Abstract

By now most of us have heard of various instances where mathematical work has ignited public concern and outcry. These have occurred across many areas of mathematics, including finance (Collateralised Debt Obligations), number theory and cryptography (mass surveillance tools), AI (problematic outcomes in machine learning), statistics (the 2020 A-levels grading fiasco), algorithms (bias in predictive policing/judicial tools), etc. Most of the concerns raised in the public were directed at the OUTCOMES of such work, but not the processes involved in DOING the mathematical work itself. However, these crucial ethical issues do not appear from nowhere; they often come from the WAY in which mathematical work is carried out. In this talk I will discuss the various steps we take when we use mathematics to assist in solving a tangible problem. Such work is never done in one day. Instead, it is carried out over time and broken up into stages. It is this PROCESS, from identifying the problem, to formulating how a solution might be implemented, to collecting the relevant data and making the appropriate calculations, to deploying the output in a tangible world, and finally following up on that deployment, that can be scrutinised. By understanding the various shortcomings in such processes, we can begin to see where and how ethical issues are "injected" into our final products.