@Mark, Ben, Guys, I've posted an article on this issue a few months ago, too: (I wouldn't call this "AI" though)
Monday, July 23, 2012 *Issues with Like-Dislike Voting in Web 2.0 and Social Media, and Various Defects in Social Ranking and Rating Systems - Confused and Vaguge Design and Measure - Psychology of the Crowds - Corrupted Society Preferences and Suggestions. In Facebook, Youtube, Twitter, TV Networks...* http://artificial-mind.blogspot.com/2012/07/issues-with-like-dislike-voting-in-web.html Give it a shot if you wish. IMO this "bug" is everywhere, the entire pop-culture and mass media is "poisoned". For example in my observations at least in Bulgarian TV market, the shows that get the highest rating are the ones who are aired on the assumed as highest-rating TV during prime-time. It doesn't matter what the artistic qualities of the show are. We use to say in discussions with friends that people will watch whatever they are given to watch, the masses/majority "likes it" not because of its intrinsic qualities, but because it's presented/transmitted by an "important"/high-ranked figure (an assumed to be influential media) and because the majority watches it, and the majority believes it's "majority" and wants to be part of the majority. Etc. There's a vicious cycle of popular things getting more popular and "viral", because they are popular at a given initial moment in a given coordinates, and the other items, which are not popular in that very moment can't never catch up, because the viewers initially see only the items on top and they click only items on top. Users don't have enough of a patience to examine all comments, all TV channels, all similar videos, all possibilities, and the low rating/viewing count etc. of the others bias their opinion, they assume that if it's not popular, it's "not important" or "not good. More than 5-10-20 comments or options are already too much. Once a particular "critical mass" is crossed, the item goes even more viral, because it seems more important and is copied again and again etc., goes to TVs and the "popularity" force over the primarily neutral viewers gets even stronger - they wouldn't see this, if it wasn't aired on NBC, ABC, National TV, MTV, wouldn't go, search and pick it themselves, they are just consuming whatever "food" they're given. Also the people who tend to vote or comment, and the ones who tend to share are particular type of users (as Ben notes about the content of the usual top comments on Youtube), there are many biases, and the measures are very vague; there are 1000 reasons to like or dislike a video, which are all mixed up, and the votes of all kinds of haters, children, old men etc. are treated as equal - sorry, but they are not equal. ... My solution for the problem, especially on Video Sharing sites would be to make voting more granular, to explicitly mark what precisely do you like. (Something that is probably too elaborate to be popular), there must be many ratings, a multi-dimensional rating. The engagement (time spent on video and the way it's spend) should be also used in the measure. I.e. somebody may come and just "dislike", without watching. Or if you have seen just 1/10 of the piece, your vote's weight is just 1/10, or it depends on the votes of the others (and what they have seen), or you may vote for different temporal parts of the video, too. I agree with Ben's proposal for those networks and subgroups - particular types of videos are liked/interesting for particular groups of people, and they are obviously not interesting for other groups, so votes from different groups may be excluded in particular presentations of the rating (the rating shouldn't be just one single-number for all, it should depend on the viewer and on the aspects he's interested in, and the opinions of which other viewers etc.) One may choose to see different aspects, different votes, different groups of voters (demographically, geographically; also based on their clustering based on previous votes and their interests); also perhaps in different periods of time etc. (There are some of those in movie ratings, products ratings.) It matters who's voting, what's his experience, knowledge, preferences, history, age etc. and a dynamic view should be offered. Also the top-comments shouldn't strictly go to the top, and the most "popular" items shouldn't go on top of a list with proposals, first page etc. There should be some randomness and to make users do a little search (checking different possibilities) and then decide, not just blindly consume. Another idea is that the rating gets meaningful after the same user rates at least N other items, including both N likes and N dislikes which are in a cluster/a related/similar topic/artist/genre (other pop-songs, videos, films, products), so that a relative measure is created for a user. Cheers, ** Todor "Tosh" Arnaudov ** * -- Twenkid Research:* http://research.twenkid.com -- *Self-Improving General Intelligence Conference*: http://artificial-mind.blogspot.com/2012/07/news-sigi-2012-1-first-sigi-agi.html *-- Todor Arnaudov's Researches Blog**: *http://artificial-mind.blogspot.com ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
