What Systems Do Musiceuticals Require?

I've been wondering about Musiceutical system support. It's clearly all about user experience (what a surprise).

In an obvious first step, devices have to support selection by Musiceutical. Listeners need to be able to say: "At this very moment, I want more energy." "At this very moment, I want inspiration and/or focus." Or perhaps even "deal my current sadness or melancholy." So obviously, providers need access to the device UI. Users need to be able to select Musiceuticals as easily as Playlists. If you control the whole experience, as Apple does, this is easy. If you sell commodity MP3 players differentiated only by physical design, you can alter your software to support Musiceuticals, but where do you get the songlists? Where do you get the data?

So the more interesting question is the data. This leads to two questions, which look similar, but are totally different. Both need research and new ideas:

1: "What affects my MOOD?"

2: "What affects MY mood?"

The difference here is: how to approach Musiceutical effectiveness in general vs. how to tailor them for varying taste, background, age, culture, etc.

This blog will already be long, so for today, I'll just offer initial thoughts about Question 1. As is typical in our Web 2.0 world, there are two approaches: Algorithms and People.

The algorithmic system needs to understand songs. It needs to deconstruct music to find relationships not visible simply by knowing artist and style. It's patently obvious that a single artist can have different songs with very different Musiceutical effects. I have to return to Pandora on this. Today, their focus is similarity for it's own sake, not for impact on emotions. They respond to the request, "Give me more songs like this." But you have to believe that a discussion of algorithmic approach among Pandora architects and engineers will instantly reach a higher level of sophistication than it would among a typical group. It's likely that an algorithmic system will require a "seed--a song that performs as desired leading to new options. I'd also expect that some subset of Pandora's "similarity algorithms" would be more applicable that others (e.g. cadence might be more important than instrument choice).

It also seems obvious that in this case, if a single "seed" is good, many seeds are better. A user who suggests more "effective" songs upfront or rates song effectiveness along the way will surely increase the effectiveness of any particular Musiceutical algorithm.

The other approach is People. With the right perceived value and painless user interface, large numbers of people will rate songs for Musiceutical application and effectiveness (e.g. this song "targets energy" and is "moderately effective at doing so"). This may be the shortest path to success since it requires less new thinking about the problem and can be a smooth expansion on current song ranking tools. For something to rise to the level of a "-ceutical," there's an implication of science or analytics beyond "voting," but this approach may provide better short-term results.

Both of these raise an immediate question: "Am I like the algorithm?" or "Am I like the others providing input?" In other words, "What affects MY mood?" That's for another day, but in the meantime, email ideas about algorithms or voting methods and I'll share them here.