WHEN IT COMES to machines, humans are eternal optimistic that the next great advance will make them truly easy to use. In computers, this means graphical interfaces, then voice, and now artificial intelligence. Donald Norman has been arguing for decades that it takes more than optimism and a new technology to make a usable interface.

Norman first came to prominence with his 1988 book The Design of Everyday Things. Quite a bit of the thinking in that book was, famously, inspired by a visit to England's Cambridge University, where the everyday things such as light switches and water taps drove him crazy. But his involvement with usability started earlier with two experiences: using UNIX, and studying the Three-Mile Island nuclear power plant accident.

The frustrations of UNIX inspired his 1981 paper The Trouble with UNIX. Three Mile Island, "led me to realise that the errors were not because of human error but bad design," he said. Working for NASA, he began developing design principles that would eliminate those problems. Then came England, where, he says, "I discovered that all the principles I was developing for airplanes and power plants applied to water faucets and light switches."

And also to software and the computers that run it. The attitude in software design up until then, he says, was: "This is for us experts, and if you mess up you shouldn't be using it." One of his favourite lines from a user manual of the day was, "Even experienced users have been known to make this mistake."

This was when every program had a different set of commands; the Mac changed all that by giving developers a toolbox that made it easier to follow consistent rules than to create their own.

Norman worked at Apple as a "user experience architect" from 1993 to 1997, the years when usability was arriving on every software company's agenda; they were all setting up user testing labs, and almost everyone you spoke to in the relatively new discipline of human-computer interaction cited Norman as an inspiration.

Norman's latest book, The Design of Future Things contains an apparent reversal. Where, in 1988, he was arguing that technology needed to be designed to make it easier for humans, today he argues that nonetheless, since humans are more adaptable than machines, if we are to work successfully with the much more complex cars, appliances, and other devices of the future we are the ones who will have to change, at least to some extent. He especially takes issue with the prevailing notion that smarter computers will inevitably be easier for us to interact with.

Instead, he says, "I'm thinking that people are from Earth and machines are from outer space." After all, "Machines are logical and we are not." We don't share life experience with machines, nor do they have empathy.

"I don't go for the singularity arguments," he adds, "but I do worry about the hybrid, where we're going to add more and more prostheses – electronic, then nano, eventually biological – into ourselves, and so we'll be a different species with perfect memory, better eyesight, better hearing, stronger." What happens, he asks, when this year's models of arithmetic unit and memory enhancers are incompatible with the models of five years ago?

Norman, who has been in and out of academia and industry throughout his career, is currently teaching at Northwestern University, lecturing MBA students about the role of design in their companies and products. We are talking the day after DARPA announced the winners of the Urban Challenge, its latest contest for a self-driving car – exactly the kind of future thing he's worrying about.

"We have to be careful about thinking that cars can drive themselves," he says. "But I do believe that in some future day cars will drive themselves and be better, safer, and more relaxing." People who still really want to drive, he says, can do so on a track, just like you do now if you want to ride a horse. µ