[Sigia-l] Concepts and Categories [longish] was mixing apples and oranges and tomatoes

[Tanya Rabourn]

Andrew Otwell wrote:
> Bah. We encounter redundancies every day in real life and have
> learned to handle them fine. I think we're over-thinking how much
> overt consideration people (other than us) give to the purity of
> categorization schemes.

Maybe. Wouldn't be the first time I personally have indulged in a little
over-thinking. But the research that MS did
(http://www.microsoft.com/usability/UEPostings/HCI-kirstenrisden.doc)
seemed to match my own anecdotal evidence. However, it's always possible
that my experience is skewed. I'd like to see something like MS did but
with a more specialized web site. They were testing Yahoo like content.

PeterV wrote:
> Instead, a category is often defined by a "typical example", which is
> then extended in many directions by certain rules. The book explains it
> a lot better of course, it completely shows the fallacy of traditional
> taxonomies.

That book by Lakoff is excellent. I think understanding how we form
concepts and categories is essential for sussing out a user's mental
model.

There's a really long history of psychologists trying to figure out how we
construct categories.

Don't know if this is useful, but here's how I understand it.

Briefly, I think a history of these theories go something like this:
There's the Classical view which assumed that we categorized concepts by
definitions/rules. Any member of a category must have "necessary and
sufficient features," i.e. all squares have four sides, nothing outside
the category "square" has four sides.

Then the prototype view which assumed that we hold an ideal example of
what would belong in a category and any new concept we met would require
us to compare it to the ideal to see if it was a good fit. This view
explained fuzzy boundaries much better than a rule based theory. Keep in
mind that you may have never seen such a, for example, perfect chair, but
what you have in mind for the category "chair" is an idealized example.

The exemplar view however, supposes that you are comparing a new candidate
for the category "chair" with a specific prior example of a chair.

I think this is where someone decided "maybe it's both." That theory
proposes that novices use exemplars for concepts and experts use
prototypes. So, when you first encounter a chair, that actual chair serves
as the benchmark for candidates for the chair category. When you've become
an expert on the whole chair thing, then you are able to construct an
idealized example, a prototype, to compare new candidates for membership
in the chair category.

The latest, which I think Lakoff refers to as an "idealized cognitive
model," addresses problems with both the exemplar and prototype views.
Such things as goal-derived and ad hoc categories, eg. animals likely to
be seen on a safari. I won't go into it since it's much too complex to sum
up here and I've gone on too long anyway. It's one of those annoying
theories that brings up more questions than answers. However, it seems to
solve a lot of problems with the previous theories.

-Tanya
___________________________________
Tanya Rabourn <rabourn at columbia.edu>
[User Services Consultant]
AcIS R & D <www.columbia.edu/acis/rad>
tel: 212.854.0295