Suggestion Machine

Playlists: Discovery, Simplified

As a reader, one of the most difficult things I face is figuring out what to read next. It’s a big problem—the amount of time I spend reading correlates directly with how much I’m enjoying my current book. This puts a tremendous amount of importance on the power of suggestion. If I crack open the wrong book, it could slow down my reading for weeks, maybe even months. Conversely, if I find just the right writer or novel, it spurs on my reading and I find myself sailing the world through the power of fiction.  

When we designed Chronicle 2, we spent a lot of time thinking about how to use the power of suggestion for good. We spent time looking at the current offerings. One thing we discovered is that formulaic platforms fixate on certain types of similarities: authors who are known to be similar, books of similar genres, similar subject matter. They focus on the metadata of the fiction and use large samples to make broad generalizations about what kinds of things might share commonalities.

Computerized Suggestion, Total Confusion

One of the most famous pictures from Minnesota is a snapshot of the Spoonbridge. This is a large sculpture: a gray spoon bent over a pond and, on its end, sits a red cherry. This sculpture is a fixture of Minnesota for a few reasons, but those who have seen it up close know that it’s imposing, that seeing it in the greater context of the city rising in the near-distance, with all the noise and rush of traffic and throngs of peers, is a unique experience.

Let’s say that I wanted to use a suggestion algorithm in order to replicate this aesthetic experience that I get from viewing Spoonbridge. The types of suggestions I’d receive from an automated function (i.e. a website that uses large samples to make recommendations) would fall into two categories:

1.     Sculptures of similar size and content (large sculptures that involve food)

2.     Sculptures that were created by the same or similar artists

Notice that neither of the above does anything to capture the actual aesthetic experience that I have when viewing Spoonbridge: the artificial intelligence has no input for this complex, multi-faceted experience that actually explains why I enjoy this particular sculpture in the first place: the city, the noise, its juxtaposition with the pond. The algorithm totally misses the point.

Experience, Textured

The example of Spoonbridge captures the problem of computerized suggestion in fiction, as well. Computers do not understand the texture of experience nor the minute differences in language that mark the distinctions between individual works. Just because two stories tackle similar subject matter doesn’t mean that you’ll enjoy them both. The texture of language is dense, rich, and replete with nuance that only humans can understand.

Playlists are Chronicle’s concealed weapon of suggestion: curated by humans who love reading, who understand the nuts-and-bolts of language. The system isn’t perfect: recommendation is a bit of a black art. But with so much reading time at stake, we think that moving the needle is important. Spend less time searching; more time sailing the world.

Romance, Redefined

Today we’re launching a new Playlist: Romance. We’re proud of the stories in this playlist, not just for their individual strengths but also how they mix together. As a whole, we believe they present both similarity and difference, and in doing so capture the essence of modern-day love. As a celebration of this work, we’re offering the Romance Playlist free to all Chronicle users through the end of the month. We know stories. Come and see the difference.