I've been thinking about the fairness of the c-index system. according to the FAQs "The c-index is a number between zero and 100 that aims to give a rough idea of the quality of an image. The primary use of the c-index is to enable users to sort galleries in order to bring the best images to the top." Thinking about the system and how it works, i believe it actual may not be as fair as it should be. Allow me to explain: The c-index is based on the votes it gets (needing at least 5), views and downloads, correct? I find problems in the way these individual rankings work together to create an overall rating.
First-off, there is a lot of grey area in the area of voting. obviously a 10 is 'perfect' but what exactly is a perfect image? what does a 5 mean? one might accept it to mean that it has as many 'good' aspects as 'bad'? however, therein lies a problem. for one, there is a lot of it has to do with preferences. for example, if i prefer fractal images and another person prefers photography, we might give different ratings. i might prefer fractals because i mostly do fractals, as a result, i might know what i should be 'looking for' as a bias which to rate the image, while another person who isn't as familiar might not be basing his rating on the 'correct' priciples. to sum it up, preference of image 'genre' plays a large part. also, what the numbers 1-10 mean. does a 9 mean perfect except for maybe a little fuzzy/blurred section? is an 8 fuzzy and excess noise? i believe the fact that a number scale is used to grade, creates a canyon of grey area, an image that is an 8 to me might be a 10 to someone else.
I also find issues within using downloads and views as a factor. I may need to view the image to vote on it, but just because i think its a good image, doesn't mean i want to download it. maybe i never download images, i just rate them. however, i am hurting their rating even though i think it is good because i didn't download it. the only people who download the image are the people who might want to use it as a desktop background or rework it possibly. hbut not every person who look at my image wants it on their desktop. conclusively, using the 'ratio' of downloads versus views can ultimately be an image rating killer.
I propose several possible solutions:
1) users have a prefence profile. possible bases for prefernces: listing image genres from most favorite to 'least'; preference of image dynamics (ex: i like darker colored images over brighter colored images). these are just some examples of preference possiblilties.
2) 'weighted' impact of rating. if i am very skilled at making fractal images and normally upload them, then maybe my vote on an image should 'count more' than someone who is more involved in photography. this is simply for some above issues i pointed out (knowing what to look for, experience in the genre). at the moment i am not sure how that would work out because of the random voting booth and such, that would be something that the specifics would have to be discussed. but another benefit is though that the caedes member who have been here longer have more of an image than the newer members. currently with the way the system works, a new member has as much 'say' in the rating as the oldest, which is sometimes a problem.
people rate images sometimes because of quality. othe rtimes, because someone really likes the image, they give it a ten, just because they like it. as i quoted above: "The c-index...aims to give a rough idea of the quality of an image..." just because a bunch of people like an image and give it a ten, doesn't mean is a good 'quality' image.
First-off, there is a lot of grey area in the area of voting. obviously a 10 is 'perfect' but what exactly is a perfect image? what does a 5 mean? one might accept it to mean that it has as many 'good' aspects as 'bad'? however, therein lies a problem. for one, there is a lot of it has to do with preferences. for example, if i prefer fractal images and another person prefers photography, we might give different ratings. i might prefer fractals because i mostly do fractals, as a result, i might know what i should be 'looking for' as a bias which to rate the image, while another person who isn't as familiar might not be basing his rating on the 'correct' priciples. to sum it up, preference of image 'genre' plays a large part. also, what the numbers 1-10 mean. does a 9 mean perfect except for maybe a little fuzzy/blurred section? is an 8 fuzzy and excess noise? i believe the fact that a number scale is used to grade, creates a canyon of grey area, an image that is an 8 to me might be a 10 to someone else.
I also find issues within using downloads and views as a factor. I may need to view the image to vote on it, but just because i think its a good image, doesn't mean i want to download it. maybe i never download images, i just rate them. however, i am hurting their rating even though i think it is good because i didn't download it. the only people who download the image are the people who might want to use it as a desktop background or rework it possibly. hbut not every person who look at my image wants it on their desktop. conclusively, using the 'ratio' of downloads versus views can ultimately be an image rating killer.
I propose several possible solutions:
1) users have a prefence profile. possible bases for prefernces: listing image genres from most favorite to 'least'; preference of image dynamics (ex: i like darker colored images over brighter colored images). these are just some examples of preference possiblilties.
2) 'weighted' impact of rating. if i am very skilled at making fractal images and normally upload them, then maybe my vote on an image should 'count more' than someone who is more involved in photography. this is simply for some above issues i pointed out (knowing what to look for, experience in the genre). at the moment i am not sure how that would work out because of the random voting booth and such, that would be something that the specifics would have to be discussed. but another benefit is though that the caedes member who have been here longer have more of an image than the newer members. currently with the way the system works, a new member has as much 'say' in the rating as the oldest, which is sometimes a problem.
people rate images sometimes because of quality. othe rtimes, because someone really likes the image, they give it a ten, just because they like it. as i quoted above: "The c-index...aims to give a rough idea of the quality of an image..." just because a bunch of people like an image and give it a ten, doesn't mean is a good 'quality' image.
there rests my case.