HCI Seminar 357

Day 7 Notes

10/02/02

 

1. One-Minute Papers

Points

Context-sensitive help/personalization of help

Customization and personalization are different

That well built systems seem to be able to effectively "learn" how to

customize/personalize content for an individual.

 

Questions

What can we learn from transaction logs vis-à-vis help?

Is help a learning tool? [e.g., in e-books]  does help go to ‘subject’ as well as ‘system’ [YES!]

What is distinction between explicit and implicit user specification? How do they affect retrieval?

What is the relationship between user-built content based solely on a

 user's experience with a system v. content built on that user's

 relationship with other system users?

 

2. User studies: methods and decisions

 

Christel et al paper illustrates two user studies…first suggesting changes in skims and second demonstrating importance of integration of audio and video information.  Note the design decisions made:

            Research goals (what to emphasize?)

            Treatments (various skims…note the choices in each as well as among each)

            Tasks (factfinding and gisting in first, gisting in second)—note how these are operationalized…consider the many variations.  The actual instruments/stimuli are crucial and there are many decisions to be made that affect results

            Instruments (in addition to the stimuli for tasks, questionnaires or interview schedules, protocols for running the subjects, data gathering tools (e.g., website administration?)

            Dependent variables  (what will actually be measured—accuracy, speed, satisfaction in this case---how will these be captured?)

            Subjects (who to include, how to motivate, how/whether to categorize)

            Procedures (how to run subjects, e.g, individually vs groups; setting, e.g., workstation, room; assignment and ordering of subjects, assignment and ordering of tasks; what to do under special conditions (NEED TO PILOT!), whether to provide training (and if so, what the effects/biases will be)

            Analyses (what statistical or interpretation techniques to use, e.g., if audio/videotaped, will you do verbatim transcripts?)

            Interpretation  (making sense out of the results, relating to the interfaces, linking to literature/explaining why, etc.)

            Writing results (how to organize, what to include, where to send)

            Logistics: IRB process, recruiting subjects, running subjects, consent forms, payments, etc.

 

 

Koenemann & Belkin paper provides another example where scores of design decisions are made to address the question of user control and relevance feedback in particular

 

Note the task was not finding documents but constructing a good SDI (routing) query for a topic.  Note use of non-LIS subjects (avoiding a common criticism)

Note strong evidence for use of interaction with corresponding performance payoff

 

Marchionini & Mu paper summarizes four sets of studies using two very different data collection techniques (usability study and eye tracking study).  Note the importance of task creation.

 

Relate this to our Open Video studies.  The tasks---developing and          

 

 

See IRB procedures:  http://research.unc.edu/ors/ethics.html

 

3. RAVE update

 

4. Readings for next time :

Learning from eye movements: read Jacob (ACM DL)

Biometrics: read Pankanti, Bolle, & Jain  (handout)

Location-aware computing (moved up from last day, handout)

User modeling (Fischer, handout)

 

5 One-minute paper

What was the big point you learned in class today?

What is the main, unanswered question you leave class with today?