Statistical Science


Dr Ricardo Silva's work on algorithmic fairness mentioned in New Scientist article

30 March 2017

A recent manuscript by Dr Ricardo Silva, on ways of enforcing fairness in predictive modelling, has been given coverage by the New Scientist magazine this week. The manuscript, posted in arXiv earlier this month, prescribes a way of making explicit causal assumptions about how demographic background of an individual may lead to unfair, discriminative, treatment, such as in racially biased stop-and-frisk policing and loan approvals. The work makes the link of matters of fairness to mathematical formalisms of causality and counterfactuals, and suggests ways to test and communicate how automated decision making from data can be certified to respect fairness guidelines.

The work was done in collaboration with Dr Matt Kusner, Dr Joshua Loftus and Dr Chris Russell from the Alan Turing Institute, where Dr Silva is also a Faculty Fellow.