Human face of the human interface

IN THE LAST 30 years, humans have changed less than computers have. In a 1985 experiment, John M. Carroll had a research assistant simulate an intelligent help system.

"We found that everybody attributed a theory of mind to the software agent, and it was not correct. They made inferences that would be perfectly plausible for you or me helping each other." Yet designers are still trying. "The latest is in human-robot interaction. I know I sound like an old fart, but when you look at the papers in HRI I find them laughable in that they're worrying about how human society will accommodate robots, and then I see what they mean by a robot. It's like worrying how you will accommodate a vacuum cleaner."

That you can now get music out of an iPhone without reading a manual is partly acculturation - but also the result of hard work on the part of researchers into human factors. Carroll, who founded the User Interface Institute at IBM's Watson research lab in 1984 and is now professor of information sciences and technology at Penn State University, was a founder of the field and is author of more than 300 papers and 14 books.

Much has improved. The Web, for example, is "in many ways so well crafted, so good for designing easy software." Even so, "There's still a lot of subtlety - and not so subtle - missing. My pet issue was always to focus on the agency of human beings. I take it very seriously that people are active, trying to figure things out, trying to get control, trying to regulate themselves and their environment. It's not unique to me - it's a common vision in psychology - but it's still not what people think of even now when they think in this vernacular way about 'easy to use'. People conflate appropriate for humans with simple. It's like an 80 percent solution." Carroll has spent his working life in computer science, but says he has the soul of a psychologist. "I've been in human-computer interaction my whole career, but I don't think I ever became a computer scientist," he says. "I was the head of a computer science department [at Virginia Tech] for five years and never felt like I was one of them."

In terms of remaining sceptical, he adds, it may be good to stay something of an outsider. Too often an "intuitive" interface merely means the thing the speaker is used to. This is a particular problem when it comes to designing supposedly "intelligent" interfaces: humans are a moving target.

"Part of the nature of human beings is to construct a theory of mind of anything that's remotely intelligent," he says. "I think people think of help systems and user interface agents by constructing a theory of mind, and the problem is that if the theory of mind is very simple and boring then it might actually be useful - predictable, understandable. But if you try to get beyond that, and especially if you lie - and by lying I mean cases where designers have gone to some effort to project intelligence that their software does not have in any serious sense - it may have it in an idealistic sense but it doesn't do anything intelligent, creative, human - you get a breakdown."

Carroll still works on the themes that have pervaded his career: "Concerns about people being active and agentive, and needing error support, help in sense-making, getting better feedback - these are still important themes in HCI." Where he's changed is to move a lot more toward collaborative systems, with a particular interest in local civic groups. At Virginia, he was a founder of the Blacksburg Electronic Viillage, which "through sheer luck was the first major Web-based community network." At Penn State, he is extending this work into improving the computer interface that supports community groups. "The problem in the civic sectors is that the infrastructures are so bad." µ