Abstract: Artificial Intelligence has taken the world by storm – the data science world has focused predominantly on these predictive applications and machine learning techniques. But as organisations have embraced data driven decision making, the questions we need to answer are more intervention-based and this demands a different toolkit – causal inference. We are excited by the developments happening in academia and the rapid advancement of open source functionality and in this tutorial we will share our learning experience of applying these techniques in business.
By completing this workshop, you will develop an understanding of the core principles of causal inference and how it differs from AI. You will see how in this context often data is not enough, and incorporating business knowledge and understanding is key. You will also become familiar with tools to implement causal inference in your own decision making solutions.
Lesson 1: Why do we need causal inference?
Distinguishing between causal and prediction problems. At the end of this lesson you will be comfortable with the distinction between inference and prediction.
Lesson 2: What is causal inference?
We talk through a pragmatic workflow – the Causal Inference Recipe. A framework of steps to take for any causal inference question with observational data. You will be able to explain the theory of the core components of the recipe.
Lesson 3: How do you do causal inference?
We will talk through a selection of the open source tools for causal inference. We will then go through a notebook that uses the recipe step by step.
Bio: Alice Grout-Smith is a Data Science Manager at Jaguar Land Rover (JLR). Over the past 5 years at JLR she has enjoyed being part of the data science team that has enabled £300m+ of value to the business. After presenting at the Open Data Science Conference back in 2019 on Hierarchical Bayesian Models, Alice and the team have been busy pursuing the exciting applications of causal inference. They have observed that data science in a business context often involves making interventions and taking actions, requiring techniques beyond traditional machine learning. Prior to joining JLR, Alice graduated from the University of Oxford with a degree in Chemistry and a Masters specialising in Quantum Mechanics, which was later published.