This morning I was listening to the Madden Brothers' "We Are Done", which came up on The Echo Nest Discovery list this week. I have the sinking/tingling feeling, and we'll see if I'm right because this isn't necessarily one of my actual talents, that this song is going to become ubiquitous enough in my environment that I'll look back wistfully on listening to it entirely voluntarily.  

But listening to it also made me think about Fun's "We Are Young" and Pitbull and JLo's World Cup song "We Are One" and Taylor Swift's "We Are Never Ever Getting Back Together". We are, we are, we are, we are. This kind of pointless internal word-association used to dissipate harmlessly inside my head. But now I have the resources to indulge it at scale.  

So I made a Spotify playlist of 100 songs that follow the title-pattern "We Are [something]". And then I realized there were more of them, so I made a playlist of 1000 of them. And having done that, it was trivial to make similar playlists for "I Am [something]", You Are [something]", "He Is [something]", "She Is [something]" and "They Are [something]", so I did that, too. And then I have a thing that will summarize the contents of a playlist in various ways, so I ran it on these because why not?  

The first thing one finds is that I, You and We songs are way more prevalent than She and They songs. And although there are plenty of He songs, they are disconcertingly overwhelmingly religious, which is kind of different. So I kept the He/She/They playlists for your amusement, but I only analyzed I, You and We.  

Here's what The Echo Nest's listening machines report:  

Metric	I Am...	You Are...	We Are...	Power
Acousticness	0.260	0.383	0.182	0.318
Bounciness	0.398	0.412	0.411	0.041
Danceability	0.498	0.508	0.524	0.077
Energy	0.649	0.541	0.705	0.333
Instrumentalness	0.196	0.171	0.235	0.099
Loudness	-8.47	-9.58	-7.91	0.198
Mechanism	0.494	0.481	0.592	0.238
Organism	0.436	0.486	0.346	0.294
Tempo	124.0	118.1	125.6	0.137
Valence	0.424	0.421	0.421	0.007
Year	2004.4	2003.2	2007.1	0.201

 

The I/You/We scores here are average values across the 1000 songs for each title pattern. Most Echo Nest metrics are normalized to be a unit-less decimal value between 0 and 1. Loudness is customarily measured on a weird negative scale, Tempo is in beats-per-minute, and Year is obviously in years.  

The Power column measures the discriminatory power of each metric. So the two metrics that discriminate best between these three sets of songs are Acousticness and Energy. The metrics with the least power to discriminate between these sets are Valence (emotional mood), Bounciness (atmospheric density vs spiky jumpiness) and Danceability, all of which vary much more widely within each category than between them. Comparing the whole set to my earlier measurements of genre, year, popularity and country shows that the pronoun sets are about as distinct as sets based on country of origin, and more distinct than sets based on popularity, but less distinct on the whole than sets based on year or genre.  

Which is a lot more difference than I expected, actually, particularly between You and We. Individual songs can have any individual character, but taken as an aggregate, "You Are" songs are significantly calmer, more acoustic and more organic in their rhythm. "We Are" songs are more energetic, more electric, notably more mechanically driven, and louder. That is, we sing more tender songs to each other, and more anthems about ourselves together.  

We also seem to be singing more We Are songs lately. Or, to be more precise, because these 1000-song subsets are selected by popularity, more-recent We Are songs are a little more popular now in the aggregate. Attentive observers may recall that my earlier study showed correlations between time and both Acousticness and Energy over the years from 1950 to 2013. But both song-sets here are largely more-recent songs from well after the period of greatest historical change for either metric, and the magnitude of difference is significantly larger than the degree of variation predicted by year alone.  

The I Am songs fall into an interestingly conflicted middle ground. They are more acoustic and less energetic than the rousing We Are anthems, but not as tender and sensitive as the wistful You Are odes. But while the I Am songs are closer to the You Are songs in rhythmic regularity, they're closer to the We Are songs in tempo.  

So while you might reasonably expect We to be a compromise between I and You, this brief study clearly and crushingly demonstrates that the pre-computational centuries of the study of the psychology of self have been a sad speculative waste of time. Music and math prove that our individual selves float suspended between what we project onto others and what we dream that we could achieve together.