Doing an exploratory fairness analysis and measuring fairness using equal opportunity, equalized odds and disparate impact — It is no longer enough to build models that make accurate predictions. We also need to make sure that those predictions are fair. Doing so will reduce the harm of biased predictions. As a result, you will go a long way in building trust in your AI systems. To correct…

Analysing non-linear relationships with partial dependence plots (PDPs), mutual information and feature importance — When you first start driving you are less experienced and, sometimes, more reckless. As you age, you gain more experience (and sense) and it becomes less likely that you’re involved in an accident. However, this trend won’t continue forever. When you reach old age your eyesight may deteriorate or your…

Why we need to understand how our models make predictions — Should we always trust a model that performs well? A model could reject your application for a mortgage or diagnose you with cancer. The consequences of these decisions are serious and, even if they are correct, we would expect an explanation. A human would be able to tell you that…

Analysing interactions using feature importance, Friedman’s H-statistic and ICE Plots — The side effects of medication can depend on your gender. Inhaling asbestos increases the chance of lung cancer more for smokers than non-smokers. If you are more moderate/liberal your acceptance of climate change tends to increase with higher levels of education. The opposite is true for the most conservative. These…

How to explain your model in a human-friendly way — An important part of a data scientist’s role is to explain model predictions. Often, the person receiving the explanation will be non-technical. If you start talking about cost functions, hyperparameters or p-values you will be met with blank stares. We need to translate these technical concepts into layman’s terms. …

How to create global aggregations of LIME weights — LIME is a popular method for explaining how machine learning models work. It aims at explaining how individual predictions are made. The LIME Python package makes it easy to get these local explanations. However, if we want to understand how the model works as a whole it is lacking. …

Common sources of bias — historical bias, proxy variables, unbalanced datasets, algorithm choices and user interaction — On the surface, machine learning looks impartial. Algorithms have no concept of sensitive characteristics like race, ethnicity, gender or religion. So it shouldn't be possible for them to make biased decisions against certain groups. Yet, if left unchecked, they will do exactly that. To correct biased decisions we first need…

All mighty methods that can be used with any model — An agnostic ponders the existence of god. Thankfully, machine learning is not quite there. When we talk about an agnostic method we mean one that can be used with any model. We’ll elaborate on this definition and the focus will be on agnostic methods for interpretability. We’ll discuss different classifications…