random memes }

Story based reasoning and interviews

Roger Schank (a prominent AI figure) gave a talk at UCI several years back. His notion was the humans do not naturally reason deductively, rather we use something he called "story-based reasoning". In fact most of human communication is in the form of "mini-stories". This upset the deductive AI people to no end, but to me seemed to make a great deal of sense.

(There is a certain amount of recursion here).

When I have done interviews I have found it useful to get the interviewee to tell stories. In particular stories about work they have done in the past - what worked, what didn't, points of interest, how did they approach the problem.

My notion is that if you can cast your experiences as a story, bringing out points of interest both high and low, then you are likely to have gained a good understanding.

In one run of interviews we divided up so that one set of interviewers did the detailed tech questions, and I tried to pull stories from the interviewee about work they had done. Our opinions after the interviews agreed remarkably. Those interviewed who could tell good stories, generally did well on the tech questions. In a few cases the interviewee did well in the tech questions, but not at telling stories.

In those cases where a candidate did well on the tech questions, but not at telling stories, our opinions agreed. The interviewers who had asked the tech questions were usually uncomfortable with these interviewees.

(So I've just told you a story, about eliciting stories from the interviewee, to evaluate reasoning ability, in terms of stories...).