Navigating the robot car revolution

(The following transcript has only been lightly edited)

Thoko Moyo: Hello and welcome to the Harvard Kennedy School PolicyCast. I’m Thoko Moyo.

On March the 18th, 2018, 49 year old Elaine Herzberg was pushing her bicycle across a four lane road in Tempe, Arizona when she was struck by a self driving car that was being tested by the rideshare company, Uber. She died of her injuries in hospital and became the first pedestrian to die in a collision involving self driving cars. Her death magnified concerns about the safety of autonomous vehicles, and sparked debates about who should be held responsible for her death. In the end, no charges were brought against Uber. But questions about safety of these cars and how to govern them have continued. Here to discuss some of the policy issues around self driving cars is Kennedy School lecturer Mark Fagan. Mark is the head of the autonomous vehicles policy initiative at the Taubman Center for State and Local Government here at the Kennedy School. The initiative works to make sure local governments can successfully navigate game changing disruption that introducing these driverless cars will bring to our streets and to our daily lives. Welcome to policy cast, Mark.

Mark Fagan: Thank you so much.

Thoko Moyo: It’s been well over a year since Elaine Herzberg was hit and killed while crossing the street. Are there any policy interventions that you think might have prevented the tragedy happening as we look back?

Mark Fagan: The issue around safety of autonomous vehicles is actually really interesting, because one of the key drivers that supports the innovation is the safety benefits that are supposed to result from autonomous vehicles. When we look at the 30,000 people a year who die in road crashes here in the United States, the vast majority of those are as a result of human factor failures. Autonomous vehicles have the promise of being able to substantially reduce that. Yet here we are talking about the very unfortunate death of a pedestrian as a result of an autonomous vehicle.

To your question about policy and associated regulation, the answer is absolutely. It’s the obligation of all levels of government. But right now, especially at the state and local level, to ensure that the AV testing that is taking place is being done in a regulated environment and in a safe environment so that these kinds of tragedies don’t happen in the future.

Thoko Moyo: So where are we now when you come to think about issues around policy? Because one of the concerns that people have is when Uber and other apps were launched, the policy ran behind the introduction of these new ways of transportation. Where are we when it comes to self driving vehicles?

Mark Fagan: Well, you’ve hit the nail on the head from a policy perspective in terms of why the importance exists right now to start working on autonomous vehicle policy. As you described, often the technology does run ahead of regulation and policy. Here in the case of AVs, we’re really trying to get ahead of the curve. The way in which we’re doing it is we’re trying to establish the policies and the safety frameworks, and the regulatory frameworks. To ensure that as this technology develops, it is brought into our world in a very safe fashion.

You hear a lot of discussion around, “Autonomous vehicles, they’re still years or even decades away. Why do we need to worry about policy right now?” And what our perspective is, the reason you need to worry about it now is to get ahead. Right now, no one has an autonomous vehicle. They aren’t actively commercially available. This is the time to regulate, because you can really set the guide rails for how this innovative technology will be brought into our world. But to do it in a way that balances the goals of innovation. We want innovation and it brings lots of interesting opportunities. But to do it in a safe, controlled manner.

Thoko Moyo: So where is this push to get ahead with the policy coming from? Is it actually from the people that are working on the technology for self driving cars? Is it from the regulators? Where is this push coming from?

Mark Fagan: I see the dominant push coming at state and local government level, and it’s particularly the people who were caught flatfooted by the TNCs, the transportation network companies, like Uber and Lyft, who came to town with a very interesting, new innovation in mobility. And completely took control. The regulation was too late at that point. So here at the Kennedy School, we have an intersection right outside. And it is very common to see congestion there because a TNC has stopped to either pick someone up or drop someone off. Policy makers see that and know how hard it is to retroactively make regulations where you’re going to pick up and drop off people. The reason people are interested in AVs right now is you can make that regulation without having as much pushback on it.

Thoko Moyo: And let’s come back to that issue of congestion, because I think I was reading somewhere or someone might’ve told me that some of the concerns actually about self driving vehicles might be that they cause congestion because it is cheaper for them to circle around parking lots than actually go and park. But let’s come back to that because I want to take a couple steps back, and maybe just make sure that I understand first of all. When we talk about self driving vehicles in this context, what are we talking about? I know you said that they are a long ways away. But we’ve seen self driving vehicles on farms. We have them in mines. So what exactly are we talking about right now when we talk about these driverless cars?

Mark Fagan: The degree of autonomy actually has a formal structure at this point in time. The Society of Automotive Engineers have created a five level determination of the degree of autonomy. When I talk about autonomous vehicles in most people in common parlance, they’re talking about what’s referred to as level four or level five, where the vehicle is capable of driving itself without having any human intervention.

Thoko Moyo: And that’s the highest level?

Mark Fagan: Those are the highest levels. Level four has restrictions in terms of geography, so it’s geo-fenced to operate in a particular area. And level four also won’t necessarily operate in every weather condition. For example in a snow storm. Level five, and it’s a little uncertain whether we’ll actually even get to level five, is complete autonomy. Just no break, no accelerator, no steering wheel in the car. It just goes. But for we’re really talking about is level four. To answer your question about where are we in this right now, there are level four vehicles being tested in a lot of different geographies, cities around the country and around the world as well. The use cases that we anticipate seeing really break up into two paths. One hearkens back to how you opened our podcast talking about the Uber vehicle. The TNCs are very interested in using autonomous vehicles as a way of substantially reducing the cost structure of those rides. So that will be one use case. Think of a TNC or a taxi today, except there’s no driver. That’s one.

The alternative use case that we’re seeing are small shuttles, six, eight passenger vehicles, which are being used in very specific applications. So you might see them on a college campus to ferry students around, in retirement communities to move seniors around.

Thoko Moyo: Some of that’s already happening, isn’t it?

Mark Fagan: It is. The way in which I describe what’s happening right now is experimentation. There’s lots of different experiments taking place to learn what’s viable, what’s not viable. Both from a technological perspective, and also from a consumer adoption perspective. When I meet people, when I introduce myself as someone who’s interested in autonomous vehicles, I often ask them, “Would you ride in an AV?” And I get very mixed answers.

Thoko Moyo: I don’t know if I would. I get nervous about the idea, but then I also get nervous flying. And I know at some point there’s autopilots so maybe that’s just me.

Mark Fagan: Well, I don’t think it’s just you. I think there’s a fair number of people who feel that way. So these experiments that you’re seeing right now are really an opportunity to learn the technology, to learn the consumer adoption and trust. But it is also an opportunity for policy makers to set the guard rails that will exist as these technologies do commercialize.

Thoko Moyo: Okay. So let me ask you one other question though. We talk about and as you explain it, I can hear and I can certainly imagine the benefits and why this is really advantageous to us, and the future looks good. But I’m assuming that, is it that simple that autonomous vehicles are the way to go and we should do it?

Mark Fagan: It is not at all simple. And how this technology ends up being either a benefit or a cost to society in some part sits with the policy makers themselves. Earlier, you raised the question about congestion and concerns about that. There is not clarity on what will happen if a AVs really become at scale, a key mobility force. And here’s why. You mentioned earlier the cost structure. If you eliminate the driver from a taxi or a TNC and replace it with an autonomous vehicle, you’re basically cutting the cost of that ride in half. At that point, people who are taking transit today or walking or cycling may actually find it more attractive to take an AV ride. So we may be creating rides. And therefore, potentially adding congestion.

The other concern that makes it not so clear that this is the answer is the technology itself. There is a ways to go in terms of ensuring that the safety benefits that we are looking for actually get achieved. For listeners who are interested and have an experiment in their area, it’s well worth just having fun and going in taking a ride. In Providence, Rhode Island, the state of Rhode Island and the city have an AV shuttle. It’s a five mile shuttle. And when I ride it, I find myself having two emotions. On one hand, I am awed by the technology. And in the other hand, I’m horrified at how far the technology still has to go to be developed. So let me start on the positive side. At 3:00 AM, this vehicle is perfectly capable of operating those five miles flawlessly. And to think that you can write a computer program that will take a vehicle, and it will operate for five miles, it will stop at bus stops. It will stop at intersections with a stop sign. It will stop at red lights, is really quite astonishing. But when you ride on it at 10 in the morning, you see something a little bit different. Because it’s in mixed traffic. There’s a lot going on. And there are some elements where it’s fine, it’s just going down the road or pulling off to a stop and it can accommodate that.

Thoko Moyo: And these shuttles are operating in general areas or in a contained setting?

Mark Fagan: It is not a contained setting. It is a on the street corridor. They actually designed it very cleverly because there is no transit on that route today. It serves a lower socioeconomic community. So in addition to testing the AV, they’re now providing a free transit service to the residents in community. But when you ask what are you concerned about, it gets to a four way stop sign. Now think about this for yourself. When you get to a four way stop sign in a car, how do you determine when it’s safe for you to progress? A big part of how you make that determination is you’re looking in the eyes of the other driver. You’re seeing how their car movement is indicating. Are they just creeping because there’s a little bit of space, or are they making a statement to you, “Ma’am, step aside, I’m going through.” The AV has very little capability to understand that level of sophistication. Yet if you really want AVs to be operating everywhere, it’s got to get to that level. So that’s the kind of challenge we see

Thoko Moyo: But isn’t the other side true as well that the point is in the future you will have more AVs on the road, so it’s almost equal. They’ll be making judgements at a four way stop here for AV vehicles. So there has to be some sort of technology that allows them to read whatever signals they are. So you take out the human element, the subjective piece of it, out of the equation.

Mark Fagan: If we fast forwarded let’s say 20 or 30 years, the scenario you described is exactly right. When the majority of vehicles are AVs, these issues will be much smaller. Our challenge is over the next 10 or 20 years, the AVs will be the minority. The majority will be you and me driving. And how that transition takes place is really quite challenging. If we come back to policy, one of the challenges that our policy makers at the municipal level are facing is how do I work that interaction during this transition period?

Thoko Moyo: And would it be fair to also ask that a policy question would be around the rollout of the Avs? Sort of making sure, I don’t know if this even make sense, but the proportions of AVs versus regular people driven cars. Does that factor at all or not really?

Mark Fagan: Well it does in an interesting way. The way in which it’s likely to roll out is in some geographic areas. So here in Boston, maybe the Back Bay would be an area where you would have AV operations taking place. It’s less that the policymaker will dictate there will be this many AVs and this many traditional cars, but I think they can achieve something quite similar to that by determining what the geography is that you’re going to allow the operation to take place in, and how many licenses they issue for AVs.

Thoko Moyo: So when you were describing the shuttle in Rhode Island, and I was trying to imagine you there, and you talked about how you reduced the cost. An obvious question and maybe we’re going to get to it later on, but I should ask this now, that was in my mind is this obviously also starts to creep into questions about the future of jobs. You eliminate someone driving the car, you reduce the cost, but that’s someone out of a job. We talked a little bit about congestion because maybe more people will want to take these rides because it’s cheaper. It just raises the question of what does that do to climate change efforts of you have more congestion, you immediately think more issues around climate change. So tell me a little bit about that.

Mark Fagan: Well, let’s start with the second issue around congestion and climate change, because we have two forces acting on us simultaneously. At least right now, it appears that the autonomous vehicles will also be electric vehicles. If we get to that point, then the carbon issue at least is diminished. In terms of the number of rides, as I was mentioning before, we really aren’t sure how that’s going to play out. You referred earlier to the potential of instead of finding a parking spot for my autonomous vehicle while I run in for a cup of coffee, I just let my autonomous vehicle ride around. We have a term for that. They’re called zombie cars. And our thought is from a policy perspective to be very aggressive in preventing those. We can do that through pricing. Because the autonomous vehicle we track. Everywhere it is at every point in time. We know how many people are in it and the like. So one of the interesting policy questions is how do we use in particular, pricing as a way of trying to incentivize the behavior we want? To address the congestion concern you raised, one thing we’ll do is we will heavily tax zombie cars. So yeah, you want your zombie car to run around? Ten dollars a mile.

Thoko Moyo: So you would do that by, so you’d see this car so you’re tracking it, and you see it driving around, and you levy a charge on that? Or just, so just explain what that price-

Mark Fagan: So what would actually happen is the computer software that runs the AV would be downloaded periodically. And you would determine how many miles it operated with no one in it. And you’d send them a bill. It’s just like the automated toll taking we have today here in town on the Mass pike, same thing. We just send you a bill in the mail. So that is one way. But that’s clearly a stick. More on the carrot side, in order to try and address the congestion concern, a number of policymakers are looking at what incentives they can provide to encourage a shared ride model. It’s most likely that these AVs will be operated by a fleet. Whether that’s Uber or Maven or some new company, don’t know, but it will be fleet operated. It’s highly unlikely in the near term it’s going to be you and me having our own. The fleet is a good step forward. But what you really want is two or three people in the vehicle going from point A to point B. Again, because we know where that vehicle is and how many people are in that vehicle at any point in time, we can actually price the use of our streets to encourage shared rides.

So for example, if you were going to operate a and and it was picking someone up at South Station here in Boston, which is a commuter rail hub, and taking them to the Seaport area, which is a new innovation hub here in town. And it’s a couple of miles, we would love for AVs to solve that last mile problem for people so that they don’t feel they need to drive their car in. And the transit options are limited there. So in that case, we would maybe not even charge you to use our roads in that environment. However, if you were picking someone up at a transit location and taking them to a destination that’s also on that same transit location. So it was literally in parallel, we might tax the heck out of you for doing that. So that’s how policy is trying to come together to think about how we shape this outcome.

Thoko Moyo: So you’ve mentioned a couple of times, you said you can track with a car is you can track who’s in the car, etc. That just raises the question of oh my gosh, we’re being watched. Data privacy. So let’s talk about that because I think that is something that just instinctively people would recoil from. And I’m sure these are discussions that you’ve had when you think about it from a policy perspective.

Mark Fagan: You’re absolutely right. The data issue is one of the most important and one of the most controversial. Here at the Kennedy School, we run a set of policy scrums, which are you can think of as a hackathon or a design sprint. With local governments to help them catalyze their thinking on autonomous vehicles policy. We ran one, it was actually our first, it was here with the city of Boston. And I was facilitating a group of people, and we were talking about how the role of information is utilized. And I must say I’ve facilitated a lot of sessions in my day, but this was one of the most where one of the TNCs who would be an AV provider, and a representative from the regional planning authority were going at it toe-to-toe over, “I want that data.” “You can’t have that data.” “I want that data.” So it is really important.

Thoko Moyo: What were some of the issues? What are the actual concerts? Break it down. So what would I be concerned about? What would the ordinary person on the street be worried about in terms of the data that’s being tracked?

Mark Fagan: So the ordinary person, and I’ll use your phrase on the street because that’s the extreme. To operate the AV, there are a number of outward facing video cameras. So these cameras are literally seeming just like I’m seeing you in this recording. They are seeing what’s happening on the street around them. Now it needs that because we want to see that a pedestrian is walking her bicycle across the street. I need to stop and not hit her. But, it is also picking up you. It’s seeing you-

Thoko Moyo: And I didn’t consent.

Mark Fagan: You didn’t consent. And I would imagine that the vast majority of us feel I have an expectation of privacy when I’m walking on the sidewalk. So, who is actually controls that information? Well the camera is owned by the AV provider. Therefore, you could argue, I think they would argue this is my data. It’s my information. Well, that information probably has a market value to it, right? How that plays out is really in its infancy of really coming to move from a here’s a problem to here as a policy. But there are a number of communities who are starting to work on that issue. I’ll take you now inside the vehicle to you. So we dealt with the person outside on the street. You’re an occupant, a rider. Well, this vehicle knows where you originated, where you terminated. It may have an inward facing camera and actually knows, sees you visually, physically. May realize, “She looks really upset today. Maybe I should push some ads to her for you could use a little massage.”

Thoko Moyo: Oh God no. That’s just my angry face.

Mark Fagan: Right, exactly. How that plays out is yet to be determined. But there is a rationale for having an inward facing camera. Because what happens if there’s a crisis in the vehicle where you have a heart attack or something like that? So there are a lot of issues here. These are more interpersonal issues. These are between you as an and how that information is shared. And my expectation is that the rules of the game there will be fairly restrictive. But I’ll take you to a little bit more mundane, but from a policy perspective, very important role of data. Which is if you are responsible for operating and running the streets of Manhattan, or Chicago, or Kansas city, the data that these AVs are picking up is actually very valuable to you. Because it will tell you everything from at any time of day, what the speed is on every road in your city. It will allow you to know how many cars are backed up at a light at any point in time. It’ll even tell you where there are potholes. So you can imagine that that will be very valuable as the head of DPW in Detroit. Who has that information? Well it’s the AV operator who has it. AV operator could well say, “But that’s my data. If I share that ride level data with you, you’re subject to Freedom of Information Acts, it could easily become public.” We’ve been battling with this issue for a little bit. And in our scrum at Kansas City, I think we came up with what I think at least at the moment seems to be a very viable solution on this problem. Instead of the agent, the government agency asking for the data, just give me this mountain of data. Instead, pose the questions you need answered. And that may be tell me where the signal light timing if it were changed, could improve the flow of traffic-

Thoko Moyo: Got it. So it’s a need to know, but specifically, okay.

Mark Fagan: Precisely. So I’ve wandered around a little bit in this tar pit around the data. It’s a very thorny issue and one we still haven’t gotten answers to yet.

Thoko Moyo: So tell me, it sounds like you’re learning a lot from the scrums that you’re having, the hackathons that you’re having with local governments in the U.S. Tell me a little bit more about this. The goal is you bring together government officials from different parts of the country, and we’re talking just the U.S. for now? Sorry, and they come in. So what is it that you’re hearing, seeing in these conversations? Just telling me a little bit more about your actual work.

Mark Fagan: What the scrums are focused on is they’re trying to catalyze policymaking. And the organizational principle is doing it at a city or state level. So therefore, it can be customized to the specific issues that are relevant for a local community. For example, we did one in Detroit a few weeks ago. And in Detroit, the focus was how can AVs enhance the mobility for senior citizens and the disabled? So that was the umbrella issue we wanted to focus on. And within that, we wanted to think about specific policies that the city and the state could craft that would help foster that. So, we looked at issues around what policies should dictate the design of these vehicles to facilitate seniors being able to get in and out of them and use them. We looked at policies that could facilitate the mobility for seniors to medical facilities. So that’s kind of the issues we do. In terms of who attends, it has ranged pretty broadly. And the case of Detroit where they’ve spent a lot of time thinking about autonomous vehicles. We had the public sector, we had the private sector. So providers of equipment and operations as well as some advocacy groups. And it’s bringing together-

Thoko Moyo: And they’re representing people like me?

Mark Fagan: Exactly. Exactly. It’s bringing together the key stakeholders to have a facilitated discussion that is how our process works now. From the perspective of our hosts, it’s a way for them to get a diverse group of people together and try and herd the cats as it were, and get some direction on how they want to move forward.

Thoko Moyo: So this is at the local government level. Is there a role for federal policy making?

Mark Fagan: So I’m smiling as you say that, because the answer to that is certainly yes, but. The way in which we regulate and manage vehicles today, traditional vehicles that you and I drive, is that the federal government sets safety standards. And then the auto producers certify to meeting those standards. At this point, we don’t have that. There’ve been a couple of attempts at federal-based regulation to get there, but they haven’t made their way through Congress. So what is happening is in lieu of that, you’re seeing a lot of action at the state and local level. In some ways as a policy maker, but also as a citizen, I’m fine with that. Because right now, the testing that’s taking place with AVs is at a local level. We’re testing in Boston, they were testing in Chandler, Arizona, testing in Providence, Rhode Island. If you’re testing locally, you ought to have a fair amount of control about how that testing takes place.

Thoko Moyo: So let’s take it out of the U.S., and we’re running out of time. I probably only have a few more minutes with you. But let’s end with looking at what’s happening elsewhere in the world? Where have self driving vehicles taken off? I have in mind China, what’s happening there?

Mark Fagan: Well, I’ll get to China in one second. But I want to actually start in Singapore. Because Singapore was one of the first geographies to really test and start to operate AVs. It turns out that one of our local company, an MIT Media Lab spinoff called nuTonomy, developed a technology that they took to Singapore. And the work in Singapore has been very successful. The efforts are really in a geo fenced area, so it’s controlled as I was suggesting before. But they’re up and operating and it’s quite exciting.

Now we’ll move to China. China is making real strides on this as well. I suspect you we’ll continue to see some activity in this area, very robust growth. But it isn’t just there, it’s happening around the globe. There are a number of experiments in Europe, in the mid East, and in Latin America as well. We did a couple of scrums in Buenos Aires. Interestingly, just to tie back to your question or your observation about jobs a little while ago. The scrum in Buenos Aires focused on the economic development opportunities that arise because of this new technology. And you really need to think about that. Because as you said earlier, jobs are a big issue. If I just take us back to the U.S. for one closing comment, there are 4 million people in the United States who earn a living through driving. Autonomous vehicles represent a dramatic threat to those individuals. And how that is managed, and the future of this transition, and the policies that go along with it will really dictate how not only it improves mobility, but how it affects the broader community as well.

Thoko Moyo: But I can’t let you end on something that’s a downer. So you have to say either how you see it playing out, how these millions that will be affected. What are some of the options and solutions around that? Or just give me a good news story from elsewhere in the world.

Mark Fagan: All right, I’ll do both for you. The reason that the 4 million people are important but not a downer is that the transition to autonomous vehicles will take a very long time. As we referred to before, this is decades away to be a substantial portion of the fleet. And I think we’ll have enough time to effectively transition.

Thoko Moyo: And I guess it plays into conversations, broader conversations about the future of jobs, future work, which is training people for the jobs of the future, which may not include the current jobs that we have. So it’s part of that bigger conversation-

Mark Fagan: Exactly. I want to close coming back to actually a number I threw out at the very beginning. In the United States, 30,000 people die every year as a result of on the road crashes. Autonomous vehicles offer the promise, not proven yet, but offer the promise that that number substantially gets reduced. That’s what motivating I think a lot of the technological development that you’re seeing, and it’s also motivating us here at the Kennedy School to think about the policies that will bring this potential safer mode of transportation into common use.

Thoko Moyo: So are you saying that when you get into a driverless car, you are fine, you are happy? What’s your reaction? Because you mentioned earlier on that you ask people and you get mixed reactions. Where do you sit in the debate?

Mark Fagan: I’m ready for it.

Thoko Moyo: You’re ready for it?

Mark Fagan: Yep.

Thoko Moyo: Thank you so, so much Mark for coming in to talk to us. This has been really interesting, and all the best with your work at the Autonomous Vehicles Initiative at the Taubman Center at the Kennedy School.

Mark Fagan: Thank you very much.