[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

[ba-unrev-talk] Contextual Ratings

Eugene just reminded Jack and I that it was Dick Karpinski
who originally brought IBIS to our attention. (Jack has
identified so *many* useful resources, that I naturally
assumed it had been him.)    (01)

The whole episode is a perfect illustration of the power of ratings.
Dick clued us in. I gave it a very high rating. Jack took a look, and
did likewise. The group then had a flag to rally around -- a useful
technology that makes sense for the purposes we want to pursue.    (02)

That result reminded me of the contextual nature of ratings. IBIS
may be excellent or so-so in it's class -- compared to other dialog
structuring tools. On the other hand, it may be great or miserable
for *our purposes*. If we want to build a house, and we're looking
for architecture software, we might give IBIS a minus 400 on our
scale, regardless of how excellent a hammer it might be for the nail
it was designed for.    (03)

It occurs to me that the problem of "contextualizing" ratings can be
solved easily with "local knowledge" systems, but only with great
difficulty in global systems.    (04)

I'm supposing now that I have some sort of web browser that gives
me a view similar to the world wide outline, and that I have some
capacity for merging the ratings of others to get a "meta rating".    (05)

To make sense of those ratings, I need one of two things:
  a) Some way to look through the lens of a local knowledge base
      and see what those folks have to say. (For example, a knowledge
      base/DKR consisting of structured dialog power users would give
      me a great venue for approximating a true rating of IBIS as a tool    (06)

      for such purposes.)
  b) Some way filter the ratings that others have supplied, so I can
      identify the ones that most closely approximate my circumstances.    (07)

There are advantages and disadvantages to each approach:
a) Going through a local knowlege base is less flexible. I can only
    look through one lens at a time. But it's fairly easy to do. In
    practice, it would be the same as (and probably part of) signing
    up to be on a mailing list.    (08)

    So everyone on the ruby-language mailing list would be a
    contributor to the ruby knowledge base, and when I look at URL's
    rating through the "ruby lens", I would get the combined ratings
    of ruby aficianados. Looking through the "XML-based
    documentation" lens (another mailing list I happen to be on) would
    give me ratings from that group. I would have to look through each
    lens separately, but each lens would be fairly easy to acquire.    (09)

(b) The ability to filter ratings would be more flexible. I could create    (010)

     multiple sets of filters and combine them to get a single result.
     It might be easier to identify multiple possibilities that lie
     a single continuum, as well.    (011)

     For example, suppose the IBIS URL has a rating of 4 stars from
     a user with the "rating dimension" identified as "Dialog mapping".
     The existence of that dimensional information lets me search for
     all dialog mapping tools, and create a picture of the extant
continuum.    (012)

     On the other hand, the same user or another user might give it
     2 stars as a "remote cooperation mechanism". Hmm. Food for
     thought. Time to search for other alternatives!    (013)

     The difficulty, of course, is that the rating dimensions themselves    (014)

     need some sort of global ontology, if apples are ever to be
     compared with apples.    (015)

(c) The difficulty of global ontologies being what they are, a third
     alternative springs to mind. If mailing-list membership can be
     used as a way to define "roles", then each rating I make can
     automatically be categorized with respect to a least that subset of    (016)

     the roles I play. (Much harder would be the task of identifying the    (017)

     kinds of projects I have worked on, as the "background" against
     which I make my ratings. But if that were possible, I think that
     the combination of the tools & technologies I am interested in,
     along with the projects I have worked on, would most likely
     constitute an accurate and fairly complete assessment of the
     "commonality" that others might be interested in, when looking at
     ratings.    (018)

     For example, if I were a high-end IT professional with mega-bucks
     in servers requiring milli-second response times, I would be highly    (019)

     interested in the ratings of other high-end IT professionals. On
     other hand, as a one-man development shop, I'm less interested in
     the reactions of people who are managing 120 developers!    (020)

Thoughts for the day...
:_)    (021)