Rick Marshall wrote:
> re sorting names
> 
> day to day use in english speaking countries - i find it indispensible -
> it's the way my users think.
> 
> while from a technical perspective - ie finding a given record - the
> index can be anything useful - like a ssn, but when you sit down next to
> users and watch their interaction with data over a whole day or week,
> they are noticeably more comfortable with meaningfully ordered lists.
> 
> i even have one user who just can't work with company names sorted
> technically correct, when the company is somebody's name - they always
> put them in the system as "lastname firstname pty ltd" - which is
> technically wrong, but i can't seem to get them to change.
> 
> sorting and the use of sorted lists is both a technical issue (accessing
> the data) and a psychological issue (comfort and efficiency in human
> interaction with the data). [snip]

Perhaps I should change the Subject when I change the subject, but 
speaking of the way users think...

I have noticed over the years that whenever a menu, table or other kind 
of list too long to be taken in as a whole is presented in sorted order, 
a significant percentage of users will request it in another order.

When I've probed these requests, it has rarely been a case of multiple 
possible sort keys, but rather the user's desire to cluster the data in 
ways that put related items in close proximity. To give an obvious and 
typical example, in a list of widget properties, few people report 
issues with "X" and "Y" coordinate properties, but many people think 
"height" and "width" are too far apart.

Hierarchical organizations are often suggested, e.g., grouping the 
height and width properties together under a structured Dimension 
property. But this introduces the problem of short term memory load. In 
a long list, users are often hard-pressed to recall the names of things, 
much less the names of containers of (containers of (containers of)) 
things. Thus, when a hierarchical grouping is provided, a significant 
percentage of users will request a command (or default) to open all 
levels of heirarchy at once, so that, e.g., height and width can be 
scanned for individually as well as by heirarchy.

In the laboratory, IIRC, it is the usual result that sorted order allows 
the fastest possible search in long lists, and that any sort of 
clustering or heirarchical aggregation slows down searches, even if 
subjects believe it has speeded them up. (As a counter-intuitive result, 
I liken this to Apple's well-studied finding that menus at the top of 
the screen (when there is only one screen) are faster to access than 
menus in individual windows, even though the users being observed often 
do not believe it.)

It is certainly true that once a desired item has been located, a 
clustered list allows faster access to related information...if the 
relationship expressed by the hierarchy happens to match the 
relationship trail followed by the user. Often, it does not.

I only offer this as another field observation of human nature, and 
leave it to others to draw conclusions.

Bob Foster
http://xmlbuddy.com/