What does an AI ethicist do?

Microsoft was one of the earliest companies to begin discussing and advocating for an ethical perspective on artificial intelligence. The issue began to take off at the company in 2016, when CEO Satya Nadella spoke at a developer conference about how the company viewed some of the ethical issues around AI, and later that year published an article about these issues. Nadella’s primary focus was on Microsoft’s orientation toward using AI to augment human capabilities and building trust into intelligent products. The next year, Microsoft’s R&D head Eric Horvitz partnered with Microsoft’s president and chief legal officer Brad Smith to form Aether, a cross-functional committee addressing AI and ethics in engineering and research.

With these foundations laid, in 2018, Microsoft established a full-time position in AI policy and ethics. Tim O’Brien, who has been with Microsoft for 15 years as a general manager, first in platform strategy and then global communications, took on the role.

Of course, many organizations are increasingly paying attention to ethical issues around AI. In a 2018 Deloitte survey, 32% of AI-aware executives ranked the ethical risks of AI as one of their top three AI-related concerns. Microsoft and O’Brien are essentially bellwethers on this issue — the figurative sheep at the head of the flock — in creating a role focused on, as O’Brien puts it, AI ethics “advocacy and evangelism.”

I talked with O’Brien to find out how his role came about, what he does in it, and what kinds of policies might emerge at Microsoft because of his work. I also asked him how the AI ethicist role might relate to similar positions at other companies.

How the AI Ethicist Job Came About

O’Brien told me that he was inspired by the early steps Microsoft’s leaders had taken on AI ethics, and that he asked to pursue the topic for the company as his primary job. He’d been a product marketer for various tech companies early in his career and spent much of the last 15 years at Microsoft pitching developers on why they should build stuff on the company’s platforms.

I asked O’Brien if he was hiding a philosophy or theology degree on his LinkedIn profile, but he laughingly said no. He said he did have some exposure to ethics issues in his business law classes as an MBA student and that he went to a Jesuit high school — a Catholic order known for its focus on ethics and philosophy. That’s it, though, for his formal educational preparation in ethics.

His interest developed when was running Microsoft’s Global Communications function outside the U.S., a role that had him traveling constantly and frequently exposed to the “geopolitics of technology,” as he put it. He began to realize, he said, that leaving all of businesses’ ethical and policy issues to the lawyers was not sufficient. O’Brien was seeing that, as data and processing move to the cloud, technology, policy, and geography issues all were starting to collide.

O’Brien was quick to point out that he’s standing on the shoulders of giants. “Most of the ideas I talk about are not mine,” he said. He mentioned that Microsoft president Smith has been preaching the AI ethics gospel for several years now, and indeed Smith speaks and writes regularly about the need to consider the ethical, social, and political aspects of the technology. O’Brien sees part of his role as helping to disseminate Smith’s ideas at scale — Smith is a powerful steward for responsible AI at Microsoft, but he’s a busy guy. And the ethical issues around AI are sufficiently complex and nuanced that Smith can’t campaign for them all.

How an AI Ethicist Spends His Time

O’Brien’s job began in May 2018, and he began traveling around the world almost immediately. His primary activities have included giving speeches at conferences, meeting with Microsoft customers, leading a research effort, talking with analysts and researchers, and coordinating activity across Microsoft.

O’Brien’s presentations have focused on topics such as avoiding algorithmic bias and creating transparency in AI models. Some of Microsoft’s ideas on these issues have been described in the press. It’s important to clarify that although O’Brien’s job title is focused on AI, he promotes ethical perspectives on all information technologies. O’Brien said he views AI as the “spark that lit ethics in tech,” including analytics, the internet of things, and virtual and augmented reality, as well as AI. Many solutions in the real world are, of course, hybrids of these technologies.

The goal of the research initiative he heads is to develop a global perspective on tech ethics. In his previous job, O’Brien observed that different societies around the world have very different perspectives on privacy and ethics. Just within Europe, for example, U.K. citizens are willing to tolerate video camera monitoring on London’s central High Street, perhaps because of IRA bombings of the past, while Germans are much more privacy oriented, influenced by the former intrusions of East German Stasi spies. O’Brien suggested that in China, the public is tolerant of AI-driven applications like facial recognition and social credit scores at least in part because social order is a key tenet of Confucian moral philosophy.

Microsoft’s AI ethics research project involves ethnographic analysis of different cultures, gathered through close observation of behaviors, and advice from external academics such as Erin Meyer of INSEAD, who wrote The Culture Map. It may result in a number of publications, depending on the findings. My feeling, and O’Brien’s as well, is that there aren’t a lot of sources out there on global tech ethics. This research could start to fill that gap.

Within Microsoft, O’Brien is trying to extend the community of people who are focused on the ethics topic. He sent me a collection of materials suggesting that lots of people are already engaged. The materials include, for example, a blog post on Microsoft guidelines about developing responsible conversation bots (no doubt occasioned by the off-the-rails comments from the Tay AI chatbot research project in 2016). They also include a presentation and article on intelligible models in health care, and an internal project for a “learning door” that recognizes (with opt-in) who is coming in and out of Microsoft buildings. O’Brien said he works closely with Smith’s legal team and also has a matrixed reporting relationship to Eric Horvitz, technical fellow and director at Microsoft Research Labs. Of course, Microsoft is a big company, but awareness of AI ethics seems to be catching on rapidly and broadly there.

This bottom-up, middle-out approach to change is what CEO Nadella is trying to inculcate within the Microsoft culture. The days of asking for a million dollars and a hundred people for a new project are over, O’Brien says. The desired approach is to get an idea in action and try to attract the needed resources over time.

AI Policy Development at Microsoft

O’Brien expects that there will eventually be a collection of policies within Microsoft about how to use AI and related technologies. Some have already emerged, such as Smith’s 2018 directive on facial recognition technologies, which stated through a blog post that Microsoft would not sell its technology for facial recognition in ways that can “adversely affect consumers and citizens” or “encroach on democratic freedoms and human rights.”

O’Brien believes there will be a number of other policies that the company will define, from avoiding algorithmic bias to model transparency to specific applications like predictive policing. Microsoft’s activism in the privacy and ethics space isn’t new; it has battled the U.S. government on the privacy of sensitive customer data in the cloud (the case was eventually vacated by the U.S. Supreme Court after legislative action), and also filed a brief in support of Apple’s privacy position on the iPhone linked to the shootings that took place in 2015 in San Bernadino, California.

But there are many questions on which concrete policies remain to be developed. O’Brien mentions, for example, mistakes made by AI systems and legal remediation for harm from AI. He believes that the most interesting part of the field is the many unresolved questions within it.

Microsoft is a big player in AI and tech more broadly, so there is little doubt in my mind that having an AI ethicist role is a good idea for the company. In the time since O’Brien was named to the role, Salesforce.com and a few other vendors have created similar positions. Other companies that are not major IT vendors may not want to advocate as strongly in public about AI-related issues. And it wouldn’t be a good idea to create an AI ethicist role without the strong leadership and support from senior executives that Microsoft exhibits. However, as I have suggested, many other businesses will face issues similar to the ones that O’Brien is wrestling with for Microsoft, and having an internal role or group to address them will be useful.