Reaction to readings 10.25.94

Designing Case Studies

• Rigour! That is what this author wants. While ethnography is a neat word, if it wants to be a science, then it has to play by stringent rules.
• Much of the discussion about how to construct frameworks applies to design in general. (IMHO)
• The boxes form very neat summaries. If you were to put together a talk on this, then the boxes would be the overheads to use.
• The author is promoting ethnographic methods, but does not want programmers rushing off half cocked. The emphasis is on the need for planning if you want to get useful results.
• The paper is a great complement to the Blomberg paper. Blomberg gives the fuzzy approach with a sprinkling of ethnographic caveats. This paper adds some scientific rigour to the study, and would help any new practitioner in getting more out of their cases.
• The key tone here is the dry humor borne of experience: "you will have a stronger appreciation of how little time is left for data collection and analysis".

• It's impeccable. Coupled with the Blomberg paper, it works great. On its own it's a bit disheartening, not giving the reader same enthusiastic glow about ethnographics.
• It would help comprehension a lot if the two different boxes were differentiated visually: overview boxes listing the next steps in one style and advice boxes summarizing the text in another.
• Better page-breaking software would help too. Box 6.4 is one silly example.

Let's Get Real

• I think Landauer is right - theory is useful in narrowly applied areas.
• General HCI theories are pretty thin on the ground. A similar problem permeates most of computer science. Specific theories are plentiful, while general theories are overly general and hard to apply to specific cases.
• Programs are rarely debugged by application of Turing machines or by Boolean analysis. Similarly, HCI rarely be debugged by using low-level theories.
• The author wants a 'usefulness and usability oriented design discipline'. He would be teaching the Terry Winograd design curriculum, focusing on user studies, prototypes, interviews and similar techniques.
• The user is unpredictable and error-prone. No theory is going to capture these characteristics in a generic model.
• Iterative design-test-revise is a useful rapid prototyping approach.

• The problem is that after much moaning about the useless-ness of theory, the cure is 'test-revise' cycles and user-centered-design. Empiric testing of each design will be enough to improve HCI.
• Landauer does not take the next step: with all this new empirical data we can now start to analyze and derive theories of usability. The oft-bemoaned lack of empirical data would change if Landauer got his way. Interestingly this lack of data affects the rest of computer science too...
• I think Landauer is overly pessimistic - theory has its place. He is right in that theory alone will not get you a good interface. The whole system is too complex - but the right theory can give a very good starting point.