rating engines...

there is a problem on the web today - it is impossible to find what might exactly match your requirements, its easy to find what everyone likes, but not what people like you like.

there are so many rating engines out there, the most used is maybe the FaceBook 'Like' engine. Rating engines in todays Web 2.0 world are very important...why?

1. the avenues for posting stuff have increased.

2. the ease of capturing information and posting it has increased.

3. its easy to find an audience

4. Because of 1, 2 and 3, the inherent vanity of people have suddenly found an outlet.

and because of all this, the amount of crap has increased exponentially on the web. but then what is crap for me, might not be for you.

so how do you find out what is not crap for you? and also how do you also find out what is not crap in eyes of people who know about this crap, the SME's so as to say.

and that is exactly why rating engines are important. the ones that i know of, simply count the number of likes / dislikes. but that has one big problem, the vote of 'i know shit about this subject' counts the same as of 'i won a nobel prize in this subject'.

so my solution, weightages. Every person should have a weightage, so if my weightage in a certain area is high, the vote i give counts for more then someone else who has less weightage.

but then how does weightage get decided for me, how does the engine know that i know more about this subject then someone else. but before that a little more explanation.

so i post a picture, or write a blog, which is liked by 100^x people (really big number meaning it was really liked), my weightage increases by x. so next time i like someone's picture or blog, my weightage is x (more then other's whoes pictures or posts were not that liked), and consequently, the picture or blogpost i am liking gets a higher rating.

but then the problem remains, how do you determine popularity based on mass liking and based on critical liking, today only the first gets importance.

So i take a picture of a cat, which is very cute and all the girls and animal lovers love it, but its not that great a picture, and my pics gets liked by 100 people (all simple beings who have average knowledge about photograpy, are normal humans), my pic gets a rating of 100. but on the other hand, if i take a time freeze pic of a cat jumping, and apply some techniques to it which makes the pic a really good photograph from a professionals point of view but not really cute. my pic is like by a couple of photgrapher who have a combined weightage of 100, but not by any of the simple beings, well both pictures then have equal weightage. Which means the most popular, and the most critically acclaimed have same weightages.

and how does the system know that i know more about photography then all the amateur photographers who just got a nice DSLR, well cyclic weightages. the total weightage given to a picture i liked will in turn affect my weightage as well, because if a picture i liked was liked by a 1000 people it basically says that my choice matches those 1000 people.

so for this to work, it has to be cyclic, the photographer liking my picture and then my pictures weightage hitting the roof later on should in return affect the photographers weightage as well, because his choice is good. Which basically means each weightage will in turn affect all the weightages in the system.

that is where the technical difficulty comes, imagine the computing power you need for this.

also as a last point, i am not sure if there should be a negative rating in the system, mathematically that is.

i would love to discuss this, its an idea which has been with me for sometime now and might be fundamentally wrong, please let me know if it is.

PS: a major problem i see in this working is discoverability, in the sense, i can like something only if i come across it. and for that there has to be a central place where i look for everything, a single platform. that today is maybe google, so the rating engine has to be built into google. maybe it already is which means as usual google beats me in implementation... :-)

also this has to work with tagging maybe, because then only can you categorize. Although am still not sure if categorization is required at all, or just the rating system will take care of it. well maybe it is, not sure.

the model derives a lot form darwins laws btw...