Bring in the serendipity dial – for search, music and beyond


At the client workshop I ran earlier in the week, I raised the concept of the “serendipity dial” (something I have written about many years ago in the context of creating enhanced serendipity, and more recently asking Last.FM to introduce a serendipity dial.)

Source: Click on the image for a very large size version including artist names.

The image above shows the similarities between different musicians, as determined by the users of Last.FM. If you like an artist, you are very likely to like other artists positioned close by, and far less likely to like artists positioned on the other side of the chart. This is an example of collaborative filtering, whereby many users behaviors can be used to predict what others with similar musical tastes will like.

Over the last decade the rise of collaborative filtering hasn’t quite met my expectations, but still is used extensively in music, movies, news, and other content, and implicitly in search.

Now that we have some basic success at this level, it is overdue for us to get serendipity dials for us to decide how pinpointed or dispersed we want our results to be.

In the case of music, we may only want music that will definitely like, or we may feel a bit more experimental.

In search results, we again could look for a wider dispersion of results, which could help us find what we were really looking for but didn’t know how to ask for it.

And in news, the serendipity dial would avoid the supposed dangers of a tightly focused “Daily Me” news site that doesn’t go beyond our core interests.