Re: [Usability] User problems and practices with modern desktopsystems

[<john erling blad aftenposten no>]

My current occupation is archival systems so I'm familiar with some of
the problems with search, retrieval and presentation of documents.

Hm, this became a bit lengthy..

> your email reflects a lot of what I'm thinking about a while.
> I wasn't shure if this list is the right place for this ideas. And
> please forgive me my english :)

My english is even worse! :)

> > Some interesting observations made by the researchers on
> > classification:
> > 4.1 Observed Classification Practices
> > "Archiving: To all of our users, archiving was decidedly an
important
> > matter.  For this reason, a fair amount of effort was invested, both
> > in creating elaborate file system structures and in labelling them
> > adequately.... Proper classification was perceived as difficult: In
> > our study the users expended considerable cognitive effort with
regard
> > to the classification of documents, and the labelling of folders and
> > documents."

One very important problem is "how do you build a classification system
for an unknown future problem". It is nearly impossible but due to a
large number of known cases some classification schemes has emerged. A
real problem is that those are pretty hard to learn so an average user
does not use them.

> > I think the point here is that it takes a fair amount of work to
> > classify all of your user documents.  I don't know if there is a
good
> > solution to this though, as I don't really agree with the idea of
> > using metadata. and dumping everything to one directory.

Dumping all to a single directory is pretty much what you do on the
desktop. This is nearly useless as an aid for retrieval.

It seems like people who develop systems tend to use this work style,
introducing directories as they need them for some sub parts of their
system, without ever realizing that most users have to work with a large
number of documents.

A journalist might write 10-20-30 articles a day (or even more), or
something like 2 - 10 000 articles a year. It will soon be apparent that
you can't continue to dump articles in a single directory for very long.

> We don't have to stick on old metaphers whatever it cost anymore. the
> world in computers is an other than world than in real life. For
> example I don't have gimmicks in a folder in an other folder and so
on.
> May it be more helpfull to show the user that things in a computer are
> not limited as things in the real world?

Alternate grouping might be very useful. In fact, you can dump
everything in a single directory if you could regroup the data somehow.

> So I would like to have something like a mind mapping strukture: data
> xy is in relation to job x, projekt y done at June04 and so on.
> Additional with relations to the mimetype or possible applications.
> These mapping may be created by hand and by the system. By hand maybe
> with arrows to icons representing the projects or with popup menues.
> The system may generate links when I create a document by reusing an
> older one, or because I copy 'n paste content.
> the data are stored in kind of relational database and the links are
> m:n relations to other objects.
> searching will be a matter of selecting the patterns the data belongs
> to.
> A folder named "job application" will now be an icon opening a window
> with all data mapped to it. selecting an application shows the
> intersection.

If you work on one document while you open some other document, you
could assume those two documents are related somehow. You could even use
automatic classification to enhance further some relations while you
prune other.

If the user chooses to store some kind of data (an article) in one
directory, and he at some later point in time inspects relations for one
related object (an other article for example) he could get a direct link
to the first data or the enclosing directory.

In a file manager this could be done through the popup menu assigned to
the object or by some kind of left pane (in the old Nautilus).

Other explanations of this could be by describing the directory
structure as one dimension assigned to an object. (It is really a mix of
dimensions in a hierarchy.) It could be other equally important
dimensions. My documents could be such a dimension, but perhaps owners
would be a better dimension. This dimension exists already in the
current file system. Music could also be one such dimension or pictures.

In some classification schemes this is folded into a strict hierarchical
system which is easy to recreate in a file system. Other system tries to
use a graph.

A real world classification scheme is not so simple at all and is more
easily described as an higher dimensional matrix, and even an matrix
where some dimension don't apply if some other dimension apply and vica
verca. For example you might have both a music (giving artist for
example) and a picture dimension (giving names of those you have
pictured) if you have stored image and sound in the same file but it
won't be very likely you need an "nobel prizes" dimension. But then, at
some point your fellow band member gets an nobel prize in astrophysics..

Now, most classification schemes don't use dimensions but rather
discrete topics. That idea is much easier to grasp. A common situation
is to mark a class field as "biography" and then peruse a person field
as describing the person you describe. This leads to some rather amusing
errors.

If you open an MSword document you will find that you can add notes in
various fields in the property box. Those fields can also to some extent
be changed to suit the problem at hand. It is a good idea to be able to
change the fields but I somehow think that some kind of selection
between schemes work better as it is possible to enforce a specific
methodology for a specific problem. If you try to classify a biography
you know you *must* set the persons name but don't have to specify what
kind of music instrument he or she plays.

Of course, the default scheme should set as much information as possible
from automatic means. It is not very difficult to find person names in
text. Neither is it very difficult to find geographic references, or
even references to technical stuff from howtos.

What is difficult is to find a way to move from a previous unknown
problem description (find the nobel prize winner) and to a known one. I
belive Medusa could be somewhat nice to have as a fallback.

So what does archive systems use the directory structure for? Very often
they create a filename with a unique document id and place it within a
project folder or a folder with a production date.

John