|
|
Deirdre and Jerry,
Perhaps I should have known better than to get myself mired in the
QWERTY swamp. I wanted to sort out what is and is not proven, and
what is missing in Liebowitz' and Margolis' analysis, and what is
missing in Deirdre's. This naturally involved making 'blackboard'
points that lack the force of other forms of argument. Jerry, I do
not wish to take issue with all that Liebowitz and Margolis wrote.
A couple things I wrote were ill-considered, such as referring
to 'the continued lock-in of QWERTY.' Although I am persuaded
that other keyboards are as much as 10 percent more efficient than
QWERTY (see below), 'lock-in' is perhaps not the best term to use.
What I hold to be essential and still persuasive about Paul David's
argument are the basic points (1) that positive feedbacks
conditioned the adoption and standardization process, and (2) that
our use of QWERTY today can only be understood on the basis of
history, not on the basis of a-priori technological factors or
ergonomic and economic optimization.
Deirdre, we're talking past one another. It may help if I make a
partial concession. The issue of whether a workplace's conversion
is hindered by the mutual externalities (and thus positive
feedbacks) between typists and keyboards (i.e., keyboards outside
the present workplace) only becomes relevant if the workplace-
specific benefits of conversion/retraining outweigh the workplace-
specific costs. As far as I know, this has not been shown.
Moreover, the only study I've heard of (other than Dvorak's) that
addresses that issue--namely the 1950s GSA study discussed by
Liebowitz and Margolis--suggests that it is not the case.
I stand by my basic point that one cannot address whether a specific
case is path dependent without taking into consideration those
positive feedbacks. Moreover, one can argue that the QWERTY standard
was established by a path-dependent process even if its persistence
today can be explained by conversion costs alone. And those
conversion/retraining costs are not as low as you seem to think
obvious. On the basis of the evidence presented by Liebowitz and
Margolis, one does much better learning the Dvorak keyboard from the
start than converting to it from QWERTY.
Jerry, as to my point that alternatives to QWERTY are 2 to 10
percent more efficient: first, Liebowitz and Margolis present
evidence from ergonomic studies that the Dvorak keyboard is 2 to 6
percent superior to QWERTY for typists trained from a 'tabula rasa'
state. L&M's argument is that this is far less than the claims of
Dvorak advocates and that the margin of superiority is smaller
than the costs of conversion, so that there is no market failure
in persisting with QWERTY.
Now, L&M's chief source for their discussion of ergonomics is a 1983
study by Norman and Rumelhart, who conclude there that the relative
advantage of the Dvorak keyboard over QWERTY is about 5 percent.
Norman later concluded, on the basis of conducting a longer series
of studies, that the advantage of Dvorak is in fact about 10 percent
(Donald A. Norman, 1990, The Design of Everyday Things, p. 147 and
endnote). As Norman notes, this is far less than claims of Dvorak
enthusiasts but hardly unimportant in view of the amount of typing
done.
Moreover, last week I received the following email from a
computer scientist at the University of New South Wales (who had
found me through the archive of a past EH.Res discussion):
Noel and McDonald (Proceedings of Interface 89 241-245)
used simulated annealing to find the optimum keyboard
layout. They found that Dvorak was about 10% better
than qwerty, and that the best possible layout was 1.2%
better than Dvorak.
So perhaps the relevant experts are forming a consensus around the
estimate of 10 percent. Beyond that, the concept of finding the
'optimum' keyboard sounds interesting! By the way, 'annealing' in
physics, chemistry, and metallurgy means to add energy to a system
to enable it to reconfigure from a 'locked-in' state to a lower-
energy (i.e., preferred) state. (Arthur mentions this somewhere as
an example of how economists should draw their analogies not only
from Newtonian physics.) I'm not sure yet what simulated annealing
means in this context--perhaps it's an algorithm for progressively
'shaking up' a given keyboard and then testing whether the result
is superior, more or less as in an eye exam or a
computational-general-equilibrium algorithm.
Now, I'm not prepared to argue that even a 10-percent advantage,
if true, necessarily means that the social benefits to a general
conversion to the Dvorak keyboard would outweigh the costs. For
this issue, Deirdre's question is certainly relevant. However, if
conversion costs do prove lower than demonstrated benefits, then
we could say that it is precisely the mutual externalities of
everybody's choice depending on everybody else's, and thus
transaction costs in organizing a transition, that lock us in to
QWERTY. (There might also be a transaction cost, of course, in
persuading people that Dvorak is better.)
Yes, those last few sentences were a 'blackboard speculation.'
But perhaps they help sort out a couple analytical issues.
Douglas Puffert
|
|
|
 Send comments and questions to admin@eh.net
|
||