For me this is really surprising as Like has very little value, what little value is has is confounded by it lacks any explicit understanding of intent. Search is about finding what is being sought, which is much harder than it sounds, particularly with massive amounts of information, or when searching across contexts and influences. Like really doesn't add much of anything of value to this. Why somebody likes something is really important to understand, or more importantly even if a person actually likes what they placed a like on, or if they were using it as a proxy for a bookmark to hold on to something so to return later, or even if the Like is a social statement.
One of my trips to California I was with friends and we were trying to sort out where to grab something to eat. One friend suggested In-N-Out, she figured it was well liked and even the guy who is vegetarian would be in on it as he liked it on Facebook (she remembered). The vegetarian in the group strongly preferred not to go there and wanted an option with better vegetarian offerings. When he was asked why he put a Facebook Like on it he said, "I like hanging out with my friends there as it makes them happy, but I usually have eaten before, or will after. Now I am hungry and wish to eat, so I really prefer something other than In-N-Out." This triggered everybody talking about their doing similar things with Like in Facebook, which really didn't mean they liked what they clicked "Like" on.
Facebook Like, much like the often problematic star ratings, adds more ambiguity (or another value point that has no clear meaning that can be reliably used for search or predictors). My favorite recommendations from Facebook are those similar to "Those who like food also like sleep.", which gives me the option to like sleep. (We can cure cancer if we keep this intelligent thinking up.)
What is Next? The Past!
So, if this augmented ambiguity from using Facebook Like in search is problematic leads you to think, "What is better?" Well, a look back to 2005 or 2006 at Yahoo! is a very good place to start. Somewhere in this timeframe Yahoo Search did something smart, no freakishly smart (actually connecting two things together that made a giant difference for search). Yahoo! had its own social bookmarking service "My Web", which was somewhat similar to Delicious (which Yahoo acquired). The second version of MyWeb (MyWeb2) made it easy to see one's own bookmarks that you yourself tagged in your own context, your friends bookmarks they had tagged with their tag terms in their context, and everybody's. Yahoo! incorporated the tags and social connections from MyWeb2 into their search. This dramatically improved the search, if you were using MyWeb2 and particularly if you had stated people you were connected to.
At this point Yahoo! not only caught up to Google but passed it by a large margin for me. Why? Google was very good at finding good results, often good enough. Yahoo! with MyWeb2 built in and using my 60 to 70 people I was connected to started surfacing exactly what I was looking for. This was happening regularly. This was search Nirvana. Let's step back slightly to understand why.
Proper Social Understandings Improve Search Precision
One of the interesting things about people tagging content to store it in services like MyWeb2 or Delicious (or any other folksonomy tagging service) is people almost alway only tag things they have interest in. Based on the assumption (which holds up well) that people hold on to thing they like, but when they drift from that they usually will add tags that state that deference.
Search is difficult because of contextual influences and ambiguity. Having tagging done by people whom you know can help with that contextualization. People whom you know having tagged things around what you are seeking and use the terms in similar manner to the way you do has value. Well, no not really, it has insanely great value. The key is sorting out similar affinities (as close as possible) and similar term use helps to further remove ambiguity, which becomes clearer when you can parse things through the lens of a granular social network. With just 60 to 70 people my world of search was turned upside down in a very positive way. All search results that had been bookmarked and tagged by people I was connected to were annotated with their their name and often tags.
This giant step forward for Yahoo! did not last long as after a few months the experiment was over and Yahoo search returned to being not as helpful as Google search, which is just good enough.
The Yahoo experiment was not perfect, but it was much closer than most anything else to that point. Holding it back was the lack of people you were connected to. The more people you were connected to, to some degree, was helpful. Also, very few people knew about this experiment (it didn't seem like an experiment at the time, as it seemed it could only grow, but Yahoo really didn't seem to know how to get the word out or talk about this value, it was an information geek thing (yes, I could fall into that grouping). But, the piece missing that would have been most helpful, was the ability to garden and craft your relationships to those with whom you connected.
The gardening and contextualizing those with whom you are connected is really powerful. It doesn't need to be publicly exposed but the tools and service can make giant leaps forward if we have this. Most of this contextualization is assumed by tools and services, but having explicit crafting takes the guessing out. Being able to add fuzzy (roughly defined) semantic terms to attract what you value from that person closer while keeping the things of less of value at bey, can be helpful. This is core to the model of attraction (draft) idea that has been my frame for much around me for years. Being able to tag or annotate "Jim" with cycling, food, social search, design, and baseball will help search bring things roughly related to those topics or terms close to me, but may not give as high of relevance for his passion for early 1990s dot matrix printers nor Hobbits.
The next step for this as in terms of products and services also has happened. An enterprise social bookmarking service, Connectbeam (now gone) took the next step (Lotus Dogear, now Lotus Bookmarks in their Connections tools is somewhat similar) by bringing this same social tagging into the work environment and then surfacing that added value into search results. What set Connectbeam apart from others doing similar efforts was it helped people understand the social components better than most. They had some really good social interaction designs around the connecting people, that really started to get at some of the tough nuances that are really hard to crack outside the early adopter types using service (only 5 to 15% of most orgs will fall into that early adopter mindset, the rest are really lost with this). This crafting and understanding social interactions allowed Connectbeam to have the potential to drastically improve search, (search is a very expensive and painful proposition at every organization I have run across). The social interactions needed for comfort, familiarity, and producing value is central to getting any service right, but the hurdle is big but there is a large positive value if you get that right for social tagging. Sadly, Scuttle and thin not well thought through attempts at social tagging really do not add to much either.
Spending time to understand the keys to getting it right and selecting tools that do it well or working with vendors to get there will pay off.