Ambient Recommendation
Last week I was in Portland, Oregon—a city I was visiting for the first time, and was mostly unfamiliar with. I hadn’t done much advance research since the trip was very impromptu and I wasn’t going to be there long, but I was interested in trying to visit at least a few good restaurants during my stay. On a couple of afternoons I was downtown, wandering by myself and looking for lunch. One of the days I tried using both Yelp’s iPhone app and another iPhone app called GoodRec to see if I could uncover any interesting places nearby. They both pointed out a lot of options, but the experience left me a bit nonplussed for two reasons:
- Both presented me with too many options. I could see that there were a lot of places to eat around me—I didn’t need an iPhone app to tell me that.
- Both had the traditional methods of narrowing your options (cuisine type, for example), but none of them quite applied to me. I was just looking to discover somewhere new and cool, somewhere that would be an experience, a hidden gem, a hole in the wall, or local favorite—I wasn’t just looking for, say, the nearest Mexican or Japanese place.
I ended up eating at Deschutes Brewery just because I was passing it on the street and thought it looked good (it was decent).
The next day, in the same situation, I happened to idly look at the nearby public activity on Brightkite, and noticed that someone who looked cool had posted a photo of their lunch (a grilled cheese sandwich and tomato soup) at a nearby place called Blue Plate Lunch Counter. I decided, based on the photo and my perception of the person’s probable good taste, that Blue Plate looked like it was worth a visit. And it turned out to be exactly what I was looking for: excellent food, very friendly, mom-and-pop operation, very Portland.
I thought of this story again a few days later while I was eating lunch at Frankie’s Spuntino, a cool little Italian restaurant in my neighborhood in New York. This time it wasn’t the restaurant I was discovering, but rather the music they were playing. It was a grey, moody day, and I found myself really appreciating the dour folk-country songs I was hearing there. So I opened Shazam to find out what one of them was called and discovered it was “Silver Raven” by Gene Clark (one of the founders of the Byrds). Thanks to the hipster staff at Frankie’s I had discovered a song I like, and thanks to Shazam I was able to identify and save a record of it.
I can think of a lot of other examples of how this sort of casual, ambient recommendation and discovery has worked far better for me than more formal systems designed for those purposes. Music recommendation services have always left me a bit cold, but I love the simple feature of Last.fm that allows me to see what my friends are playing. I generally find travel guides fairly useless for anything beyond the most obvious tourist attractions, but I’ve managed to file away tons of interesting, off-the-beaten path places to visit gleaned from my friends Delicious bookmarks and Flickr photos (one example: a cool little shop that I visited in Barcelona because I saw a photo of it on Flickr).
I think the reasons these more casual recommendation and discovery methods work better for me are three-fold:
- They allow me to employ my fuzzy, intuitive perception of peoples’ broader personality and taste to determine how likely I am to like the things they like (I thought the person on Brightkite looked cool, so I trusted her taste; I think my Last.fm friends are cool, so I trust that new stuff I see them playing will be interesting to me).
- They aren’t explicitly recommendation systems, but rather allow people to implicitly recommend things just by going about their normal business (someone likes a web page so they post it to Delicious to remember it later, the hipsters at Frankies like Gene Clark so they play his music while they work and I hear it incidentally). I think people are more likely to participate in this kind of system than one where they are expected to formally recommend things.
- They don’t require me to narrow what I’m looking for by overly specific criteria (“I want something that sounds like The Beatles,” for example, or “Give me the nearest low-priced Mexican restaurant”), but rather allow me to say simply “Show me something I might like.”
My theories about this have fascinated me for years, and I’ve often wondered if I’m an outlier in my preference for these sort of “ambient” modes of discovery, or if they might apply to people more broadly. Fortunately, this sort of thinking is the underpinning for one of the three projects I’m working on right now, so I’ll likely be finding out in the next few months!
