| |
In lugnet.admin.general, Steve Bliss writes:
> > In other words, the "one vote from the system of zero" is only conceptual
> > -- it produces a needed counter-behavior in the voting, but it doesn't
> > literally and mechanically need to be carried out...just mathematically
> > carried out.
>
> And it at least as much sense to have an implicit 'indifferent' vote • ^^^^^^^^^^^^^
> from the system as it does to have a special rule for the case of zero
> votes.
Steve, I'm glad you mentioned that -- because from a "makes sense" point of
view, I think it's much harder to explain to voters that the article's score
is sum(V,1,n)/(n+1) [n voters] than it is simply to explain that the article's
score is simply sum(V,1,n)/n [n-1 human voters & 1 computer voter]. They're
*exactly* the same thing mathematically, but it should "look and feel" like
the latter.
Some empirical evidence to support that: I ran some simulations and also
poked around setting some votes on various articles, and let me say, boy does
it look and feel strange to start with a score of 0 and 0 voters, and then
cast a vote of x and have it immediately show up as a score of x/2 with 1
voter. I think to myself, "Wha? I just cast a vote of +100...why does it
say +50? That's a bug." Of course it's not really a bug -- but it feels
that way; the system's vote of 0 is "hidden" there implicitly in the division
by n+1.
So I think it makes sense for the mathematical to meet the practical halfway:
What I've got running now does this: If there are zero human voters, then
the score is simply defined to be zero. If there are n human voters, then
the number of voters is displayed as n+1. So now when I go to a previously
unvoted message and I see Score:0 Votes:0 and I cast a vote of +100, my
immediate feedback from the system is now Score:+50 Votes:2. That now "feels
like it makes sense." Thus, the "oddity" of the dampener has switched from:
- Why did it divide my +100 by 2 and get +50 when I'm the only voter?
to:
- Hmm, why now are two voters shown? Well, at least the math there makes
sense, assuming the other voter cast a 0.
(It's the lesser of two evils.) Voters will just have to be aware that
the system auto-votes a 0 at the same time as the first human voter.
Now, if a third voter (a second human voter, that is) shows up and also casts
a +100, then they'll see Score:+67 Votes:3. And if a fourth voter (a third
human voter, that is) shows up and also casts a +100, then they'll see
Score:+75 Votes:4. The math always "lines up" this way, and the only anomaly
is the jump from 0 to 2 upon the first vote. (Easy to explain to voters.)
Anyway, so much for those empirical tests. Now on to the domain of values --
the user interface.
For math's sake, it's easiest to work out the system in terms of real numbers
on the unit interval [-1,+1]. But from a user standpoint, numbers like -.45,
+.37, +.09, and -.88 are hard to parse quickly. A range of [-10,+10] is even
worse, because that gives numbers like -4.5, +3.7, +0.9, and -8.8. I think
the most intuitive range is [-100,+100], which gives "friendly" numbers like
+45, +37, +9, and -88, which look and feel a lot like percentages, especially
up in the 90's. And no unsightly decimal point (which, uh, isn't even a '.'
symbol in most parts of the world :)...
And I'm pretty excited that it won't be too difficult for an article to
achieve a +90 rating, yet it will still be "impossible" for an article ever
to achieve the theoretical maximum +100 rating. The most we'll probably ever
see in practice is +97 or +98... It would probably take something like LEGO
Direct announcing bulk purchases of every element, and 199 people all rating
the message +100, in order to get a composite score of +100 (it would be 99.5
internally, but would display as +100 in most listings).
I want the theoretical upper bound of +100 to be *extremely* elusive, just
like trying to get to Warp 10. :)
OK, now for the *actual* user interface. A slider which glides between -100
and +100 would be perfect, but that's not part of HTML forms, so it has to
be a drop-down list box with a few choices. (A fill-in-the-box might kinda
work, but that's prolly too much work for the user.)
I'm weary of 33's and 66's because they don't seem "round" enough, so what
I've got now in the test thingie is the 6 values of +100, +75, +50, +25, -25,
-50, -75, and -100. (0 is also a list item, but it means "I don't care" or
"erase my earlier vote" and displays as three dashes).
Here are the tentative text labels for the 7 tentative settings:
+100: Absolutely Outstanding!!!
+75: Really Good!
+50: Good
+25: Above Average
- - -
-25: Below Average
-50: Bad
-75: Really Bad!
-100: Absolutely Rotten Bad!!!
I don't know how to avoid judgment words like "good" and "bad" and still
have the labels "make quick sense." Any suggestions?
+100 and -100 also should probably suggest a superlative condition. I'm
not sure that "Absolutely Outstanding" suggests the highest possible
condition, although it does sound pretty high. Is there any two-word
phrase which means "Abso-freaking-lutely wicked incredibly mindblowingly
heart-poundingly jaw-droppingly awesome!!!"? (If not, that's OK.)
--Todd
p.s. I did thoroughly read all the FAQ items and the Moderation rationale
and rules at Slashdot, and while that system works great for Slashdot,
something quite different is needed here. I also scoured a Statistics text
any sorts of hints about this sort of thing, but it came up blank. I'm
becoming pretty happy, however, with this system described above -- it's
very easy to use on the interface side, it's very light on system resources,
and it actually ends up employing the classic average function (albeit in a
novel way).
|
|
Message has 1 Reply:
 | | Re: Article scoring
|
| (...) (hey, anything to help out) [snipped all the stuff I agree with] (...) How about radio buttons instead of a drop-down list? Seems more approachable for the non-geeks among us. (...) You could let people choose a score from 1 to 10. This is (...) (25 years ago, 8-Mar-00, to lugnet.admin.general)
|
Message is in Reply To:
| | Re: Article scoring
|
| (...) And it at least as much sense to have an implicit 'indifferent' vote from the system as it does to have a special rule for the case of zero votes. Steve (25 years ago, 6-Mar-00, to lugnet.admin.general)
|
52 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
