To Like or Not to Like: Advances in Rating & Social Voting Systems

I've been thinking on and off about the FB Like button for weeks now and a huge flaw just occurred to me. Here it is in plain English. There was already the problem of adding the Like button, right, to your site: i.e. what is it that people are liking? When I Like CNN, am I liking the website, the business, am I liking a set of different features, etc.? That was a problem since the beginning, the way that you classify what it is that people are liking.

I found another problem. What does liking do? What use is FB's huge dataset of things people Like going to do? What I mean to say is that Liking in this manner is extremely NON-SPECIFIC. The problem is a rating problem and a social voting problem that no one has yet solved.

The gist of it is, right now you either Like or you don't Like (not the same as Disliking, not Liking is just not clicking the like button, it's an empty set). Already it would be better if we could vote in terms of Like, Dislike, and Neutral. But even then, people in the field of sentiment analysis will tell you that they encounter problems even with a 3-pronged system of evaluations.

The problem is there is no formal language for encoding people's opinions. What would help FB would be WHAT someone Liked precisely.. What is it about Chum.ly that so-and-so likes, is it the service, is it because they have a friend who works there or a family member who blogs there? Do they merely like the design of the site? or they like the logo?

And when someone Likes something, couldn't we be more specific? Did the web page MOVE them? was it an emotion that led to the Like Behavior? a thought, a belief, an impulse? did the individual have a compulsion to press the Like button merely by virtue of social influence? (In other words, if 5 million people Liked it, I can be influenced to like it based on the sheer number of social votes).

There is nothing to compare between different Likes, they are all the same, you either Like or you do not in fact press the Like button. This is pretty useless information and won't work in the long-run - in my book! - with respect to the goal of "instant personalization".

To truly personalize, you need more information than just a Like. If I Liked the New York Times website, or a specific article.. that information is relatively useless, not completely useless, but relatively so. Did I find the information relevant, timely, useful, interesting, fun, particularly informative? And what did I Like, was it the information conveyed through the text of the article or the design of the site, or an image or a video in the blog I just read?

So we're dealing with the current "ceiling", if you will, of rating systems online, systems of social voting. Digg.com has the same limitations, if you ask me. Thus far, after using Digg for a month or so, it still cannot give me a single useful recommendation, and I fear the same is true of many more recommendation systems online.

What we need is a formal language with which to ENCODE precise information, and extensive information, to go with our individual votes. It should be as clear as possible WHAT I am voting for, and I should have at minimum 3 choices, Yes, No, and Neutral.. but ideally, I would have like 12 or 15, 20 different points on a scale, or a linguistic form for expressing ratings, opinions, for comparison between votes.

What do you think about this? Can we invent a perfect voting system and be a million light-years ahead of everybody else? We need to know WHY PEOPLE LIKE x, y, and z. Or else you're just collecting a bunch of web pages.. with little relevance, I mean concretely.. You've gone through all this trouble to implement a new protocol.. and you never thought to make it a little more precise on the voting side?

The problem is with user intention when the user votes. And then, there's no way currently to encode metadata on social votes like pressing the Like Button.

One tool in particular seems to me to be a storehouse of much potential: Likebutton.me. This site, apparently, can tell you what your friends are liking. I think you can just enter the name of a Facebook user and you can get a list somehow of what they have Liked.

This means that anyone can use the potential of the Like button and Open Graph Protocol for their OWN personalization agendas. But what we need is someone somewhere who aggregates Like votes and adds metadata to the information, like a online system of annotation for Like button clicks that can add extra input, by either interviewing people, or I don't know how. The problem is we need answers to these questions: What are people in effect Liking? What do they like about it? Why do they like it? What is the user intending when he or she votes? What is the voter SAYING? i.e. what does such and such a vote entail as a SIGNAL (in itself)?