HCI Seminar 310-89
Day 6 Notes
10/9/00
Syllabus
and links to class notes available at:
http://ils.unc.edu/~march/courses/310_f00/syllabus.html
Main
Points
Tradeoff
in supporting old and pusing the new
Innovative
interfaces only part of solution---need good info arch too
Questions
What
options to visual representation are there?
Where
will we be in 10 years?
Does
abundant display hierarchy apply visually?
What
types of data lend themselves to visual rep (e.g., like themescape)?
How
much should we strive for natural? Should we not aim for supernatural?
Why
so many startups at other places but not SILS?
What
is the difference between 3d games and info vis? (why former works well)
Given
video game (3D) usage, what will next generation design?
Principles/models
for mapping info arch to tasks and tools?
Are
we at LCD vs high end interfaces disjunction?
Is
there a single SILS web page that would make each of us happy?
2.
Readings.
Koenemann
& Belkin. Conceptual interface
example
Baseline
rank-retrieval system + one of three ‘added value’ variants.
Relevance
feedback useful (confirming past work)
BIG
Finding: User control (interaction) is both beneficial AND preferred
Christel et al.
Incremental
work embedded in a larger project (Intermedia)
Goal:
Surrogates that give an order of magnitude improvement. (skims)
Skim
creation techniques (representations)
Systematic sampling (10 sec of each
100)
Image-centric (use visual signal
processing)
Audio-centric (use audio signal
processing + text processing)
Combined image+audio
Two
kinds of tasks: fact finding (text?) and gist extraction (text)
Note
this is a within subjects design (compare to K&B above which was across
subjects)
No
statistically reliable differences, move to next iteration of skim design
Increase length of skims (and
full-length video)
Improve synchrony
Use phrases rather than words in
text processing
Pilot
test: Choppy audio worse than choppy video; audio considered more
content-bearing
Focus
on gisting, but use video images+text to ID gist
Only
the best audio with bad visual sync was stat different on visual recog task;
full video better on text task.
Results:
full is better and preferred (no surprise), but improved skim better than
others. Claim that visual content not
so crucial for gisting (we might argue that it depends on what is meant by
gisting).
Usability
must be situated, usability must be focused.
Usability
engineering lifecylce (adapted from Mayhew, 1999).
Requirements
analysis
User profile (user needs analysis)
Task analysis
Platform capabilities/constraints
General design principles
Design/testing/development
Iteration 1
Work re-engineering
(amplification and augmentation here?)
Conceptual model
Mockups
Mockup evaluation
(discount testing ala Nielsen)
Iteration 2
Screen design standards
Screen prototypes
Prototype evaluation
(informal testing, lab testing)
Iteration 3
Detailed UI
specifications and prototypes
Prototypes evaluation
(lab, field testing)
Installation
User feedback
Maintenance
4.
Interface design in context. The BLS
and Fedstats case.
User and Tasks (a taxonomy based on
interviews, focus groups, t-logs, email content analysis)
The Multiple portal visual interface
+ minimizing clicks + abundant display
RB iteration 1 and usability testing
RB iteration 2 and field testing,
revision, usability testing
The challenges of implementation in
real world environments
Next week: read Hutchinson et al.; read Jacob.
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?