(Visitor David Gilmore's response to Bryan Oakley's post)
I find my reactions to be mostly in-line with Brian Hayes' review, rather than IBM User Interface Group's hype, but neither seems to have any data from real users to guide their evaluations - though Brian of course is more of a real user than the design team.
The design of IBM's RealThings seems to raise important questions. If we ignore the fact that they break their own guidelines (e.g. using minimal screen real estate) we are still left with the question about whether mimicing real world objects should be expected to enable people to be more effectively at doing real work?
The design guidelines seem to ignore traditional views of HCI which would address the number of keystrokes (e.g. GOMS-type analyses), more recent cognitive approaches which ask how the user is meant to know something (e.g. Cognitive Walkthrough), and more radical approaches which address the context of user's tasks and goals (e.g. activity theory) - all in favour of fancy graphical design.
As a psychologist concerned with evaluations - I would like to predict that IBM's design will score highly on evaluations based around impression, preference and first reactions, but very poorly on task performance and long term satisfaction. Their online questionnaire is one of the worst HCI questionnaires I have ever seen with almost nothing about ease of use, transparency, ease of learning, etc etc. And it is packed full of leading questions.
What I would like to see is a reply from IBM's User Interface group pointing to the research that backs up their design - or at the very least a much clearer indication of how they are going about doing that research. I personally would love to see a simple head-to-head between RealCD and Apple's Audio CD Player, which is by no means perfect, but it would certainly offer the necessary contrast since it is definitely not a real-world interface.
David Gilmore