[pmwiki-users] RFC: Ratings Redux

Ben Wilson dausha at gmail.com
Tue Sep 26 22:35:34 CDT 2006


I seemed to have missed the original thread in March, but I think I'm
in the resurface discussion somewhere. At least, I put together my own
private proof-of-concept that gets pinged from time-to-time.
Naturally, I always have an opinion (although like armpits, they
sometimes stink). But, never afraid to wave my arms about, I'll fork
the thread for a discussion on Ratings. Issues I see in this
discussion (not to limit but to offer for discussion):

1. Scale granularity
2. Axes and what they represent
3. Opinions for context
4. Popularity expressed via other means.

The original discussion suggests more than a simple vote was
preferred. Early on, Hagan referred to Rating, Vitality and Popularity
(courtesy of freshmeat.net) which led to a discussion of freshmeat and
rating axis. Hmm, popularity is based on downloads---not necessarily
the best source of utility. Vitality is visible from the change log.

One suggestion is to put a little more into it, sort of the Amazon
school, but with a slight twist. First, I think a five-point scale is
easiest to wrap the mind around (1-dislike, 2-somewhat dislike,
3-neutral, 4-somewhat like, 5-like). I've seen some ten-point scales,
but what's the difference between a 7 and 8? The granularity of
five-points is easiest to categorize an opinion---in my humble
opinion. So, one issue to be resolved is which scale to use.

Second, I think a rating on two-axis is preferable (or three). I'm not
sure of what to attribute to each axis. However, a scale on the
recipe's ability to do as advertised helps, as may another on
complexity (which may really be better addressed in the
documentation). So, the second issue is how many axes and what are
they rating.

Third, I think an actual opinion helps. In many cases, there's a
problem with the recipe that may not require an opinion---just
communication with the developer.[1] However, a stated opinion helps
others see the recipe through the eyes of those who have tried it or
used it--providing context to the rating. So, the third opinion is
whether to employ an opinion in conjunction with the rating system.

For example: Somebody tries out GoogleMapAPI (GMA) and does not like
the fact that the polylines are forced to websafe colors. So, he rates
the recipe "2." However, you don't care about polylines---you want it
to read in German, or autozoom. So, while the rating helps expose an
aspect of GMA that may not be favorable, you discount the rating. My
point is a two- or three-axis rating means little without the context
provided by an opinion.

Popularity may be possible in other ways. Perhaps those of us who are
active in the community could have a section in our Profiles page that
lists recipes that we use that _aren't_ ours. For example, I've been a
fan of Page TOC from the beginning,[2] but I tend to employ recipes I
wrote.  Hmm, the flip approach would be to have a list of people
employing the recipe, regardless of their opinion. So, a fourth issue
is a method of showing popularity in a quantifiable way.

Ben Wilson

[1]:  On the general topic of recipe maintenance, I think recipe
maintainers should have a way to be contacted listed in their
profile---I'm terrible about this.
[2]:  Although I'm not as much a fan of folding. I drink the koolaid
with a smile.




More information about the pmwiki-users mailing list