How to Think about Data Science

Book Preface

Data science is an exciting and fast-evolving field that is no longer the exclusive domain of computer scientists and mathematicians. Today, many other professionals such as engineers, physicists, security experts, lawyers and hospital administrators approach this field and often make very valuable contributions thanks to their unique perspectives and backgrounds.
Despite the abundance of excellent books and resources on data science (many of which I reference here) the standard focus is on solving specific problems programmatically while ignoring the bigger picture. My goal with How to Think about Data Science is to present the field of data science in a critical, conceptual and holistic manner where all the rele-vant facets (technical, social and philosophical) are integrated. Data science is made up of many straightforward ideas whose combination gives rise to an ecosystem consisting of multiple, interconnected elements. These include topics as important as the rational choice of algorithms, the nature of data bias and algorithmic discrimination, data privacy and cybersecurity, as well as being able to understand the risks involved in the deployment of artificial intelligence tools.
This book contains no programming code – there are already plenty of ‘how to’ books and online courses out there detailing algorithms and their implementation. Here, we become literate in data science by asking the why behind data science problems and by learning to reason about these problems holistically instead of following recipes. Thus, we take the scientist’s approach of focusing on asking the right question as this is the single most im-portant factor contributing to the success of a data science project. We discuss key concepts and how they are linked together, the pros and cons of various methods, all while abstract-ing away from the minutiae of the implementation. Developing a holistic understanding is a most valuable skill that will survive specific programming fashions, working styles and computing platforms.
This book is accessible to readers without a strong computational background (although you should be numerically literate), but it is also of much interest to data science practition-ers. For example, a programmer in a start-up who wears many different hats might benefit from reading the chapter on algorithmic bias (Chapter 5) or the chapter on cybersecurity (Chapter 6). Likewise, a regulator working on data privacy will benefit from having a critical understanding of the different types of supervised learning methods, their limitations and areas of applicability (Chapter 3).
The advice I give here is the distillation of my experience, both as a practitioner and as a mentor of the many data scientists I have trained in my career. I hope you enjoy reading this book as much as I have enjoyed writing it.

What is the difference between a regular cook and a chef? A cook may follow recipes and create edible dishes. However, knowing what ingredients should be used and how they should be prepared, combined and presented makes all the difference. What really adds value here is knowing the right tools, ingredients and processes and how they should be orchestrated to create a terrific dish. Data science is no different: Anyone can run a clustering algorithm or train a neural network with minimal effort, but what really matters is knowing, for a specific dataset, what questions should be answered, what algorithms should be used to answer each one of these questions, as well as the ethical and data privacy issues that must be contemplated. The data scientist who can answer these questions can not only follow recipes but is also capable of applying the right algorithms to answer the right questions while minimising potentially discriminating outputs. This book focuses on these relevant questions; if you wish to cook a terrific dish, this book will undoubtedly help you.
Jordi Conesa i Caralt, PhD
Associate Professor of Computer Science
Universitat Oberta de Catalunya