'Likes' are a Flawed Currency

Anecdotally, I used to surf the internet a lot in high school in the late 90's/early 00's and not a ton of my friends were online, so there wasn't much shared link wise among us. Then, in college, there was a group forum that a bunch of us were a part of and one of the more popular parts of it were the links people posted. Then, after college digg, Facebook, reddit, pintrest and twitter all found a place amongst the population at large for sharing things.

I guess for most people sharing is not about currency or ranking at all, it's about sharing something interesting. It's more akin to telling a friend about music that you like, a book that you are reading, or a newspaper article you thought was insightful. When you think of those things, it is rarely about ranking them, it is almost always about sharing.

People share with other people because it makes their life better or their friends' lives better. They do it because they care about their friends, not because they care about you or what you created.

Not only is 10 an arbitrary number, but it sounds like a lot less than some people would like to give, and a lot more than most people will give any way.

Why not have the best of both worlds by using a PageRank-alike system to determine the weight of any given like.

That way, if someone is spamming likes they won't affect any one item very much compared to someone who likes only sparingly. You can go the whole hog and have the system where heavily liked users are more powerful than non-heavily liked users also.

The simple solution is to post-process likes in a different way. See http://www.evanmiller.org/how-not-to-sort-by-average-rating.... for a suggestion that makes them more like a rating, despite the fact that some people use them differently than others.

But we can do other things. Here is a fun one. If a person likes 2 different photos, we can say that is a connection between them of strength 1/log(# photos liked by that person). The total connection between any two photos is the sum of the connections. From each photo we can now find other photos that are in some sense likely to be liked by people who like the first. People enjoy following connections like these.

This system is similar in principle to the one that I wrote about at http://bentilly.blogspot.com/2011/02/finding-related-items.h.... Which drove a noticeable sales bump.

So the challenge is not to remove likes as a mechanism because it doesn't give you the information that you want. It is to find ways to process the data you get from it to pull out useful information.

Capping it artificially stifles users who may actually enjoy photography more than others. A critic would presumable like photos far more often than a casual user.

"Please like us on facebook to unlock this functionality of software that you thought you already had full access to."

"To enter this contest, like us on facebook."

Now, the number of like's your product has is completely uncorrelated to the number of people who actually "Like" your product. You are effectively saying "look at all of these people we convinced to arbitrarily click a button"

People are strongly motivated to look good in the eyes of others, so most people don't like everything, that's just creepy.

It alerts everyone that I'm friends with that I "liked" something.

I'm not going to go around handing out likes for things because it will make me seem annoying and not genuine. I also don't want to represent myself as someone that likes Wal-Mart or Samsung unless those are things I actually like.

Another question I'd pose: what's the real goal of a "like" in the various contexts it is used? In many Facebook applications (e.g. liking a photo, comment, etc.), it appears to be as much about filling in missing social cueing -- a one-bit emoticon, if you will -- than about informing relevance. In other contexts on Facebook, there's a definite notion of relevance (e.g. liking a page).

Given that, I suppose I must ask: how much additional value can really be added to this single low-information stream vs. simply using additional active or passive metrics to inform relevance? Is "like dilution" really such a big problem? If it is, there are other approaches to consider.

Instead of creating artificial scarcity it may be worth using another IR technique and inverse-weight a signal based on dilution criteria. I.e. if one user expends LOTS of "likes", they effectively place a lower value on each of those. If another user uses "likes" very infrequently, we can infer those carry more weight. A straw-man example: turning the proposed technique on its head, the summed value of a user's likes might be capped at 10 points each [week, etc.], with no limit on the number of likes. The value of any like within that window is 10/[number of likes this week]. Users can click the "recommend/like/etc" button as many times as they want, and the system will dutifully record and interpret that data according to the site's needs.

For one thing we need to pin down the word "like", which means something specific on Facebook and often something different in other places unless those places are playing the Facebook like game.

My gut feeling about social is that it a game of statistical sampling so it never hurts to get more data, you can always decide to throw some of it out later. So getting as many interactions as possible makes sense.

* Discovery: Good but unpopular content is being buried by popular content

* The noise makers are the taste makers: What is popular is being decided by the most active users

* Popular begets more popular: Rankings that are based on popularity only reinforces the popularity of top-ranked items.

Some communities may decide that this is not a problem and therefore ranking content by Likes works as is.

However if you want to address those issues, then yes, rationing the available Likes is one way to approach it. You could also weight the ranking formulas by percentage of Likes per user (ie. users that have older accounts and give out less Likes overall are weighted differently)

Likes that cause donations to go to your favorite charity!

(I am an acquaintance of one of the founders, but not related in any other way.)

This just puts the problem somewhere else.

Now, the new problem is a follower becoming "a flawed" currency. People who try to bot likes will now bot followers to get more likes.

Is there evidence that a significant # of users are doing this?

You wish they were "recommendations" instead. But this is already a solved problem. When I like something enough that I want to recommend it to someone else, I use email. Or I call. Or I mention it in the hallway. Or I comment about it on Google+.

This type of recommendation existed long before we were born. It doesn't really need any help from you, but if you want to add a "recommendation" feature, go for it. But this concept doesn't replace the completely different concept you might call "like-based recommendations". This is a new concept that wasn't possible before, and now you're talking about crippling it by setting artificial limits!