Expert systems are a clear cut example of an artificial intelligence technology brought successfully to a number of specialized markets, so successfully that the curse of AI falls heavily on expert systems. It seems pretentious to call them artificial intelligence. So it must be adding insult to injury to suggest that even the term "expert system" is pretentious.
One of the standard "mysteries" of artificial intelligence, posed in most articles and beginning courses on the subject, is the difficulty of simulating a nonexpert. Why, the puzzle goes, is it easier to simulate knowledge and behavior of an expert than to simulate the knowledge and behavior of a child of three? or of a chimpanzee?
The real mystery is why people keep asking the question, "Isn't the answer obvious?" The first synonym for "expert" in Rodale's The Synonym Finder is "specialist." Specialization is a narrowing of focus, and there is no mystery at all about why it's easier to simulate behavior in a narrow domain than in a broad one. True, narrowing of focus is not one of the defining features of expert systems, but I'd be willing to bet that it is the one that makes them successful. Perhaps it would be more modest and more meaningful to call expert systems "specialized systems."
Nuke the Innumerate
According to the editors of CD-ROM End User, who use the following expression as a running head throughout their useful journal: 
(Digital Data x Mbytes) * CD-ROM> E=mc{2} means, "The sum of megabytes of digital data plus CD-ROM is greater than... "(the ellipsis is theirs). On the principle that from a false premise any conclusion can be derived, I suppose, that they can assign any meaning they like to this meaningless sequence of symbols. Perhaps I shouldn't be so hard on the editors; after all, they are only using the expression as techie-looking illustration. But the expression is so egregiously innumerate that it calls into question the technical accuracy of everything in the magazine. To cite one example, should we trust the math of people who don't recognize that they've used different symbols for multiplication in one expression?
My interpretation of the expression is, "When you combine megabytes of digital data on CD-ROM ignorantly, the results are more than likely to blow up in your face."
See Tog Flame
I praised CD-ROM End User here recently, and I still think it's an informative publication. I also praised Dave Thompson's Micro Cornucopia, which, alas, appears to have published its last issue. Dave, weary of the grind and eager to get on with other things, sold the publication to Miller-Freeman, who probably won't continue Micro C as a magazine. Too bad.
Another of my favorite publications, Apple's developer magazine Apple Direct, stubbed its toe recently when it announced Apple's new hardware products days ahead of the official announcement date. Another setback for the anti-Perestroika cultural reeducation program at Apple.
What I like most about Apple Direct, apart from its timely information on new products, is a user interface design column written by Mr. Tognazzini. (Sometimes he's Bruce, sometimes Bruce "Tog," but usually he's just Tog, which is also his AppleLink address.) Some of Tog's humor may not please either Apple brass or developers: "I have forwarded your link ... into the very heart of Apple Engineering Land; and, rest assured, we shall eventually do something official." Tog's opinions, of which he has his full share, frequently deviate from the party line, as when he said that he would "rather die than defend this silly business of throwing disks into the trash to eject them."
In that same March issue that preannounced the new hardware, Tog had some sage things to say about user testing on the cheap. Among the radical ideas is that user testing should be done by members of the design team, using no more than three people per design iteration. User testing can show you things you'd never guess about your product and your users. Tog relates an experience in user testing of an in-box tutorial in which one apparently trivial question proved to be the most difficult of all. The "difficult" questions posed no problems for the test subjects, but the "easy" question proved to be the most difficult to phrase unambiguously. He just wanted to find out whether the test subject was using a color or black-and-white monitor. It took six iterations to get the question right.