 | | | | | | |
| |
|
In lugnet.admin.general, Thomas Main writes:
> [...]
> I know that there are some good aspects to article rating...such as filtering
> noise and creating a "best of LUGNET" sort of highlight page...but I don't
> think the current system is the best tool for the job. Consider if, instead
> of a 0-100 ten point scale, there was just an option to rate an article as
> "I think this is noteworthy" By default, the articles an individual didn't
> think were great would sink to the bottom and articles that stood out as
> particularly important would rise to the top. The ambiguity of the ratings
> would be diminished too...I don't know what a rating of "30" versus "40"
> means...
Well, it doesn't really mean (and isn't supposed to mean) anything profound
but simply that the 40 is 10% higher than 30 on the recommendation-to-read
scale. Similarly, an 80 is simply 10% higher than a 70 -- nothing profound.
But the difference between an 80 and a 30 is more meaningful, and the
difference between a 95 (for example some of Brad's announcements) and a 23
(for example some bickering and name-calling in .debate) is profound.
The ratings are intended mainly for casual and less-active readers -- to help
them locate things they might find interesting or useful more quickly. There
is still a lot "to do" of course (such as top-N recent listings) before this
will truly become helpful.
> but I can understand "look at these articles...a majority of LUGNET members
> thought these were excellent." Simplifying the system would also eliminate
> some of the subjectivity of the scores the articles receive. Someone might
> think "40" is a ok score whereas someone else might choose "60" as a low
> score.
Do you think that a standardized rating-recommentation info-page would help?
I've been following Richard Franks's suggestion for the past couple of weeks
and trying to consider 50 an "average post" midpoint. Most of the things,
I end up marking 50 or 60, with some 40's and 70's and occasional 80's, 90's
30's, 100's, and 20's. That seems to produce consistent results from day to
day, and is quick.
Also, do you think that the default rating should be 0 rather than 50? For
a default of 0 would mean that articles tended almost always to go upwards
in rating over time, rather than either upwards or downwards -- in other words,
no one would feel that their post was ever "marked down from a 50 to a 30 or
40," but rather that their post was "marked up from a 0 to a 30 or 40."
> There's less confusion about a system that just uses "noteworthy" as a
> "good" score and lets other messages default to "no comment."
Do you mean like a 0 (zero) and 1 (one), or 0 (zero) and 100 (one hundred) and
averaging those rather than more gradations in-between?
--Todd
| | | | | | | | | | | | | | | | |
| |
|
Todd Lehman wrote in message ...
> In lugnet.admin.general, Thomas Main writes:
> > [...]
> > but I can understand "look at these articles...a majority of LUGNET • members
> > thought these were excellent." Simplifying the system would also • eliminate
> > some of the subjectivity of the scores the articles receive. Someone • might
> > think "40" is a ok score whereas someone else might choose "60" as a low
> > score.
>
> Do you think that a standardized rating-recommentation info-page would
help?
Yes.
I would also suggest that some explanatory text (i.e. how ratings
should be interpreted) be added to the rating histogram page itself.
I think that some people might interpret a low rating as "a bad post"
in general, which is not the objective, of course.
For the casual user, the intent of the rating system should be
as clear as possible.
> Also, do you think that the default rating should be 0 rather than 50?
I think that the chance for misinterpretation of the rating system
would be the same regardless of the starting point. The current
system works well for those who understand what it means.
--
John
(remove the obvious to reply)
| | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Todd Lehman writes:
<snip>
> Do you think that a standardized rating-recommentation info-page would help?
Yes, that's a great idea.
<snip>
> Also, do you think that the default rating should be 0 rather than 50? For
> a default of 0 would mean that articles tended almost always to go upwards
> in rating over time, rather than either upwards or downwards -- in other • words,
> no one would feel that their post was ever "marked down from a 50 to a 30 or
> 40," but rather that their post was "marked up from a 0 to a 30 or 40."
Yes, I think that would alleviate the perception that certain people don't
approve of the posts someone else is making...I like the idea of a post getting
points for its merits and "floating up" rather than being marked down.
<snip>
> Do you mean like a 0 (zero) and 1 (one), or 0 (zero) and 100 (one hundred) and
> averaging those rather than more gradations in-between?
>
> --Todd
Yep, I think so. I see you as the kind of person who likes gradations...you
see the need to keep adding to and refining the newsgroups on LUGNET, you like
details, compartmentalizations, complexities, hairsplits... :) I prefer simple
things (as long as they work). The newsgroup rating system would be simpler
for me if there were less choices...and I think it would be more
objective (of course the recommendations page would also address this problem).
Thank you for considering my ideas.
--
Thomas Main
main@appstate.edu
| | | | | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Thomas Main writes:
> [...] The newsgroup rating system would be simpler for me if there were
> less choices...and I think it would be more objective (of course the
> recommendations page would also address this problem).
Do you mean that from the point of view of a producer or a consumer of the
rating information (or both)?
As a producer of ratings, it is certainly your right to treat the rating
levels more coarsely if that helps you produce ratings more comfortably or
more quickly or more meaningfully to you. In other words, if you wanted,
you could apply this approach:
- Liked it: mark it High (100 -- the rightmost choice)
- Didn't like it: mark it Low (0 -- the leftmost choice)
- No opinion: leave it blank
The system is designed to work whether people always rate 0 or 100 and never
use anything in between (ultra-coarse), or whether they use the 11-point scale
from 0 to 100 by 10's (moderately coarse), or whether they use the 101-point
scale from 0 to 100 by 1's (very fine). (There's currently no interface which
actually uses the whole range 0 to 100 by 1's, but the system will handle it
if someone wrote a custom client to submit fine-grained ratings. The finer-
grained ratings will be more useful for things like sets and websites, of
course, and not too useful for news articles since they are so temporal.)
As a consumer of ratings, any method for calculating a rating which averages
input ends up producing some sort of multi-position scale. That is, even if
only two inputs are is 0 and 100, the average of several values still might be
something anywhere in-between such as 57 or 83.
As a consumer of ratings, would it help you more or less if the output had
fewer degrees of freedom? Siskel & Ebert used a 5-degree system for rating
movies: -2, -1, 0, +1, +2. Other rating methods include use 4-star or 5-star
systems and some of these even output 8 or 10 positions by giving halfs as well
as wholes (i.e., "3 1/2 stars"). Then there's the classic primary-school
rating system of A, B, C, D, F, sometimes with +'s and -'s, giving a 15-point
scale. And then there's the classic secondary-school rating system of a 0 to
4 (or 5) point system producing extremely detailed average GPA's with 4
significant digits.
Giving a two-digit rating 0 to 100 which is intuitive in the sense of a
percentage seems like the simplest general-purpose way to go for something
where the domain of input (articles) spans the entire emotions-range from
incredibly exciting to incredibly disgusting.
Would you find it useful as a consumer or producer of ratings if you had the
option to specify how many choices (radio buttons) you saw when you rated
messages and how many scale-steps you saw when you viewed ratings?
--Todd
| | | | | | | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Todd Lehman writes:
> In lugnet.admin.general, Thomas Main writes:
> > [...] The newsgroup rating system would be simpler for me if there were
> > less choices...and I think it would be more objective (of course the
> > recommendations page would also address this problem).
>
> Do you mean that from the point of view of a producer or a consumer of the
> rating information (or both)?
I think both would be useful...
>
> As a producer of ratings, it is certainly your right to treat the rating
> levels more coarsely if that helps you produce ratings more comfortably or
> more quickly or more meaningfully to you. In other words, if you wanted,
> you could apply this approach:
>
> - Liked it: mark it High (100 -- the rightmost choice)
> - Didn't like it: mark it Low (0 -- the leftmost choice)
> - No opinion: leave it blank
And the few times I have used the system...that's the way I've used it. I
marked a few articles I read at "100" -- I don't think I would ever rate an
article "0" though...I only rate things that I think are really important (who
cares if I think something is really *not* important...I guess I just don't
like to spend too much time focusing on the negative). One thing I really
don't understand is who goes to the trouble of rating articles that everyone
realizes are not important....like maybe the "me too" posts (in general groups,
not the new LEGO ones) or auction posts that just convey information about an
auction. To me, it seems like a big waste of time on the part of the person
rating.
>
> The system is designed to work whether people always rate 0 or 100 and never
> use anything in between (ultra-coarse), or whether they use the 11-point scale
> from 0 to 100 by 10's (moderately coarse), or whether they use the 101-point
> scale from 0 to 100 by 1's (very fine). (There's currently no interface which
> actually uses the whole range 0 to 100 by 1's, but the system will handle it
> if someone wrote a custom client to submit fine-grained ratings. The finer-
> grained ratings will be more useful for things like sets and websites, of
> course, and not too useful for news articles since they are so temporal.)
>
> As a consumer of ratings, any method for calculating a rating which averages
> input ends up producing some sort of multi-position scale. That is, even if
> only two inputs are is 0 and 100, the average of several values still might be
> something anywhere in-between such as 57 or 83.
I realize this...and that's fine...I think the resulting "average" is more
representational of the entire group's opinions, rather than the "weighted"
opinions of one or two persons who chose to use coarse versus fine rating
scale. For instance...since folks have the option of rating a post "60" and
most people will take advantage of the available options and rate things in the
gray area between 0 and 100. If I continue to just dole out "100s" that's
skewing the average and it's not fair. I either have to ask for a simpler
system or try to use the "fine" system as it's intended.
>
> As a consumer of ratings, would it help you more or less if the output had
> fewer degrees of freedom? Siskel & Ebert used a 5-degree system for rating
> movies: -2, -1, 0, +1, +2. Other rating methods include use 4-star or 5-star
> systems and some of these even output 8 or 10 positions by giving halfs as • well
> as wholes (i.e., "3 1/2 stars"). Then there's the classic primary-school
> rating system of A, B, C, D, F, sometimes with +'s and -'s, giving a 15-point
> scale. And then there's the classic secondary-school rating system of a 0 to
> 4 (or 5) point system producing extremely detailed average GPA's with 4
> significant digits.
>
> Giving a two-digit rating 0 to 100 which is intuitive in the sense of a
> percentage seems like the simplest general-purpose way to go for something
> where the domain of input (articles) spans the entire emotions-range from
> incredibly exciting to incredibly disgusting.
>
> Would you find it useful as a consumer or producer of ratings if you had the
> option to specify how many choices (radio buttons) you saw when you rated
> messages and how many scale-steps you saw when you viewed ratings?
>
> --Todd
As a consumer, I'd rather not see the ratings at all ;) Usually I read LUGNET
through a newsreader, so I don't see the ratings, but occasionally I peek at
the web page and see that this or that article was rated a certain way and I
think, "Wow, someone has a lot of extra time on their hands," or "Why would
someone rate this article so low," or "Wow, that rating seems petty and
vindictive to me...I wonder who is doing all this rating anyway?"
As a producer...well, I think it's best if I don't rate posts because I don't
want to mess up averages by giving out 100s and I don't want to have to think a
long time to try to come up with a number between 0 and 100 (particularly, if,
you're like me and would rate the same message "40" one day and "60" another
depending on when you read it, what you were thinking about at the time, and
all the other random things that contribute to scoring subjectively.)
--
Thomas Main
main@appstate.edu
| | | | | | | | | | | | | | | | | | | | | | | | |
| |
|
Thomas Main wrote:
> And the few times I have used the system...that's the way I've used it. I
> marked a few articles I read at "100" -- I don't think I would ever rate an
> article "0" though...I only rate things that I think are really important (who
> cares if I think something is really *not* important...I guess I just don't
> like to spend too much time focusing on the negative). One thing I really
> don't understand is who goes to the trouble of rating articles that everyone
> realizes are not important....like maybe the "me too" posts (in general groups,
> not the new LEGO ones) or auction posts that just convey information about an
> auction. To me, it seems like a big waste of time on the part of the person
> rating.
I just thought of a way to alleviate this problem.
Don't publish the average rating unless there are at least N values (N =
10?).
If there are fewer than that many values, just say "insufficient sample"
or
something like that. That way only the articles that people feel are
important
will will get ratings, and articles that are irrelevant will continue to
be ignored.
I also think the idea of starting at 0 and going up is a good one, but
that
is an independent decision.
/Eric McC/
| | | | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Charles Eric McCarthy writes:
> I just thought of a way to alleviate this problem.
> Don't publish the average rating unless there are at least
> N values (N =10?).
I think that if N can be defined by the users, this would work well. I
personally think that if even one person rates a message that some information
is better than none. However that's just me, and I think that new users would
like to have N a little higher. If it could be changed much like the
"skip-filter", that would work well. I like 10 as a default value, but I
personally would change my setting to 1.
Ben Roller
| | | | | | | | | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Thomas Main writes:
> [...] One thing I really don't understand is who goes to the trouble of
> rating articles that everyone realizes are not important....like maybe the
> "me too" posts (in general groups, not the new LEGO ones) or auction posts
> that just convey information about an auction. To me, it seems like a big
> waste of time on the part of the person rating.
Probably depends on how quickly they can rate articles. Through the website,
it's a little cumbersome to give input on everything -- a lot of scrolling and
mouse-clicking and waiting, etc. But using a custom newsreader client, it can
be as fast as a single keystroke.
For example, if I want to rate an article 70, I just press the "7" key on my
keyboard and it queues up a 70 for that article (which it sends to the server
in the background) and then immediately shows me the next article. If I could
(theoretically) actually read 1 article per second, I could actually rate 1
article per second. But my brain doesn't work that fast. :)
My overhead for rating something I've read is probably 1/2 second per article.
I have to hit some key to advance to the next article anyway, so it might as
well be one of the rating keys. If I don't have an opinion, I just hit a non-
rating skip-to-the-next-article key and don't mark the article.
> As a consumer, I'd rather not see the ratings at all ;) Usually I read
> LUGNET through a newsreader, so I don't see the ratings, but occasionally I
> peek at the web page and see that this or that article was rated a certain
> way and I think, "Wow, someone has a lot of extra time on their hands," or
> "Why would someone rate this article so low," or "Wow, that rating seems
> petty and vindictive to me...I wonder who is doing all this rating anyway?"
How about search results? Sometime down the road (a long way, probably) the
search results could take the ratings into account (at your discretion at
search-time) and you could ask the search engine to give higher priority to
articles with higher scores. Would you find that useful?
> As a producer...well, I think it's best if I don't rate posts because I
> don't want to mess up averages by giving out 100s and I don't want to have
> to think a long time to try to come up with a number between 0 and 100
> (particularly, if, you're like me and would rate the same message "40" one
> day and "60" another depending on when you read it, what you were thinking
> about at the time, and all the other random things that contribute to
> scoring subjectively.)
Well, maybe it's not something for you then. (Certainly, it's not for
everyone, because it does take a bit of time and concentration sometimes.)
What if, instead of giving out 0's and 100's, you could give out either a
25 or a 75? (Or some other pair?) Would you find that less intimidating?
--Todd
| | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Todd Lehman writes:
> For example, if I want to rate an article 70, I just press the "7" key on my
> keyboard and it queues up a 70 for that article (which it sends to the server
> in the background) and then immediately shows me the next article.
Apologies if this has been asked before and I missed it, but is this available?
It sounds *extremely* useful!
> How about search results? Sometime down the road (a long way, probably) the
> search results could take the ratings into account (at your discretion at
> search-time) and you could ask the search engine to give higher priority to
> articles with higher scores. Would you find that useful?
Yep! Yep! Yepyepyepyepyepyep! Useful! Yepyepyep!! Yeeeep!
Actually, if that included fuller search capabilities as well (date, author,
subject, etc) then I'm sure I wouldn't be the only one doing bad Seasame-Street
alien impersonations :)
Richard
| | | | | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Richard Franks writes:
> Apologies if this has been asked before and I missed it, but is this
> available? It sounds *extremely* useful!
Mine is a horrible hack crufted together to run in text mode with Curses on
my particular home machine, but Jeremy Sproat has written a general-purpose
platform-independent newsreader in Java, and I think he might be considering
adding rating capability to it. (Or was that Dan Boger?)
(Search for "GUI streaming newsreader Jeremy Sproat Dan Boger"...)
--Todd
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Todd Lehman writes:
> In lugnet.admin.general, Richard Franks writes:
> > Apologies if this has been asked before and I missed it, but is this
> > available? It sounds *extremely* useful!
>
> Mine is a horrible hack crufted together to run in text mode with Curses on
> my particular home machine,
If you mean that it would be virtually impossible for me to do some hacking of
my own to get it working, or you're embarrassed to share the source, then fine!
Otherwise, I'm still interested :P
> but Jeremy Sproat has written a general-purpose
> platform-independent newsreader in Java, and I think he might be considering
> adding rating capability to it. (Or was that Dan Boger?)
Yup - these are cool developments, but ideally I want something low overhead -
Linux has started thrashing on my p166 already, and once I start up a few java
instances it crawls even more..
Richard
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Richard Franks writes:
> > Mine is a horrible hack crufted together to run in text mode with Curses on
> > my particular home machine,
>
> If you mean that it would be virtually impossible for me to do some hacking
> of my own to get it working, or you're embarrassed to share the source, then
> fine! Otherwise, I'm still interested :P
It's not particularly bad code or anything like that, it's just that it was
an evolve-mode prototype -- didn't know Curses at all before digging in (still
don't know it well) and wasn't sure it would even end up working. It also is
still using an older undocumented pre-avid.cgi gateway to the server for its
incoming feed, so until I update that to avid.cgi, I can't release the code.
But maybe it would be a useful example client if cleaned up a bit and released
with the understanding of no little or support being offered to get it up and
running (I just haven't the time to support it).
> Yup - these are cool developments, but ideally I want something low overhead
> - Linux has started thrashing on my p166 already, and once I start up a few
> java instances it crawls even more..
What I made is pretty low-overhead -- it just uses Perl5 and the Curses.pm
Perl library and runs probably any Linux (although many of the colors are
currently hard-coded for my settings) and typically consumes about 2-4 MB of
RAM while active. Lemme think about what would be involved in making it
releaseable...
--Todd
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.off-topic.geek, Todd Lehman writes:
> until I update that to avid.cgi, I can't release the code.
> But maybe it would be a useful example client if cleaned up a bit and released
Funky!
> with the understanding of no little or support being offered to get it up and
> running (I just haven't the time to support it).
Yup - I forgot to mention that I expected no support for it :) In fact, if
someone doesn't take up the challenge before me (I couldn't justify it until
June, so it's likely they will!), it would make a good way to get a bit more
perl experience.
Richard
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Todd Lehman writes:
> but Jeremy Sproat has written a general-purpose
> platform-independent newsreader in Java, and I think he might be considering
> adding rating capability to it. (Or was that Dan Boger?)
Well, the Java client is mine, but I'm not considering adding a rating
capability until after I can get posting to work (1). Dan is working on a
non-Java client (Perl?), with which he does plan on supporting article rating.
Cheers,
- jsproat
1. Gotta learn cookies. Actually, I just gotta take the time to implement
cookies using java.net.URLConnection, but I've been swamped lately.
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
On Tue, 18 Apr 2000 23:07:52 GMT Todd Lehman <lehman@javanet.com> wrote
concerning 'Re: Article rating (was: Re: the latest news)':
> In lugnet.admin.general, Richard Franks writes:
> > Apologies if this has been asked before and I missed it, but is this
> > available? It sounds *extremely* useful!
>
> Mine is a horrible hack crufted together to run in text mode with Curses on
> my particular home machine, but Jeremy Sproat has written a general-purpose
> platform-independent newsreader in Java, and I think he might be considering
> adding rating capability to it. (Or was that Dan Boger?)
I'm still working on my perl/tk based streamer... it's coming along
slowly, since work keeps bugging me. What about addind the
X-lugnet-rating header to avid.cgi, though? Without it, the client
can rate, but won't be able to see other people's ratings... Also, is
it ok for the client to use the raw.cgi to get specific messages, when
needed? Or is that not a part of the published API?
:)
Dan
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Dan Boger writes:
> I'm still working on my perl/tk based streamer... it's coming along
> slowly, since work keeps bugging me. What about addind the
> X-lugnet-rating header to avid.cgi, though? Without it, the client
> can rate, but won't be able to see other people's ratings...
The avid.cgi script is meant to serve things continuously, not backward in
time, so adding the header to that doesn't fit its design very well. (It
would also add load to it.) (But serving the ratings via something specially
constructed to handle back-in-time or since-some-time queries efficiently
through another separate script is planned.)
> Also, is
> it ok for the client to use the raw.cgi to get specific messages, when
> needed? Or is that not a part of the published API?
It's not a particularly streamlined script like avid.cgi is, but it's not
particularly inefficient either. It's more efficient than the HTML display
of articles, for one thing, but still really intended for only interactive
display.
It doesn't have a published API, and it wasn't intended to be called from
user agents other than web browsers, but if you access it randomly (i.e.,
right when your app needs to display it to you) without hammering on it
(fetching zillions of articles in rapid succesion), it shouldn't be a problem.
In that sense, your agent would be acting like a browser.
For fetching multiple articles, instead of using raw.cgi, open an NNTP
connection and send:
1. GROUP <group> to change into a group, followed by
2. ARTICLE <artnum> to get the article data.
Repeat at 2 if fetching multiple messages from a single group, otherwise
repeat at 1 if fetching multiple messages from different groups.
--Todd
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Richard Franks writes:
> Actually, if that included fuller search capabilities as well (date, author,
> subject, etc) then I'm sure I wouldn't be the only one doing bad Seasame-
> Street
> alien impersonations :)
No, you wouldn't be alone!! I'll join with Big Bird any day for those
enhancements.
-Shiri
| | | | | | | | | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Todd Lehman writes:
> [...]
> Well, maybe it's not something for you then. (Certainly, it's not for
> everyone, because it does take a bit of time and concentration sometimes.)
Yikes, I gotta watch my wording. There's nothing to read between the lines
there -- those are two separate statements. I realise that rating takes time
and that not everyone has time or wants to spend it.
> What if, instead of giving out 0's and 100's, you could give out either a
> 25 or a 75? (Or some other pair?) Would you find that less intimidating?
Intimidating was a poor word choice... I mean, would you find that more
inviting (or less unpleasant)?
--Todd
| | | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Todd Lehman writes:
> Do you think that a standardized rating-recommentation info-page would help?
Maybe something expaining the "judge whatever you feel" philosophy? Rather than
'an average post should have 50', 'a more than average post should have 60'
etc..
> I've been following Richard Franks's suggestion for the past couple of weeks
> and trying to consider 50 an "average post" midpoint.
Erm, I'm flattered! :) But I really didn't mean to say that anyone should give
average posts '50', or any other figure. LUGNET is full of worthy (average)
posts, I totally respect someone who would therefore rate an 'average' post as
40, 30 or whatever.
What I meant was that people will tend to see '50' as an average - eg Scott who
automatically considers <20 as a terrible mark (worthy of justification)..
which is why I feel it would solve a lot of problems to remove the ratings from
general view.. and let people use tables to see the most popular posts.
I would probably rate quite a few more posts if the scores were removed - there
is a lot of fluff around.. but it's *really* not worth it (to me) to make
someone feel disenfranchised(1) by rating an otherwise harmless post as 0.
> Also, do you think that the default rating should be 0 rather than 50? For
> a default of 0 would mean that articles tended almost always to go upwards
> in rating over time, rather than either upwards or downwards -- in other
> words, no one would feel that their post was ever "marked down from a 50 to a
> 30 or 40," but rather that their post was "marked up from a 0 to a 30 or 40."
No - I think the current default=50 paradigm is great - just that by making the
entire focus of the rating mechanism the scores themselves, people
understandably get hung up on them.
Richard
1 - Unwelcome, picked upon, boring, unworthy, afraid to post, etc - some of
these I've seen people compain about, others I've felt myself to varying
extents; as a result of the rating system.
| | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Richard Franks writes:
> > I've been following Richard Franks's suggestion for the past couple of weeks
> > and trying to consider 50 an "average post" midpoint.
>
> Erm, I'm flattered! :) But I really didn't mean to say that anyone should
> give average posts '50', or any other figure. LUGNET is full of worthy
> (average) posts, I totally respect someone who would therefore rate an
> 'average' post as 40, 30 or whatever.
I wanted to try that out, to see how well it worked and, after doing that
statistical analysis last week (in reply to your original message about what
50 could mean) was encouraged that the average average was already near 50
(I think it was 56 or something like that).
So far, it feels comfortable (to me, at least) to mark a 50 if my reaction was
an average or typical reaction (read: the average over all articles I've ever
read, not the predicted reaction of the average person; the ratings are
supposed to be personal recommendations/reactions).
> What I meant was that people will tend to see '50' as an average - eg Scott
> who automatically considers <20 as a terrible mark (worthy of justification)
> ..which is why I feel it would solve a lot of problems to remove the ratings
> from general view.. and let people use tables to see the most popular posts.
You mean...like...don't ever show the composite numbers or raw data? Just
use the composite ratings to produce listings, but never show the numbers?
I wonder how that would affect custom clients (like the ones DanB and JeremyS
are writing) which certainly could benefit from knowing the numbers. I think
in order for those software clients to be most useful, they need that raw data.
Or do you mean hiding the ratings from general view by default, and perhaps
making the user go through some sort of brief overview page in order to enable
the view of the numbers?
> I would probably rate quite a few more posts if the scores were removed -
> there is a lot of fluff around.. but it's *really* not worth it (to me) to
> make someone feel disenfranchised(1) by rating an otherwise harmless post as
> 0.
Hmm. Well, I wouldn't recommend marking a harmless post a 0 -- save 0 for
harmful posts, unless you're restricting yourself to using exclusively 0 or
100. A zero should indicate that the article *shouldn't* have been posted,
for whatever reason -- that it was just pure noise or actually hurtful. Or,
more specifically, that no one should bother to read the article. IMHO, there
is lots and lots of harmless fluff that people should read -- for fun or
entertainment.
For example, when someone puts up photos of a new model that blows everyone
away, and it winds up with a score of, say, 90+, it tends to generate several
"wow, that's cool -- keep it up!" types of messages. I've been marking those
40, 50, or 60 depending on how much new (useful) information they add. If
the reply just says "that's cool" I tend to mark it a 40 (or sometimes 30);
if the reply helped me find something I might've missed, I mark it a 60 (or
sometimes 70). But most replies of that type are just harmless fluff --
people patting each other on the back. Those are important messages to the
community as a whole, even though taken along many are just noise that
doesn't particularly enrich the reading experience. (Who but the poster of
a cool model wants to read 20 "wow, that's cool" replies? But when those
replies talk about things in specific that they like, that becomes much more
helpful and less fluffy to other readers besides the original poster).
> > Also, do you think that the default rating should be 0 rather than 50? For
> > a default of 0 would mean that articles tended almost always to go upwards
> > in rating over time, rather than either upwards or downwards -- in other
> > words, no one would feel that their post was ever "marked down from a 50
> > to a 30 or 40," but rather that their post was "marked up from a 0 to a 30
> > or 40."
>
> No - I think the current default=50 paradigm is great - just that by making
> the entire focus of the rating mechanism the scores themselves, people
> understandably get hung up on them.
What if...hmm...what if, instead of a number 0 to 100, there were a small
horizontal colored bar representing the number graphically?
--Todd
>
> Richard
>
> 1 - Unwelcome, picked upon, boring, unworthy, afraid to post, etc - some of
> these I've seen people compain about, others I've felt myself to varying
> extents; as a result of the rating system.
| | | | | | | | | | | | | | | | | |
| |
|
In lugnet.admin.general, Todd Lehman writes:
> I think in order for those software clients to be most useful, they need that
> raw data.
Agreed - as I see it, the problems created by the rating system aren't because
some posts are bigger than others, but rather because people can't avoid seeing
the ratings that they recieved. Especially when the ratings are for harmless
'me-too' posts - if it really stinks, or is really great.. then it is good when
the author sees the rating!
> > I would probably rate quite a few more posts if the scores were removed -
> > there is a lot of fluff around.. but it's *really* not worth it (to me) to
> > make someone feel disenfranchised(1) by rating an otherwise harmless post as
> > 0.
>
> Hmm. Well, I wouldn't recommend marking a harmless post a 0 -- save 0 for
> harmful posts, unless you're restricting yourself to using exclusively 0 or
> 100. A zero should indicate that the article *shouldn't* have been posted,
Yup - it wasn't a great example.. probably should have made it '30' instead!
> What if...hmm...what if, instead of a number 0 to 100, there were a small
> horizontal colored bar representing the number graphically?
Do you mean in the multiple-post group listings? If it was easy to tell a 35,
from a 50, then the casual (mostly-harmless) authors of me-toos would still
feel bad on a regular basis.
What about a simple 3-colour interface:
00-33% - blue
34-66% - green
67-100%- red
Or to distinguish between real turkeys, and super-posts:
00-15% - black/dark blue
16-29% - blue
30-70% - green
71-84% - red
85-100%- bright red
The latter is an un-even scale - the benefit of it is that most people will get
a uniform grading (less bad feelings).. but it is easy to scan between great
posts and horrid posts.
This might actually be better than removing the ratings from the general
display - the concerns that I had were mainly that 'harmless' fluff posts were
being put to 30-49.. (which is great from a system-sorting POV), but people
were feeling bad because of it.
Richard
| | | | | | | | | | | | | | | | |
Richard Franks wrote:
[snip]
> Or to distinguish between real turkeys, and super-posts:
> 00-15% - black/dark blue
> 16-29% - blue
> 30-70% - green
> 71-84% - red
> 85-100%- bright red
[snip]
I think it should be the other way around.
Red for low bandwidth, violet for high bandwidth.
/Eric McC/
| | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Charles Eric McCarthy writes:
> Richard Franks wrote:
> [snip]
> > Or to distinguish between real turkeys, and super-posts:
> > 00-15% - black/dark blue
> > 16-29% - blue
> > 30-70% - green
> > 71-84% - red
> > 85-100%- bright red
> [snip]
>
> I think it should be the other way around.
> Red for low bandwidth, violet for high bandwidth.
Yeah, the colours are a bit squiffy!
Bronze, light bronze, silver, light gold, gold?
(Silver being light grey or white)
Richard
| | | | | | | | | | | | | | | | | |
In lugnet.admin.general, Charles Eric McCarthy writes:
> Richard Franks wrote:
> [snip]
> > Or to distinguish between real turkeys, and super-posts:
> > 00-15% - black/dark blue
> > 16-29% - blue
> > 30-70% - green
> > 71-84% - red
> > 85-100%- bright red
> [snip]
>
> I think it should be the other way around.
> Red for low bandwidth, violet for high bandwidth.
Blue and purple/violet are automatically out anyway because of traditional link
colorings. Green is out because green text looks horrible on white background.
A statistically significant portion of the population is also some form of
colorblind -- and brightnesses of a single color work just as well for relaying
information on a single scale as two colors. Two colors are most effective
when a two-dimensional scale is needed (not really the case here).
--Todd
| | | | | | |
