Interview with Paul Green on Usability-in-Driving

The following interview was recorded on April 19, 2005.

Dr. Paul Green hosted MIUPA’s December 2004 meeting at the University of Michigan Transportation Research Institute. Besides describing UMTRI’s mission and projects, he took us upstairs to let us try the driving simulator. Tim Keirnan caught up with Dr. Green four months later to discuss more usability-in-driving issues, current projects at UMTRI, and the value of side-impact airbags in today’s new cars.

Interviewer: Timothy Keirnan

Simulator in profile

Tim: It’s been about 4 months since your presentation and demonstration at UMTRI. Is there anything you’d care to add to what you told us that night?

Paul Green: Yes, we have a couple things going on currently. One is the workload studies that will kick off within the next month or so. So that’s one development.

The second development is an interesting use for the simulator that may happen in the near future, and that is to use it more as a theater. From time to time we have visitors come in and there are some very sophisticated multi-screen analysis systems we’ve developed here at UMTRI for looking at field operational tests. You take a vehicle out on the road, you put a whole bunch of cameras in it, you record hundreds of engineering variables, and you want to show people what’s happening. However, there is so much going on—the drivers’ facial expressions and where they are looking, the forward scene, a radar view of the targets ahead, strip charts of steering wheel angle, accelerator angle, and speed, etc., etc., that all of it cannot be presented simultaneously on a single screen. And we have a group of visitors in the facility who want to watch the field operational tests—how do we show them all what’s going on?

So what we decided to do was hook up some computers to the projectors in the sim lab and they’ll bring all that data in and project them on the walls for the visitors to easily see.

Tim: Wow.

Paul Green with the Simulator

Paul Green: We had a little test run today and it actually works out ok. The projectors obviously don’t have the resolution you’d get with a good monitor, but it’s impossible to crowd a lot of people around two monitors in someone’s office.

Tim: Sure. So this would allow people to see what’s being recorded in your field test as it happens; there must be so much data flooding in. This must present the data in a way that visitors can see what’s going on in the moment.

Paul Green: Right, and it’s the view that the analyst sees. You get the data, but like I said if you were to try the standard single projector situation you’d have in most conference rooms, it’s not enough. You’d need two projectors. So hey, we’ve got a room hooked up for three, for four! So we’ll use that.

Tim: Excellent! Your new driving simulator reminds me of that 360 degree movie in the round at Disney’s Epcot, where every wall is a movie. Very immersive.
The first thing you talked about, the workload studies, does that mean the same as cognitive load? Where you’re driving with things distracting you in addition to the regular tasks of driving.

Paul Green: Yes.

Tim: Cool. You know, your MIUPA meeting at UMTRI was one of the best meetings we had last year, between your presentation of what UMTRI does and your letting us drive the driving simulator afterwards. Would you like to tell the readers what UMTRI does in case they don’t know, and how usability fits into the picture of your research there?

Paul Green: Sure. UMTRI’s mission is to carry out research on driving; to understand how driving occurs and to try to make driving safer. So we look at questions relating to headlights and what people can see, we look at crash data to try and find where the problems are. There are people here worried about impact tolerance so they take dummies and put them on sleds and see what happens. There were some discussions here about what happens to a soldier in a Humvee when it rolls over. UMTRI is very interested in understanding how electronic systems, crash avoidance systems, and so forth, can make driving safer. So it’s a lot of different issues involving psychology, mechanical engineering, electrical engineering, civil engineering, biomechanics, public health, and medicine. And trying to pull together information from all those disciplines to say “here’s the kinds of things we need to do to make cars and driving safer”.

What’s important to realize is we’re not sitting back thinking “Gee, what would we like to do?” We are competing against other universities and companies for federal government and industry dollars to do projects, some of which they lay out and some of which we lay out. So it’s not like we’re sitting back thinking about what we want to do, we have to go out and be the cheapest, be the best, et cetera, just like any company does. That’s the first part.

You asked about usability. I think it’s pretty commonly accepted that usability and safety are tied together. If people have trouble getting in and out of a car in a hurry, if they can’t figure out what the controls and displays are, if they don’t understand warning messages, there are going to be safety issues. So our experience, which agrees with the perspective a lot of people have, is that if it’s easy to use then it’s probably safer to use.

Tim: I’ve read two articles recently about distracted driving and I’m amazed at the number of drivers who really don’t think about their responsibility to their safety and the safety of fellow drivers on our roads when it comes to cell phone use.

Paul Green: Absolutely.

Tim: Is there a definitive study to which I could refer people about how unwise it is to talk on a phone while driving a vehicle? So many people think “I’ve gotten away with it this long so I must be great at multitasking” but their eyes glaze over when I tell them about cognitive loading and response times and stuff like that.

Paul Green: Sure. The article cited most often is a study by Redelmeier & Tibshirani that appeared in the New England Journal of Medicine in 1997. It’s cited commonly on the Web. If you want I can look for it. A lot of people have done stuff, but their stuff is the definitive work because it uses real crash data. Other research has followed, but the basic point is the risk ratio goes up.

Tim: Right, and I was curious to ask you something related. I bet there was a similar issue when car music systems became popular. Because I’m sure there was a time when it was all a “motorist” could do to get their engine started.

Paul Green: Right.

Tim: And to get to where they needed to go—maybe only 4 miles away—without having a breakdown or a tire puncture. And eventually—I don’t know what point in automotive history it happened—they started getting these technological aids to make the driving experience more fun. So with radios you could have some entertainment going on while you were driving. From my own experience, sometimes you’re really into the music or the newscast or the audio book and you can “zone out” without even talking with someone else.

Paul Green: Right. Well, I think there are some interesting differences, though, between the car radio problem and the cell phone problem.

First, listening to a radio is passive. You’re just there and it’s just playing. It can be entertaining but it doesn’t often have too much information content except for the news. But it doesn’t require any action from the driver.

That’s not true of a phone conversation, where the other person’s expecting you to respond, expecting you to say things. And they have no awareness of your driving situation, so they act like it’s a normal conversation, as if you’re undistracted and solely attentive to the conversation. So as a consequence, there’s a great deal of additional demand there.

Also, keep in mind that, when automotive radios were being introduced, human factors did not exist as a formal discipline, so there were no experts, procedures, or equipment for making a professional assessment on possible radio-induced distraction.

Tim: I recall from our MIUPA meeting that you said if you have a passenger in your car, it’s ok for you to converse because they are a kind of “copilot.”

Paul Green: Right, but as a person on the phone with someone who is obviously not in the same car, if you look which way the traffic is going and they ask you a question and you don’t respond instantly, they can say “You’re not responding to my question, I asked you this, why aren’t you responding?” And they’ll get impatient or think your silence means something you did not intend. They have no idea there’s a car coming at you and you’re trying to avoid it. It is a very different situation. But if you talk with a passenger in your car, they know when you are facing a tense situation in traffic and will either be quiet or provide help.

Tim: I’m really disturbed by people who say “Oh, I have a hands-free cell phone kit, so I’m not a danger to myself and others.” I tell them “But you don’t have a brain-free cell phone kit, you haven’t altered your risk profile one bit.” Sometimes I’ll tell someone who calls me while driving “Look, I care about you, therefore I can’t talk to you right now because you’re making me party to your distracted driving. Call me when you hit a rest stop or pull into a parking lot.” Am I being too harsh about this?

Paul Green: Not at all. A colleague of mine recently conducted a major on-the-road study in which drivers were given test vehicles to use as their own personal vehicle for four weeks. The vehicles had a lane departure warning system. If they drifted from their lane and did not operate the turn signal to indicate a lane change, they received an auditory alert. Several drivers commented that the system made them aware how poorly they were driving when they were on the phone, something they did not realize because they were focused on the call. That is the problem. Drivers do not know what they are missing.

Tim: Here’s something more on the distraction situation. I’ve had some momentary lapses in driving myself, ahem, that could have led to a problem. I was trying to operate a radio or music player in the car and realized how important it is to have a tactile sense of the controls so you don’t have to look at the radio or whatever. Keeping the eyes on the road is so obviously important, it shouldn’t need mentioning, but so many car devices require that we take our eyes off the road.

Paul Green: Right.

Tim: I’ve noticed some systems recently whose controls fit on the steering wheel and I have to think that might be a good idea?

UMTRI simulator\'s \

Paul Green: Yes, that can help unless it causes you to think it’s easier to use the device so that you use it much more than you currently do, which is still a distraction. So you reduce the risk per operation while the car is moving, but you increase the exposure to the risk!

Tim: [Laughs]

Paul Green: And maybe get nowhere, statistically speaking! But we don’t really know that yet as a general rule, I’m just saying we need to be careful with these devices.

Tim: Sounds to me like, between the cell phone studies you’ve mentioned and the stuff we’ve seen since their publishing, radio use is not anywhere near the level of distraction that cell phone use is.

Paul Green: No, but it’s surprisingly up there.

Tim: Is it really?

Paul Green: Yes, and it really depends on the task. From what I’ve seen, the “insert the CD” task is actually fairly demanding. It’s “find the CD, remove it from the jewel case, handle it very gingerly, insert it in the slot—that is a pretty fine precision task—then to whatever you need to do to play it, go to track 6 or whatever”. It can be distracting, but it’s not as frequent as phone calls.

Tim: Tell me about projects UMTRI has done, either in the past or more recently, that made you most proud.

Paul Green: Three or four years back, we did some work on the visual occlusion method that we’re going to do a lot more of. There are a lot of ways you can assess the workload of driving. You can look at lane departures and steering wheel reversals and standard deviation of lane position. Subjective measures, physiological measures, the list goes on and on. And there’s no one measure that really stands out.

For a lot of them, the logical chain between what the performance measure shows and what the behavior is—in particular, crashes—involve a long chain. Let me give you one example.

One indicator of stress is galvanic skin response, which is a way to measure the conductivity of the skin, and when you sweat, the conductivity changes. That’s an indicator of stress. But if an event occurs that leads to stress, it is only sometime later that you can measure the presence of sweat on the surface of the skin. And the linkage between difficult driving situations, whatever they might be, or difficulty in driving, and the physiological chain that results in a change in galvanic skin response is non-trivial. “This happens, which causes the neurotransmitter to change, which circulates the blood that causes the skin…”, et cetera, et cetera. It’s a long, logical chain and, when you explain it to people, their eyes glaze over.

However, it’s pretty obvious that in order to drive you must see, and so therefore the way to measure the difficulty of a driving situation is simply to ask people to drive the simulator and close their eyes whenever they can. Therefore, the more they close their eyes, the easier the driving situation. So there’s a very simple link between the demand of driving and whether or not you can close your eyes.

It’s not trivial to measure eye closure directly, so what we do is put people in the simulator and give them a button to press. Each time they press the button, they can see the road scene for half a second, which is a typical eye glance duration. We ask people to drive different kinds of driving situations and, whenever they think they need to see the road, they press the button, see the road for half a second, then the screens go to grey. When they need to see some more, they press the button again. It’s simple way to measure, on a second-by-second basis, the demand of driving some road segments. It’s amazingly sensitive.

Tim: You know, what you said earlier almost sounds like a polygraph machine where they can measure skin responses and other nervous system functions based on stimuli.

Paul Green: Right, but in our visual occlusion case it’s “You need to see? Press the button whenever you need to see.”

Tim: That’s probably open to less interpretation.

Paul Green: Yes, it’s pretty simple, pretty straightforward, and easy to do. It doesn’t take a lot of fancy hardware, and people who are not sophisticated with human performance measures can understand it.

So, we’ve done this to look at the demand of curves as a function of the car’s position on a curve, and we’ve started using it to drive in fog and see how sight distance varies, etc. So this is a line of research we’re going to pursue and hopefully get some funding for in the next decade to try and characterize the demand of a wide range of driving situations.

The nice thing about it is you can look at some system on the vehicle that’s going to promote safety. To do that, you need to know the characteristics of the driving situation. If you’re designing a road, you need to know how difficult this road design is going to be to drive, so our method helps us quantify that. It has a lot of uses.

Tim: Sounds very worthwhile. I wish you plenty of customers who’ll want to pursue this kind of research method.

Paul Green: Right, I know what I want to do, now it’s like “Ok, who’s going to pay for it?”

Tim: You just mentioned road design. It’s occurred to me that I as a driver have concerns not just with the usability of a car’s controls, but I’m also concerned about the design of roads and intersections and signs. Those variables are outside my car and I can’t control them.

Paul Green: Right.

UMTRI simulator with traffic on the screen

Tim: Driving usability isn’t just about a car’s cockpit, in other words. Does UMTRI research issues with the usability of roads, intersections, and signage?

Paul Green: We’ve done work on signs. We haven’t done too much on roads because we tend to be vehicle-centric, due to “hey we’re in the Detroit area and motor vehicles are pretty important around here.” Now, having said that, that is beginning to shift and we’re beginning to get more civil engineers here at UMTRI. There are some major projects on intersection crashes in which I expect UMTRI will be involved.

So it’s gone from not doing too much on roads to starting to do more. In part, it’s because we need much more of an integrated view of the vehicle, the road, and the driver.

Tim: Yeah, because it is a system, isn’t it?

Paul Green: That’s right.

Tim: One last question. This is kind of personal yet I think many drivers are in the same situation. What’s your opinion on the value of side-impact airbags in newer cars? I drive an 8-year old subcompact coupe that’s very fun, very cheap to maintain, and very paid off. No rust, and I prefer the handling of a small car that zooms about in a very controlled manner as opposed to some ponderous car or truck. I know I’ve endured some emergency situations where that superb handling helped me out of a very dangerous event. And I like not burning a lot of gasoline for both environmental and wallet reasons.

Paul Green: Ok.

Tim: But I’ve noticed some press in the past couple years about side impact air bags becoming more common on new cars, at least as an option, and there are crash tests people can look at when researching a new car that show how safe the car is if hit in the side. Is there a way to quantify how much safer having side airbags is to a driver of a small car, and the odds that if I’m hit by some SUV, it’ll be a side impact? Life is great without a car payment, and I don’t know if side airbags are truly a dramatic improvement or not.

Paul Green: Well, they are…

Tim: Darn.

Paul Green: …and what you’re asking for has been suggested in many quarters and has gotten manufacturer resistance. What you’d like to be able to do is get on a website and know—based on a variety of tests—what the aggregate safety level is of a vehicle you might buy.

Tim: Right!

Paul Green: And there’s been pressure not to do it, the reason they argue being “it’s not going to be perfect, it depends on the crash situation” and that’s true. Depending on where you drive…urban situations versus rural…you’ll be exposed to different probabilities of different crash types. So it’s kind of hard to predict overall.

But to say, “it’s hard and therefore we won’t do it” is not a convincing argument. There is an approach to do this in Europe, and so I think eventually it will happen.
On the US DOT site there is crash data for new vehicles for a variety of crash scenarios—frontals, side impacts, and the like. So you can see how many stars or what the rating is for each vehicle. But your question is how do you combine all those numbers? And it’s difficult. I have seen attempts to do it but it has a long way to go.

It is true that the bigger the vehicle—meaning the more mass it has—the better it’s going do in a crash because, basically, it’s going to bash the other vehicle. But mass alone is not the only consideration. You’ll find a lot of smaller vehicles that will do pretty well in a crash. The issue is not only are they more massive, but how tall are they? The more massive vehicles may be able to batter their way through one kind of crash, but if they are tall and narrow, they are much more likely to have rollovers. And rollovers are often fatal.

Tim: Even if you’re wearing a seatbelt?

Paul Green: Yes. The vehicles getting the most attention in the press are SUVs, which must comply with government standards for trucks, not passenger cars. The truck standard for roof crush is much more lenient than the standard for passenger cars. So what happens is that in a crash, the tall narrow SUVs roll over, and while rolling, the roof crushes and breaks the necks of the occupants.

Tim: Wow…well, I don’t want that.

Paul Green: Right, and I’ve heard people tell me this—although I haven’t tested for it—that under certain circumstances, if you had a subcompact t-bone a full-size pickup truck, it could roll the pickup. But the subcompact, while bashed, would not kill its driver—some injuries certainly, but the guy in the pickup would be dead.

Tim: Wow..

Paul Green: But you’d never think in a crash between a subcompact and a pickup that the subcompact would come out better.

Tim: Yeah, how could it “beat” the pickup truck in safety terms?

Paul Green: Right, but if it hits the pickup truck in the side and causes it to roll, the driver of the pickup is dead.

Tim: Hmmm. I have two routes to get to work. If I’m still driving my small fun car while I save up for a fun small car with side airbags, my hunch is to stick with the divided highway route to avoid risk of head-ons and that sort of thing.

Paul Green: Yes, head-ons are often fatal. And by and large, the frequency of crashes per mile per vehicle on an expressway are much lower than on less controlled roads.

Tim: I can’t imagine there’d be many side-impact crashes on the divided highway.

Paul Green: Yes, the big plug for the expressways is that, not only do they move people faster, but they drive the crash rate way down.

Tim: Hmmm. Well thanks so much for spending time with me on this interview and for presenting that wonderful end-of-year meeting back in December for MIUPA at UMTRI. I had a great time there and I know the other attendees did too because I hear it mentioned four months later.

Paul Green: It was a great evening, I agree. My pleasure and drive safely.

Note to our readers: The visual occlusion approach was developed by John Senders in the mid-50s. The most commonly cited reference to Sender’s landmark work on visual occlusion is:

Senders, J.W., Kristofferson, A.B., Levison, W.H., Dietrich, C.W., and Ward, J.L. (1967b). The Attentional Demand of Automobile Driving, Highway Research Record #195, 15-33.

Related posts:

  1. MiUPA 2004 Events
  2. Interview with Tom Brinck on Focus Groups and Usability Testing
  3. Interview with Josephine Scott on the Usability of Voting Systems

Tags: , , , ,

Comments are closed.