The discourse-based interview is a method of composition research developed by Lee Odell. He used the method to find out what kind of rhetorical choices people were making, and he offered it as an alternative to protocol analysis (a version of which you went through when you did the text-technology interaction exercise). Odell believed that protocols were disruptive and that a researcher could find out about choices writers make by collecting examples of their writing--several samples of several genres--and then analyzing these documents to find places where the writer had opted for different wording, phrasing, or arrangement. The researcher would then show the writer two samples of her writing from the same genre with highlighted text. These highlights showed differences between the two documents. Then the interviewer would assure the interviewee that he was not questioning the correctness of her choice in any way, and then would ask a question something like, "I see in this letter, you addressed the reader as Dear Mr. Jones whereas in this one you said Dear Justine. Would you be willing to change the Dear Mr. Jones to Dear Ted in the first letter? Please explain why or why not." What would happen is that writers would generally have a strong opinion and would explain the social, stylistic, and rhetorical reasons for their choice.
If you use a discourse-based interview to ask someone about documents she hasn't written herself, you are finding out her knowledge of the genres and conventions associated with the documents. The point would not be to analyze her rhetorical choices but to tap her tacit, or "taken-for-granted," knowledge of her discourse field.