in between.
Applied Mathematics
I have a Doctorate (Ph.D.) in Fluid Dynamics (specifically looking at supersonic vortex breakdown), a Masters (M.Sc.) in Theoretical and Applied Fluid Dynamics and a Bachelors in Mathematics (B.Sc. Hons). I studied at the School of Mathematics at The University of Manchester – where Osborne Reynolds (of Reynolds number fame) spent his career and Alan Turing formulated the Turing test and helped develop the Manchester Computers (the world’s first stored-program and transistorised computer).
References
- Aristotle, Nicomachean Ethics, V.3. 1131a10–b15; Politics, III.9.1280 a8–15, III. 12. 1282b18–23
- Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, and Rich Zemel. Fairness through awareness, 2011.
- Arvind Narayanan. “Tutorial: 21 fairness definitions and their politics”. 2018.
- Reuben Binns. “On the Apparent Conflict Between Individual and Group Fairness.” 2019.
- Till Speicher, Hoda Heidari, Nina Grgic-Hlaca, Krishna P. Gummadi, Adish Singla, Adrian Weller, and Muhammad Bilal Zafar. A unified approach to quantifying algorithmic unfairness. Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, July 2018.