Introduction

—William B. Weeks, MD, PhD, MBA

Writing is a lonely endeavor that, to be honest, is draining. Authors put a lot of themselves into writing. Picking the right next word, getting the phrasing correct, and accurately conveying the material all take effort. To be sure, spell-check helps, as do grammatical suggestions. However, writing about technical processes and research findings requires a lot of second-guessing and ego oversight. It is not enough just to get words on paper: someone reading them might follow your suggestions, and if the words are misleading or inaccurate, they could be more harmful than helpful. There is an ethical imperative to get the work right, to revise and check and confirm the work and the words so that they accurately depict what you did as a researcher, what you found, and what the limitations of your findings are.

Nonetheless, I love to write about and conduct research. Because of its challenges, I find the research process and the conveyance thereof to be highly intellectually stimulating and engaging. But more importantly, good research, when shared, can improve the world.

After a 30-year career at Dartmouth Medical School as a professor, teacher, and health services researcher who studied health systems and how people used them, I joined Microsoft. I love working at Microsoft and have had wonderful managers here—Dr. Jim Weinstein during my time at Microsoft Research, and now Dr. Juan M. Lavista Ferres, the co-editor of this book and the leader of the AI for Good Lab, the work of which fills this book. But, further, I think that Microsoft's top leadership—Satya Nadella and Brad Smith, who fund the AI for Good Lab—seek to use their positions to do good in the world.

In teaching classes on the financial and strategic management of healthcare organizations at Dartmouth, I often contrasted the two Latin phrases that express the ethics of business and medicine: “caveat emptor” and “primum non nocere,” respectively. Caveat emptor means “let the buyer beware.” If an organization produces something and sells it and it does not work out for the customer, too bad—the customer should have done due diligence and might even have anticipated that the product was not going to be useful. Medicine has an antithetical ethic: first do no harm. Healthcare providers have a fiduciary responsibility to their patients: they have an ethical obligation to share the risks and benefits of treatment decisions and collaboratively work with patients to tailor care pathways to achieve patients' goals in a way that is consistent with their values.

With a mission “to empower every person and every organization on the planet to do more,” Microsoft's ethic aligns more with the medical one than the business one, which is why I like working there. Much like that of a provider and a patient, Microsoft seeks to have long-term and helpful relationships with its customers, ones in which customers benefit from Microsoft products in ways that are consistent with their goals and values.

Perhaps Microsoft's ethic of empowerment is most evident within Microsoft's AI for Good Lab. Considered part of Microsoft's philanthropic efforts, the Lab seeks to engage largely not-for-profit organizations in one of two ways. First, by providing Azure Cloud credits so those with data science expertise can use those credits to begin work on a particular project without incurring cloud storage and compute expenses. Second, by providing time-limited and project-specific data science expertise to those organizations that have data but do not have the advanced analytic skills to use the data to improve the world. The Lab also engages in work that addresses social problems that may not have a specific not-for-profit collaborator, like rapidly assessing damage from natural disasters or war, or providing tools that can help researchers and policymakers identify where broadband access or health inequities exist in the United States.

That work is presented in this book. By writing the book, we seek to help readers who are interested in how artificial intelligence and advanced data science techniques can be used to solve world problems by providing examples of the Lab's efforts to do so. So, part of the reason to write this book is to share knowledge with others who are interested, might learn about the approaches that the Lab has used, and, hopefully, apply those methods in a propagative way to address more problems. The world is complex, and we need as many thoughtful, curious, and motivated people who want to spend time addressing its problems as possible. We hope this book reaches them.

Moreover, I worry. As a physician, I worry about the world's health and the massive inequities in care access, quality, and outcomes that drive health disparities, within countries and across countries. As an economist, I worry that, unless efficient and effective solutions to some of the most pressing issues in the world (like climate change, humanitarian action, and health equity) are addressed, incentives that drive market behavior will worsen the plight of the disenfranchised. As a father of six and grandfather of four, I worry that the world I leave to my kids and theirs will be a worse one than the one I inherited.

So, my primary reason for writing a book that demonstrates how artificial intelligence and sophisticated data analytics can be used to solve the world's most pressing problems is because I have hope that these technologies can help, and hope assuages worry. The tools that are described herein are not panaceas—just as with a medical intervention, the choice of the tool, the approach, and the target must be clearly described, cautiously applied, and carefully interpreted.

But I am hopeful that these tools, when judiciously, rigorously, and ethically applied, can empower the world's populations to live in more just societies, avoid unnecessary harms that might otherwise befall them, and live healthier, more fruitful, and more fulfilling lives.

A Call to Action

—Juan M. Lavista Ferres, PhD, MD, MBA

Jeffrey Hammerbacher, one of the first data scientists at Facebook, once said, “The best minds of my generation are thinking about how to make people click ads.”

When I first came across Hammerbacher's quote, I was leading the metrics team for Bing. Although I wasn't working directly on ads, part of my job was to understand the trade-offs between ads and relevance, so this statement resonated deeply with me.

To be clear, search engines like Bing and Google have immensely enriched society by granting unparalleled access to information and empowering individuals in novel ways. Yet, while recognizing these contributions, the pressing challenges of our times necessitate collective innovation, creative application of new tools and methods, and solutions that attempt to solve those challenges and not just provide information access.

The problems that we seek to address within the AI for Good Lab are foundational to societal improvement. For example, each year, millions of children die before they reach the age of five, with a significant majority of these deaths being entirely preventable. The climate crisis affects hundreds of millions of people, a staggering 1.6 billion people live with severe disabilities, and half of the world's population has inadequate access to high-quality healthcare. The world needs all the help it can get.

Over the past five years, I've had the opportunity to see the remarkable ways in which artificial intelligence and technology can address some of those challenges. While they aren't silver bullets, artificial intelligence and technology can be instrumental in solving specific issues. However, one challenge is that non-profit organizations and governments, which are at the forefront of addressing these problems, often do not have the capacity to attract or retain the artificial intelligence experts that they need to solve them.

Some might find it surprising, but even though predicting which person will click on your ad and which child has a higher chance of infant mortality are vastly different in societal terms, from a pure data science standpoint, they are essentially the same problem. If we can apply AI algorithms to optimize ad clicks, why can't we direct some of our best minds and most advanced technologies toward optimizing human life, well-being, and the health of our planet?

I am optimistic and strive to leave the world better than I found it—a goal I believe is more widespread than commonly perceived. I'm profoundly grateful to Microsoft for the chance to lead our AI for Good Lab that embraces this very mission.

I'm not going to lie: demonstrating an impact in this complex world is not easy. Over the years, my successes have taught me a lot, but my failures have taught me even more about making a tangible impact. And I've learned a few lessons.

I learned that there is a huge difference between solving a problem on paper and solving one in a real-world setting. One profound realization I had was that, as humans, we are addicted to complexity: we like complex problems and complex projects. This is the reason we sent a person to the moon before we added wheels to our luggage. However, seeking complexity is the wrong approach. If we want to try to impress people and look smart, our solutions will be complex. But if we want to measurably improve the world, our solutions must be simple. And building simple solutions is much harder.

I've...