Overview Appendix A: CI Models Flow Models Sequence Models Culture Models Artifact Models > Appendix B: Previous Designs Full Design Learning Curve Error Report Problem Profile Data Export Sample Selector Appendix C: LearnLab Surveys Appendix D: Presentations / Misc

Data Export

Wizard Design

This is an early design for data export. It tried to simultaneously take into account actions over data as well as selection of data. When we tested a variant of this design, it became apparent that selection of data was going to be a much larger problem than we anticipated.

Drag and Drop Table Builder

Another early design for data export which also tried to simultaneously allow the user to add data as well as select and perform actions over data. In user testing it revealed some dangerous problems including: confusing terminology, sort functions unclear, and that users had difficulty comprehending how to manipulate the table.(U5)

Sequence-based Filtering Sketch

This idea is a GUI-based attempt at building a query to get sequenced-based information. This is a very difficult problem and an area that a future group could focus a lot of effort on. We abandoned sequence-related interfaces due to difficulty and time. (By sequence we mean "show me all the students who did this, then this, then this.")

Sequence-based Filtering Paper Prototype

This design was a well-intentioned attempt at sequential filtering in the data explorer (what later became the data export). At the top, it allowed users to set where their sequence of events begins and ends. The default settings were the beginning and end of a session, which would include all transactions unless other filters were applied below. Users could specify the start or end of a sequence with conditions that matched a given criteria, such as 'outcome = incorrect'. This design never made it past the paper prototype phase for a number of reasons:

• Users did not know if such a query would find the first sequence that matched the criteria (for example, the first sequence that started with a hint request and ended with a correct answer) or ALL sequences that matched.
• It did not suggest that transaction sequences were isolated to an individual student's session, yet also aggregated over all students' data.
• lack of intimate knowledge of the 'tutor' which generated the data being filtered

Direct Manipulation of Columns

This sketch explored the idea of directly manipulating columns.

Pre-Filtered Array of Columns

The idea of building out a row of columns with filters was one we used in our final designs. The pre-filtering idea may be re-usable in the future with users who already have an intimate understanding of the data mappings in the data shop. Two early users () in a paper-prototype version selected one variable, hit "view data," selected another, hit "view data," etc. (U9, U10) These user tests indicated to us that it was not necessary (and perhaps counter -intuitive) to require users to filter data before any data is selected. (This decision may become limiting if the amount of data grows to be very large and consume more time while loading.)