[
Lists Home |
Date Index |
Thread Index
]
On Dec 15, 2003, at 2:39 PM, Joshua Allen wrote:
> So I assume the
> main point of discussion here would be about the difference between
> hierarchy vs. list thinking.
I'm not sure that was Tufte's point, but it's a good one.
>
> Anyway, people who like outliners and mind maps are aware that outline
> style is obviously superior :-) But maybe some semantic web people
> would argue that graph style is even better. However, I think E-R
> diagrams and UML diagrams are useful in very specific domains, but
> quickly become overwhelming.
Hmm, I'll throw out the following strawman:
- Knowledge is probably represented in real human minds, books, the
Web, etc. as something like a graph of interconnected neurons,
references, hyperlinks, etc. That is EXTREMELY powerful, but somewhat
difficult to grok. To the extent that the concepts and relationships
can be formalized with E:R methodologies or true relational schemas,
this is great, but mainly because it allows flexibility and formalism,
not because it offers a really different perspective. Remember when you
learned a new subject in school, or stumble on one on the Web -- one
tends to be overwhelmed by the concepts and relationships you have to
know to make sense out of one thing, and that's densely connected with
assumptions about others. You just have to dive into it and go with
the flow ...
- ... unless you have a textbook/instructor who has decomposed it into
a nice linear sequence of axioms/theorems or a hierarchical taxonomy.
That's monumentally hard; Euclid did a fine job of the former for a
very limited set of ideas, but Aristotle's taxonomies are totally
laughable today -- we honor him for the *style* of decomposing
knowledge into taxonomic hierarchies, not for his actual taxonomies.
Good taxonomies today (thinking of SNOMED) are usually based on
centuries of research and, ahem, often cost serious money.
- Very few people giving presentations can even aspire to coming up
with a nice linear sequence of logic or a coherent hierarchy; the best
they can hope for is a few bullet points giving memorable ideas, and
some graphic devices to assist in explaining them. That's probably
fine for a typical product presentation, but as Tufte implies, it's
practically criminal for something like an analysis of the probability
of Columbia surviving re-entry.
- So, XML does encourage the "cognitive style" of presenting
information in hierarchies. That's a good thing for data interchange,
and the more work that has gone into preparing the schema and/or
taxonomy (ontology if you insist), the better, but the raw hierarchies
encode the most important relationships. But it's a happy medium
between the totally unstructured/disconnected bullet points that the
average salesperson is trying to get across, and the densely
interconnected graphs/relations of a real knowledge base.
|
