Youtube switches from Score Voting to Approval Voting

[clay shentrup]

In their product updates they describe it like this:

=> Ratings have changed from the Star system to a binary "Thumbs-Up
‘Like’" / "Thumbs-Down" system. Anything other than a 1- or 5-star
rating is rarely used on YouTube, and so we moved towards a simpler
"Like / Don't Like" model. Liking a video will save it to your “Videos
I like” list.

The official Youtube blog explains that it's because 5-star ratings
dominate, with a small fraction used for the opposite 1-star rating,
and almost nothing intermediate.

http://youtube-global.blogspot.com/2009/09/five-stars-dominate-ratings.html

That post from Youtube's official blog cites this graph of ratings.
http://4.bp.blogspot.com/_0uuFi1arkJE/SrhU9YnkyeI/AAAAAAAAALg/MiOAALxA7-c/s1600-h/+ratings+graph.jpg

Now surely people like Rob Richie would spin that into a narrative of
how Score Voting is doomed because of strategic voting. Of course an
objective person would point out that voters are self-selecting, and
people generally don't feel excited enough to even rate a video if
it's just mediocre.

An additional theory (which is supported by my own behavior) is that
people will often vote when they don't feel that the current average
rating is correct. So if something is at a 3, and I think it should be
at a 4, I'll give it a 5 to move it toward 4 as strongly as I can. And
I'll give it a 1 if it's at a 4 and I think it should be at a 3.

[clay shentrup]

I should add that this last strategy I discussed, where you only give
a max or min in the direction you think the score should go, actually
produces extremely accurate results. Here's a little Ruby program that
tests this.

NUM_VOTERS = 10000

honest_scores = (0...NUM_VOTERS).to_a.map{rand}

honest_sum = 0
strategic_sum = 0.0
sum_of = [0,0,0,0,0]
NUM_VOTERS.times do |i|
score = rand(5)
honest_sum += score
strategic_sum += sum_of[score] > strategic_sum ? 4 : 0
1.upto(4){|i| sum_of[i] += i}
end

puts "Actual sum of scores is #{honest_sum}.\n\
Strategic sum of scores is #{Integer(strategic_sum)}."

===

Sample output is:

Actual sum of scores is 20118.
Strategic sum of scores is 20000.

Actual sum of scores is 19923.
Strategic sum of scores is 20000.

Actual sum of scores is 20364.
Strategic sum of scores is 20000.

Actual sum of scores is 19949.
Strategic sum of scores is 19988.

Actual sum of scores is 19950.
Strategic sum of scores is 19996.

Actual sum of scores is 19781.
Strategic sum of scores is 20012.

Actual sum of scores is 19976.
Strategic sum of scores is 20000.

[clay shentrup]

my response to their blog.

http://youtube-global.blogspot.com/2009/09/five-stars-dominate-ratings.html?showComment=0#c6432592080090720737

[Warren Smith]

I'm actually confused about YouTube.

I thought they did NOT have a 1-to-5-star range voting system, they
had some kind of
arbitrarily-huge-#-stars system because I saw YouTube videos some guy
had rated 28,
etc. Huh?

--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)
and
math.temple.edu/~wds/homepage/works.html

[Sean Walker]

Videos were 1-5. Comments were approval style already.
Perhaps you were thinking of the number of votes, which was included
with the average?

On Apr 14, 3:44 pm, Warren Smith <warren....@gmail.com> wrote:
> I'm actually confused about YouTube.
>
> I thought they did NOT have a 1-to-5-star range voting system, they
> had some kind of
> arbitrarily-huge-#-stars system because I saw YouTube videos some guy
> had rated 28,
> etc.   Huh?
>
> --

> Warren D. Smithhttp://RangeVoting.org <-- add your endorsement (by clicking

[Warren Smith]

ok here is a Youtube video.
http://www.youtube.com/watch?v=hLKXD_Ar5CM

check the comments below the screen. FIrst one is like +10. Next is like +8
etc.

Seems to be no limit.

--
Warren D. Smith

[Sean Walker]

Oh yeah. Comments have always been a display of net approval.

On Apr 14, 4:20 pm, Warren Smith <warren....@gmail.com> wrote:
> ok here is a Youtube video.http://www.youtube.com/watch?v=hLKXD_Ar5CM

>
> check the comments below the screen.   FIrst one is like +10.  Next is like +8
> etc.
>
> Seems to be no limit.
>
> --

> Warren D. Smithhttp://RangeVoting.org <-- add your endorsement (by clicking

[clay shentrup]

yeah, that's approvals minus disapprovals, not a rating.

that underscores my biggest complaint with the new system. if you want
to take away intermediate scores, fine. but keep displaying the
_average_ of the "ratings". in other words, if 30% of people said
thumbs down, and 70% said thumbs up, that's an average of 0.7 (using a
0 for disapprove and 1 for approve, but that's arbitrary).

basically you're just asking, what fraction of people who have voted
on this video have given it a thumbs up?

[clay shentrup]

oops, this line can be completely removed from my program. it is
vestigial.

honest_scores = (0...NUM_VOTERS).to_a.map{rand}

so it's just:

NUM_VOTERS = 10000

[Cathy Woodgold]

On Apr 15, 12:06 am, clay shentrup <thebrokenlad...@gmail.com> wrote:

> that underscores my biggest complaint with the new system. if you want
> to take away intermediate scores, fine. but keep displaying the
> _average_ of the "ratings". in other words, if 30% of people said
> thumbs down, and 70% said thumbs up, that's an average of 0.7 (using a
> 0 for disapprove and 1  for approve, but that's arbitrary).

I agree. Just because voters are restricted to expressing thumbs-up/
thumbs-down
doesn't mean the results presented also have to be that simplified.


--
Subscription settings: http://groups.google.com/group/electionsciencefoundation/subscribe?hl=en

[clay shentrup]

here was my comment on their blog:

http://youtube-global.blogspot.com/2009/09/five-stars-dominate-ratings.html?commentPage=3#c6432592080090720737