[uf-discuss] hReview feedback

[Paul Bryson]

"Craig Ogg" wrote...
> If a friend told you that he would give a movie 3 out of 4 stars,
> would you really ask if he was starting from 0 or 1?   Because, after
> all, if 0 is the lower bound that is a 80% rating (4th out of 5
> possible positions) and if it is 1 is a 75% rating.  Who would want to
> see a movie that only got a "C"? [1]   I think ratings are almost all
> subjective and are all assumed to have the same lower bound which is
> always unstated.

If 0 is the lower bound, then that is a 75% rating.  If 1 is the lower 
bound, then that is 67%, or a difference of -9%.
0...........100%
0..1..2..3..4
1...2...3...4

But that isn't really important.

Oddly, the difference between even a 75% and 80% rating is significant.  In 
the world (and for some reason it is more pronounced in the US) ratings are 
used in a bell system.  People tend to vote towards the middle of their own 
internal scale unless they feel very strongly about something.  And for 
whatever reason, people tend to vote towards around 60-70% as the middle 
instead of 50%.

These two things mean that ratings don't spread out evenly, and so being 
more accurate around this particular range is especially important, not 
necessarily for a single review, but when aggregating or comparing multiple 
reviews.

There are some studies about this, but I can't think of how to track them 
down.  However, you can see some interesting trends on certain sites.  For 
instance,
http://www.resellerratings.com/seller2077.html
is nice because it lists the average store ratings total, high, and low.  In 
this context, 5% would not be a small thing.

> Given the microformats priniciples of covering the 80% case and no one
> seems to be able to come up with examples where people routinely state
> the lowerbound, I would suggest that it be removed from the format.

I am not talking about ONLY lower bounds, but also about providing a good 
way to present information that will likely be read by machines, and I think 
that the lower bound can be done at the same time (although it may not).

> [1] In the US when grading students on a bell curve, 60% = D, 70% = C,
> 80% = B, 90% = A

This may be why ratings are skewed.  People have been conditioned that 50% 
is a failing grade and is completely unacceptable in most ratings systems. 
Where I grew up, there was no D.  Anything below 70% was failing.


Atamido