Prof Simon Jackman, CEO, US Studies Centre, The University of Sydney

IMG_1315.jpg

Prof Simon Jackman, CEO, US Studies Centre, The University of Sydney gave us insight in to how to remove contamination from a protected variable (Xp) on another variable (Z), in order to remove biased data from a machine learning model. A concrete example of this is removing the effect of ethnicity on income in a model that predicts the likelihood of credit default. Here you remove the effect of ethnicity on income and then use the new variable in the model preventing certain groups being discriminated against by an algorithm. He also spoke about how ML algorithms can be used to identify where we can most effectively deploy human labour and how there are zones where we should decide to reject an algorithm’s prediction and get a human involved. He concluded by explaining how data science teams need more definite direction from those product owners with skin in the game. This will require business and policy makers to learn more about statistics and machine learning to deeply understand the pros and cons of each approach available. hashtag#algorithms hashtag#datascience hashtag#artificialintelligence hashtag#usydedsc Jesper Henrichsen Morten Henrichsen Lucas Beck

Andrew Szwec