�鶹��Ƶ

Associate Professor Margot E. Kaminski teaches, researches, and writes on law and technology. Her groundbreaking work has focused on privacy, speech, and online civil liberties, in addition to international intellectual property law and legal issues raised by artificial intelligence (AI) and robotics. She also serves as director of the Privacy Initiative at Silicon Flatirons.

What initially attracted you to this area of the law?

I worked in publishing after college, right around the time that e-books were becoming popular. That, along with the rise of social media, got me interested in how we arrange existing legal rights around new technologies—from speech to privacy to intellectual property rights.

Later, while in law school, I interned for a summer at the Electronic Frontier Foundation (EFF), a nonprofit in San Francisco that describes itself as “defending civil liberties in the digital world.” There was a team there working on national security surveillance litigation. That really opened my eyes to the central significance of human rights in the digital age. But I’ve always been interested in the nitty gritty aspects of regulation, especially the role that transparency can play. I came back to law school and co-founded a clinic, the Media Freedom and Information Access (MFIA) clinic, which litigates government accountability and transparency cases, including cases involving new technologies. We found ourselves asking who should count as a journalist, and what role information technologies were playing in the face of existing power disparities—both as tools of surveillance and tools of accountability.

By the time I became a professor, I had really honed in on privacy. It’s such a fast-moving policy space, with big implications for democracy and individual freedom. I spent a few years working on the privacy issues raised by unmanned aerial vehicles (UAVs), or drones, and then got the opportunity to go to Europe through a Fulbright grant. That’s when I turned to working increasingly on comparative data privacy law, with a focus on automated decisionmaking systems and AI.

Have legal issues raised by AI and robotics changed since you graduated from law school?

Gosh, yes. When I was in law school many of these questions felt ahead of the curve or hypothetical. Now, we have companies using facial recognition (a type of computer algorithm) to scan applicants’ faces to try to determine emotions and extrapolate personality traits, for example. We have government agencies using algorithms, including AI systems, to try to allocate benefits or catch fraud. The use of and investment in AI systems is everywhere, and lawmakers are taking the potential harms of AI systems seriously. Colorado, for example, just enacted a new law on facial recognition.

You’ve written extensively about the role of AI algorithms in decision making. How do you see the future of balancing decisional authority between humans and machines?

This is a hard one. I don’t see my role as predicting the future, necessarily. I’m more interested in trying to figure out what the law can do to ensure that human values stay on the table—that as we increasingly create and use new sociotechnical systems, we put in place whatever’s necessary to make sure we don’t lose sight of what matters. I definitely believe, for a number of reasons, that there are some decisional realms where we will always use humans. Legal decisions, for example, aren’t just about correctness and efficiency. Legitimacy, justification, accountability, even a respect for the dignitary rights of the person affected by a decision—these are all reasons why law, at least as an ideal, isn’t suited to automation.

The most interesting problems aren’t about whether to use a machine or a human, but about how to get them to work together. For example, I’ve coauthored this recent article on “humans in the loop,” or the people involved in automated decisions. Often, well-intentioned lawmakers will look at a decision made by an AI system and try to solve some set of perceived problems by requiring that a human be involved. It’s not that humans are worse decision makers than machines—in fact, humans still do a lot of things, like crossing contexts or dealing with edge cases, much better. But putting a human in the loop thoughtlessly actually creates new problems. Hybrid human-machine systems have known weaknesses and can be subject to complex failure cascades. So if we’re going to put a human in the loop of particularly significant automated decisions, we have to know why we’re putting her there, and set her (and the system) up to succeed.

You recently worked with Colorado Law’s Samuelson-Glushko Technology Law & Policy Clinic to develop comments responding to the Attorney General’s Pre-Rulemaking Considerations for the Colorado Privacy Act. Tell us about that work.

I feel so lucky to be at a school with a tech law clinic! Professor [Blake] Reid ’10 is a joy to work with, and his students (who are often also my students, from other classes) take their work very seriously and produce impressive and important output. This most recent project, responding to the Colorado AG’s office on the Colorado Privacy Act, is a great example. Two clinic students worked to exhaustively identify aspects of the act that could benefit from focused rulemaking. They did an extraordinary amount of research, ranging from technical articles on how to best design an effective consent stream, to organizational literature on how to make an impact assessment successful. They also pointed the AG’s office to resources on other privacy laws, both in Europe and in other states like California. This is particularly important as states like ours weigh the benefits of harmonization, which typically makes for lower compliance costs for businesses, with the appeal of being a policy leader in the consumer protection space.

Colorado Law’s Tech Law and Policy program, along with the law school’s Silicon Flatirons Center for Law, Technology, and Entrepreneurship, are nationally recognized. How would you like to see these programs evolve or grow?

Again, I feel very lucky to be at a school that has so many faculty members working in related spaces. Each of us does something slightly different—Professor [Kristelia] Garcia works on copyright law, Professor [Brad] Bernthal '01 on entrepreneurship, Professor Reid on telecommunications and platform law, Professor [Harry] Surden on patent law and a different area of AI—but we’re able to collectively offer our students a depth of expertise and classes that aren’t really available elsewhere, except at a few very top law schools. There is, however, always room for growth. I would love to see us be able to offer our students more privacy courses, in particular. It would be amazing to be able to offer data privacy for practitioners or an international privacy course. We also, despite the expertise on our faculty, have yet to offer a class on law and AI!

What research themes or projects are you most looking forward to digging into in the coming year?

I have a few projects I’m really excited about. This “humans in the loop” piece I already mentioned is a big one. So is a piece I’ve been revising this summer called Regulating the Risks of AI. Most laws targeting AI have been risk regulation—the kind of thing we use, for example, in environmental law, or that companies use to try to mitigate risks. Risk regulation comes with a particular set of policy baggage. It’s been fun—and challenging!—to dig into how aspects of it do and don’t work when it’s applied to algorithms and associated practices.

I’m also really looking forward this fall to getting back into a piece I’ve been calling Data as Speech Infrastructure, where I’ll be looking at data privacy laws through the lens of the First Amendment. And there’s a good chance something will come of all of the discussions I’ve been having about the data privacy implications of the Supreme Court’s decision revoking the right to abortion in Dobbs. In short, there’s always something to do.