the first one is whether ratings without comments are useful. I think they are, but I don't really want to elaborate on that...
the second one is how to improve the rating system. It turns out I've been thinking about it a bit, and here's my take on that topic. Quite clearly a 10 from someone rating every picture a 10 is worth much less than a 10 from someone who gives on average a rating of 5. The situation is clearly the same for low ratings, but due to basic human psychology, this has already been pointed out. So what I suggest is that besides the "raw" ratings, each picture gets a "renormalized" rating, calculated as follows:
for each rating, substract the average rating given by the user who rated the picture (and possibly divide by the variance of its ratings). This will result in positive ratings (above average) and in negative ratings (below average). This measure is much more robust than the current one with respect to being biased by someone with a few accounts giving extremely high (or low) ratings.
Changes to Rating System
in PhotoNet Site Help
Posted
There are obviously 2 issues:
the first one is whether ratings without comments are useful. I think they are, but I don't really want to elaborate on that...
the second one is how to improve the rating system. It turns out I've been thinking about it a bit, and here's my take on that topic. Quite clearly a 10 from someone rating every picture a 10 is worth much less than a 10 from someone who gives on average a rating of 5. The situation is clearly the same for low ratings, but due to basic human psychology, this has already been pointed out. So what I suggest is that besides the "raw" ratings, each picture gets a "renormalized" rating, calculated as follows:
for each rating, substract the average rating given by the user who rated the picture (and possibly divide by the variance of its ratings). This will result in positive ratings (above average) and in negative ratings (below average). This measure is much more robust than the current one with respect to being biased by someone with a few accounts giving extremely high (or low) ratings.
just my 2c.
Olivier