Recommendations Based on Your Mood, Not Just Your Taste

[NOTE: this post originally appeared on Datachondria, a blog dedicated to technology, data, and modern life.]

I've written elsewhere of my love for the Nike+ and its exemplary use of technology as a bridge between a solitary leisure pursuit and the support offered by a network. But I find -- as I increase my workout times -- that there are some significant feature gaps. While I like the 'powersong' feature to deliver a boost when I find myself flagging, it doesn't quite take it to the next step. For example, if I want to change music to something more appropriate to a warm-up or a warm-down, I have to pause the workout and change the playlist.

But these things have a fairly predictable place in my routine. It should be possible for the technology to anticipate them. Just like bookclubs (real or virtual) will set landmarks for progress, there's no reason why you shouldn't be able to set playlist markers. Get me to x bpm after y minutes, keep me there for z minutes, then give me a warmdown. These landmarks could be used to manage a random selection of music within nonetheless predetermined parameters.

This kind of landmarking could be helpful in all kinds of situations -- workouts, parties, coitus.

But that's really only the first step. I'd ideally want the ability to tailor my listening dynamically to my pace, location, mood, heartrate, or any number of other parameters which would increase its utility. Music at parties could change depending on the volume of the conversation, the number of participants, and the amount of alcohol thus far consumed. Think of the public safety implications.

The same is true of reading -- I'm more inclined to read business books and articles on the way to work, but a novel or entertainment news on the way home (particularly on Fridays). An electronic reading device should be able to push content to me on the basis of my habits, mood, place, location, activities. In this context, having to remember to pick up a different magazine or different section of a newspaper seems just ridiculously inefficient.

In short, we're talking about having our devices present content to us on the basis of mood, not just an aggregated taste history, or the ratings we have assigned to things.

Suppose that our interfaces could isolate trends in your reading, listening, or leisure pursuits. Suppose that iTunes Genuis or Amazon's recommendations algorithm could give you some options:

  • It looks like you've been listening to a lot of indie guitar noise merchants recently. Would you like these kinds of recommendations to be weighted higher in your results?
  • You've rather gone off James Patterson of late. Even though you own his entire published output, would you like recommendations based on this to be fact to be downplayed?

Better yet -- imagine that these engines could talk to one another.

* Note: not all of these scenarios are based on personal experience.

These are, of course, exactly the kinds of things that happen in day-to-day conversations with friends. Which is why social networking stands to enrich our exposure to culture and events so significantly. In the meantime, how do shuffle and recommendations engines acquire more precise sensitivities to our desires? Landmarks, trending, and complements would be good first steps.