Data Science in Context: Foundations, Challenges, Opportunities

If you are a technology professional leading a software product development team big or small -- this book is for you. AI Mirage causes anxiety in your organization. You are expected to deliver AI-first products with broad near human capabilities, while you worry they may fail miserably at the simplest tasks even a child can reliably do.This book covers a few common challenges where software achieved or exceeded human-like capabilities. It shows their evolution from basic data science, through machine learning & AI into revolutionary consumer products. It highlights business, organizational, and societal challenges that necessarily arise when one is successful at scale. And most importantly -- it's written in plain business language, so you don't need to learn Python or upgrade your linear algebra skills.Few years back I worked directly for Alfred and Peter, who built the foundations of Google Research. Their pivotal ACM article with intuitions on how scientific methods can be used in complex organizations like Google have proven tremendously effective for many years. Similarly, I am confident that this very book offers the wisdom essential for your enterprise to be successful in the AI wars of the next decade.

In Data Science in Context, the authors present a broad overview of Data Science. Their introduction to the field includes defintions, a new evaluatory framework, and thorough investigations of specific examples with respect to that framework. More specifically, the framework is called the "Analysis Rubric" and provides a comprehensive approach for evaluating any application of Data Science. This data-scientific method involves seven key steps: the data, the algorithm, their dependability, their understandability, clear objectives, the implications of false positives or negatives, and ethical implications. Dependability encompasses the security of both the data and the algorithm from adversarial missuse. Understandability encompasses the ability to reproduce results and the ability to explain why a specific result is reached. The authors discuss the importance of a shared language, and Data Science in Context provides a common framework and shared language in the Analysis Rubric. Data Science in Context presents a comprehensive introduction to Data Science and demonstrates how valuable an understanding of Data Science is in our modern world, given the abundance of data and the rapid progression of computational power.

Spector, Norvig, Wiggins and Wing do a wonderful job with not just the academic foundations of data science but also many of the interesting complications that emerge in the real-world. My view applies my experience teaching undergraduate and graduate computer science at Stanford and Harvard, and sixteen years at Google, five of which were spent building data systems and teams at YouTube. It is based not on a comparison to other books, but instead on my daily experience designing and building systems that protect the data of billions of users.Data science occupies a peculiar zone of computing, a zone that bridges the gap between the gritty technical topics, like storage systems and algorithms, and the human parts, the entropy of the real world. As such, a strict technical curriculum does not fully prepare an engineer to solve problems in this field. Competence requires breadth, a minimal exposure to adjacent non-technical fields. The introduction of philosophy, sociology, economics and law acknowledges their relevance, helps students recognize relevant challenges, and acknowledges that professionals in the field should recognize situations that demand depth in these independent disciplines.I am pleased that this book considers other basic human rights besides privacy. Liberty is hard to achieve without safety, a right to which the book’s discussions of security and abuse develop. The consideration of interpretability, explainability and auditability speak to the basic public right-to-know that is too often forgotten both in theory and in practice. The authors scope these discussions to data science, but some of this analysis can apply more generally to the integrity of scaled information systems.I applaud the authors for venturing to propose a set of recommendations, which help to highlight the challenges of this moment. Agree or disagree, these proposals provide a basis for an important discussion, both in class and in the broader society impacted by data science and data powered systems.

Data Science in Context offers an excellent and broad introduction to the contemporary issues surrounding the topic of data use. Many of the automated decisions that are made by machines today are fueled by the abundance of data available to corporations. At its start, data science was a purely scientific field that aimed to extract as much value as possible from the available data. But over the last few years, experts in data science have realized that the decisions that are being made by the systems they are building are often of ethical nature and their consequences are often unpredictable to individuals and to societies. This book begins by explaining what data science is in a way that is accessible to a wide audience, and then goes into more detail on the broader implications of using data science techniques. For anyone who is interested in a deep understanding of how data is used today, this book is an indispensable read.

In the rush to make everything into data science, this book takes a step back to consider the whole enterprise. Can everything really be done with data science? Should it be? These authors have worked in data science at the highest level for a long time and it really shows. I love the “rubric” they have come up with to structure the planning of new data science applications and to help anticipate any potential implications. This is a tool that can be used right out of the box.