If you haven’t read Donald Norman’s The Design of Everyday Things, I would urge you to put it on your reading list. It was a catalog of human factors design failures, and a set of prescriptions to improve design. An excellent read.
In The Design of Future Things Norman tries to recreate his earlier success, with mixed results. He is as insigntful as ever, but the material is a tad thin. To fill space he resorted to repetition and added a wholly extraneous and redundant ‘afterword’.
He proposes six rules for designers of automated or ‘smart’ machines:
Provide rich, complex, and natural signals.
Provide good conceptual models.
Make the output understandable.
Provide continual awareness without annoyance.
Exploit natural mappings.
He provides at least one negative and one positive example for each of these rules.
A major theme of the book is that machines work well if they can be fully automated and operate autonomously, or when they are completely human controlled. But the middle ground is troublesome. If automation works most of the time and then suddenly fails, trouble results. Since many tasks cannot be fully automated, it is a challenge to engineers and designers to find a way to design machines that can interact safely and usefully with humans.