Interview with Paul Green, Part Two

In this second half of my interview with Paul Green, we discuss teaching Human-Computer Interaction at the University of Michigan and the history of the Association for Computing Machinery (ACM) Special Interest Group on Computer-Human Interaction (SIGCHI).

Teaching HCI at UM

When did you become interested in User Experience?
I got into it pretty much as I was finishing my PhD, and immediately thereafter. The reason was that I was looking ahead and I knew the future was cars would be connected to computing and complicated interfaces.

The first report I wrote here was a pretty straightforward topic: an assessment on stalk controls—those things that stick out of the steering column, like turn signals and windshield wipers. I said, “That’s not the future. The future is that we’re going to have touch screens in cars.” So I wrote, “Here’s a hierarchical menu system inside a car that does all this stuff.”

I remember the report was funded by the then-Motor Vehicle Manufacturers Association (it’s gone through a bunch of iterations and it’s now the Alliance of Automotive Manufacturers). It’s the car-makers. They said to me, “Take this out—no one would ever do touchscreens. No, it’s sticks.”

But that got me, in 1982, to teach a Human-Computer Interaction course, because I knew I needed to know that material. That material would be important for motor vehicles. I read the literature, taught classes on it, and basically learned what was out there so I could do research that related to that topic.

If you look at cars right now, what have they got in them? They’re all computer systems, and people complain about the user interface.

I’m still doing work on that topic, like predicting how long it will take people to do basic things. Then, there’s the basic stuff we’ve done for years, that we’ve written standards on, and that we’re still doing.

Probably by next year we’ll have some much better data for predictions. That is, we’ll be able to predict how long it will take people to do things as well as an experiment can. So, if you compare the correlation between experimental data and the predictions in any experiment, the predictions will be as good as data, which means there will essentially be no unexplained variance. It’s part of this push to make interface analysis much more like other aspects of engineering.

Is that portable beyond touchscreen in a car? Such as for mobile phones?
I’d say that there are certain unique things about thumbing where the time data are going to be different, but the analysis procedures are going to be the same. Basically, it’s like the keystroke level model, but adapted for automotive interfaces.

You taught an HCI class is in the Industrial and Operations Engineering Program?
Yes. I still teach it—it’s still going on.

Has interest grown in your class over the years?
I’d say it’s stabilized, because now there’s SI (University of Michigan School of Information).

Also, I look for a particular kind of student. That’s often not your typical engineering student, because with engineering students, everything is taken as plug and chug. “Show me the equation, and I’ll plug it in and get the correct answer.”

As we know in the real world, especially for interface design, that’s just not what’s important. We’ve got to teach them how to do keystroke level modeling—GOMS—calculation procedures, but the first problem usually is “what’s the problem?” It’s not a problem where you know how to calculate the answer.

Furthermore, the answer to the problem doesn’t exist in a single chapter in a textbook. You use lots of ideas and methods. You figure out how to solve this problem, see what the issue is, go get a bunch of people, have them try it, learn how to process their feedback, and then offer a suggestion for a design fix. They’re used to situations where they’re told what the problem is and they solve it. Well, the first part is to figure out what it is.

Let me give you a concrete example. The next assignment they’re going to do in class is that I’m going to talk to them about the literature of how to assess a display, how to measure lighting characteristics, the different ways to determine legibility, etc, and how people have done assessments of them. I’ll talk about some journal articles where people measure them, about reading tasks, and the percent that they read correctly. Now the charge to them is “Go find a computer display you’ve experienced that’s not legible. Conduct an experiment, make some change to the display somehow, figure out how to do it, and then conduct an experiment to show that it’s better. And write a recommendation.”

They ask, “How do I find legibility problems?” Well, go experience them!

To them that’s hard because I didn’t tell them what the problem is. They have to go find the problems. I say, “Well, think about your day, and as you go through the day, every time you encounter a computer display, is it hard to read? What was difficult about it?”

“How do I do that?” they ask.

So they’re not used to that kind of open ended task where you have to think about it. They ask, “What experimental procedure should I use?” I’ve told them what the literature is, and given them some sense of how different procedures are applicable in different contexts. They need to look at the context and think about what makes sense, what’s feasible, what’s going to be compelling evidence, and then consider that they can’t rewrite the software. They’re going to have to think of some way to fake it. They’re not used to not being told exactly what to do.

It’s the kind of thing you experience all the time—you can’t do this, you don’t have a tool for that. Well, what other ways can you come up with an answer? What does the literature say? Well, there’s more literature than you can ever read. How do you find the key things that are appropriate? They’re just not used to that open ended-ness. But if they don’t know how to handle that open ended-ness, they’re going to be completely unprepared for industry.

The Early Days of ACM SIGCHI

What can you tell me about local figures in the HCI field?
I can think of one person in particular. A guy named Greg Marks*. I’ve known Greg for years—I knew him when I was a student here, and Greg is still active as a professional. Years and years ago, he was head of an ACM group called SIGSOC (Special Interact Group on Social Aspects of Computing). That group was active here on campus, and so there was a point in time when Greg was wondering, “As computing changes, what should our future be?” Some of what Greg did eventually morphed into SIGCHI, which is not well known.

The first real CHI meeting was in ’82, and what happened was that the Washington ACM chapter had a chapter meeting**, and they invited Al Chapanis as the speaker. Chapanis had been the president of the HFES (Human Factors and Ergonomics Society), and he was really well known to Human Factors people. He was doing a lot of the early work on communication and related topics, as they pertain to computer systems. As I recall, the original plan was to have a chapter meeting, and they thought 200 people would show up. Instead, they invited Chapanis, and he connected with the Human Factors community, and something like 600 people showed up to this chapter meeting.

After the chapter meeting, people said, “Oh, I think we’ve discovered something. A connection between the ACM people and the Human Factors people. What should we make of this?” There was a meeting afterwards, and I was involved, and Bob Williges*** was involved. Williges was a senior Human Factors person and I’m not sure if he had been HFES president or was president later, or executive council, but he was that kind of person. There was a lot of discussion about continuing this meeting in a collaborative manner, but what happened was that ACM had the resources, and they ran with it, and formed SIGCHI****.

Now, what happened is the false rumor in the Human Factors community that we had the opportunity to dominate in Human-Computer Interaction and we blew it. The answer is, “No.” ACM had the resources, and they ran with it. We didn’t have them. It’s not that we lost anything, and CHI’s done very well.

* Information technology researcher at the University of Michigan, and involved in planning one of the first conferences focused on Human-Computer Interaction, which took place in Ann Arbor, Michigan.

** Specifically, the Software Psychology Society held a conference in Gaithersburg, Maryland. At that time, it was clear that the ACM SIG board would approve SIGSOC renaming itself SIGCHI, and the Gaithersburg conference was one of the first gatherings for the emerging human-computer interaction community.

*** Professor Emeritus of Industrial and Systems Engineering at Virginia Tech.

**** Greg Marks was involved in discussions to change SIGSOC to an organization focused on Human-Computer Interaction around 1980. As the Gaithersburg conference was planned, it looked like ACM would approve the change to SIGCHI, and it was at this conference that attendees settled on the name “SIGCHI.” ACM subsequently approved the repurposing of the special interest group.

The Future of the Field
What do you see in the future of Human Factors as a field, both locally and internationally?
In terms of the field, it’s pretty clear that medical issues is the future. That’s the growth area, and I don’t know what else is going to grow, but people are getting older. Medical expenses are becoming a greater proportion of everybody’s income, and there are just a lot of Human Factors problems in medicine that need to be resolved. There’s this tremendous opportunity out there.

Locally, it’s really hard to predict what the scene is going to be, because there are so many independent pieces. There’s no organization, there’s no place, there’s nothing that serves as a sort of central rallying point. Okay, there are medical people doing Human Factors at the hospital, and there’s a Human Factors community, but it’s really a student community. There’s not a community of professionals in the normal sense.

Then there are the UXPA and CHI groups. There are all these people doing related things, but there’s no central unit that everybody belongs to. Quite frankly, it’s very difficult for those kinds of organizations to function anymore. Look at HFES. A lot of our local chapters are very weak. The only things that succeed are the student chapters.

It’s all because people are just busy. A lot of what used to happen was at local events, but now people say, “I don’t have the time to drive there, so I’ll just email somebody.” A lot of the face to face connections have fallen. But it’s true of many organizations and not just HFES. It’s just too difficult now, because the workload has increased and because the society has changed, and that’s unfortunate.

The other thing is that a lot of people that are younger don’t understand the value of the face to face connections that we used to have. They have to get pushed very hard to start to build them, and once they start to build them, they say, “Okay, I get it. I can’t just email somebody. I’ve got to go talk to them, and not just on the phone. I’ve got to talk face to face.”

Paul Green continues to teach and do research at the University of Michigan.