It seems that ever other website I stumble upon these days has a recommendation feature be it for songs, friends, products, restaurants it doesn't matter - there is a recommendation engine for everything. Many of these engines use variants of collaborative filtering and often preface their recommendation by stating "people who bought x also bought y" or "You may also like/know..." etc
There are times when I'm looking to make a decision and I find myself asking, "what would somebody like me want or do in this situation?" Recommendations engines first look to those who have come before me to make an intelligent suggestion (who the hell are these people?) and now they are starting to look for answers in social networks. The assumption being, if a friend likes something maybe I will also like it. Come on, give me a break. While I see some value in this approach I know my friends don't have all the answers and I know certain topics each friend is an authority on but each friend is certainly not an authority on all topics. Also, let's not forget there are topics I wouldn't want them knowing I am asking about and topics for which they won't digitize their experiences.
There are a lot of companies circling these problems but there is a piece of the puzzle that I haven't heard much talk about - it's personal preferences i.e. "why" did 'somebody' 'like that' 'thing'? Identifying personal preferences is a difficult problem. Preferences often change based on mood, location, time of day etc etc Take deciding where to eat dinner as a simple example: who am I planning to eat with, are they driving me home, who's buying, what did I have for lunch that day, are the kids joining us, how much time do we have etc etc. The answer to each question will alter the final decision and ultimately where I decide eat dinner. Furthermore there is 'brand performance' and how well it preforms against my personal preferences for in this case a restaurant that has vegan options, is kid friendly, is new/hot, etc etc
It isn't enough to know my friend "Kathy dines at Loving Hut Restaurant". Companies that are pioneering the semantic web are marking up the Subject, Predicate, and Objects triples (SPOs if you will). e.g. "Kathy", "dines at", "Loving Hut Restaurant" What I really want to know is 'why' does my friend Kathy dine at Loving Hut? Is it the all Vegan menu, the convenient location on University Avenue, the fact that it's kid friendly, has interesting tea menu, is spotless..... In order to infer 'why' Kathy likes Loving Hut one needs to chain together SPOs e.g. 'Kathy' 'is' 'Vegan', 'Kathy "likes' Clean Bathrooms' and "Loving Hut' 'has' 'Clean Bathrooms' and 'Loving Hut' 'is in' 'Palo Alto' etc etc
Marking up SPOs may be an essential first step to identifying the connections between things but IM(very unscientific)HO it's ~3 to 7 times more valuable to know 'why' the connections exist and how strong the connections are. It's the history and circumstances in which 'things' are related that holds the value- more so than the relationship itself.
Ok, I'm rambling a bit but this leads me to believe that this 'history of circumstance' for people, places, things and their relationships with one and other is where the value is. Organizing and tapping this knowledge is where my cycles are focused. In the meantime I'll ask my friends why they like the restaurants they recommend to me.