A Low-Tech Guide to Knowledge Acquisition


810:162
Intelligent Systems


The following are excerpts from The John McPhee Reader (edited by William L. Howarth, New York : Vintage Books, 1976). In particular, they are excerpts from Howarth's introduction to the book, in which he describes McPhee's laborious method for researching and writing a new piece. I think that the same approach works well when doing knowledge acquisition for a new program.

The basic process seems to be: (1) ask lots of questions, and (2) re-state the answers back to the expert for verification. The only CS-specific addition that I would make to this process is to base many of your questions on scenarios or "use cases"--pose a specific user interaction scenario to the expert, and let her work through the problem-solving process for you. This also works for reviewing protocols of experts in action or watching videotape of experts in action



When McPhee conducts an interview he tries to be as blank as his notebook pages, totally devoid of preconceptions, equipped with only the most elementary knowledge. He has found that imagining he knows a subject is a disadvantage, for that prejudice will limit his freedom to ask, to learn, to be surprised by unfolding evidence. ... he would rather risk seeming ignorant to get solid, knotty answers.



When repeating answers, he so garbles them that a new answer must be provided. ... By repeating and even fumbling their answers McPhee encourages people to embroider a topic until he has it entire.



McPhee has a passion for details, for they convince his readers that he deals in actualities.



When he starts to hear the same stories a third time, McPhee stops interviewing, returns to Princeton, and begins the torturous process of composition.



Writing a first draft is painful work for any writer, whether it moves like lightning or like glue. ... With writing comes the need for endless decisions, mostly on what not to say, what to eliminate. The process is nerve-racking and lonely. ... Writing is the same sort of game: he has spent a long time learning to move against a habitual thought or phrase, which is always the easiest, oldest rut to follow.



A good part of his style rests on knowing the professional "lingo" of a subject. He masters its vocabulary and syntax, even the jargon....



McPhee does not venerate his heroes; he replicates them for our judgment as fully as he can. His principal aim is to maintain an artist's distance and catholicity, to bring alive all the possibilities of a story, not merely a few of its stronger points. He liked Bill Bradley for the totality of his play: "He did all kinds of things he didn't have to simply because those were the dimensions of the game." The same devotion to craft, to a love of skilled enterprise itself, characterizes McPhee's approach to writing.



Eugene Wallingford ..... wallingf@cs.uni.edu ..... January 12, 2011