CS377: Reactions to 11.01 reading

Christian mogensen@cs.stanford.edu http://www-pcd.stanford.edu/mogens/377/reaction-1101.html

Woods: Clumsy Use of Technology

This reading is long - far too intense to assimilate in a single session. Starting with new concepts and definitions of error, we jump through fourteen premises. After this rapid summary of systems thinking, we discover that errors come in phenotypes and genotypes.
Jumping on to chapter 5, we look at 'clumsiness', distributed cognitive systems and impact flow diagrams. Finally, after mentioning the keyhole effect we look at an actual example from the Apollo 13 mission.
The point here is that there is a lot in this reading. A lot of these ideas are really cool: distributed systems thinking, error genotypes, impact flow diagrams.
Unfortunately, the reading selection has few examples to illustrate these ideas. The stories that are there aren't differentiated from the rest of the text. This makes it harder to disentangle comments on the case from the case itself. The short snippets and references to other cases just irritate - they assume an expert's familiarity with the literature.
In the near miss at the Davis-Besse nuclear; station (NUREG-1154), there were about ten machine failures and several erroneous actions that initiated the loss of feedwater accident and determined how it evolved.
Ch. 2 - Premises, p.13
It doesn't give enough information to help understand the point being made unless the reader is much more familiar with the incident. Similarly, the new ideas (such as the IFD) need to be illustrated in greater detail to be understood in context (to borrow a phrase).
The reading flows; like a firehose. There is a distinct jump between chapter 2 and 5 - there isn't a common thread tying the two together. The fifth chapter makes few references back to its premises.
Chapter five make the interesting point of talking about human computer cooperation: a phrase that hammers home the need for thinking about cognitive systems.
The paper seems to be claiming that designers can anticipate all interesting events. Having the system display or highlight all interesting events leads to cognitive overload again as meaningful events rarely travel alone. Building meta-events out of single events is a hard AI problem.
The Apollo 13 example is a relief - it takes one example and studies it in depth and analyses it in the light of the material discussed.

Laughery: Everybody Knows

The obvious is not obvious to everyone - we must focus on determining the users perceptions of the system; knowledge of the domain and perception of risks in the system.
This comes back to the cognitive loads mentioned in the Woods paper. Risks are rarely on our minds, as accidents are rare occurences. Hence a design must make risks obvious, rather than hiding them.
The point is that once a designer is aware of a user-hazard, it is possible to design to compensate for it (the prevent-guard-warn approach).
The paper is a call-to-arms for ergonometrists. They need to get out and talk to designers and manufacturers. Once a designer is aware of the risk of ignoring risk, she can compensate by using an ergonometrist to help her discover the hazards before they become dangerous.
This is the first time legal definitions have been brought into the discussion. It is important that designers be aware of the law. This has interesting implications, which unfortunately the paper does not discuss.
It's short - always a virtue. The paper makes a nice short survey of some of the problems in design that are so obvious we regularly overlook them. It would fit nicely after the Don Norman POET.