Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Episode 31: Music and the Brain: Guests David Rosen and Scott Miles
17 Plays3 years ago
David Rosen and Scott Miles of Secret Chord Laboratories (secretchordlaboratories.com) talk to Joe and Rolf about musical preference, the role of surprise in these preferences, what's going on in the brain, and how COVID is affecting the way we listen to music.
Discussion paper:
"A Statistical Analysis of the Relationship between Harmonic Surprise and Preference in Popular Music" (2017)
https://www.frontiersin.org/article/10.3389/fnhum.2017.00263
Special Guests: David Rosen and Scott Miles.
Recommended
Transcript
Introduction to Guests and Topic
00:00:08
Speaker
All right, welcome to Cognation. Today we have with us two guests who I'll introduce in a second. But just to make sure that you remember who we are, my name is Rolf Nelson. I'm your co-host and I'm coming to you from Providence, Rhode Island today. And I'm Joe Hardy and I'm coming from El Cerrito, California.
David Rosen's Background in Music Cognition
00:00:30
Speaker
And with us, we have two guests, David Rosen and Scott Miles, who are going to talk to us about the neuroscience of music. So welcome, David and Scott. I'll give you a second to just introduce yourselves. So David, I'll ask you to introduce yourself first.
00:00:46
Speaker
Sure. First, it's great to be here, Ralph and Joe. Thanks for having us. My name is David Rosen, cognitive neuroscientist. My background is in studying music cognition, the neuroscience of specifically jazz improvisation. I've done some work in other verbal domains as well. I'm a musician. I've been a bass player in various bands for the last, let's say, 20 years of my life.
00:01:12
Speaker
And I've been an educator and speaking at different forms, both academic and more along the lines of the music industry, with presenting the research that we're going to talk about today. And then Scott and I, I'll give you a little bit of backgrounds that he can continue. We met in our grad school days. I was studying music improvisation, looking at what happens in the brains of jazz musicians. And he came to my lab at Trexel University in Philadelphia. And he can tell you a funny story about how he found me.
00:01:42
Speaker
Um, but he came looking at music, uh, perception on the other side of things that popular music and we are paths kind of crossed then in 2015. And that's when we started, uh, you know, doing research together. Um, and from that point, uh, you know, we started secret cord laboratories together, um, which is a music tech, uh, startup company. And we began that in April of 2019. So that's kind of my background and how Scott and I know each other and kind of the work that I've been doing over the last decade or so.
Scott Miles' Journey into Music Research
00:02:10
Speaker
Great, okay, and I guess you've already given some information about Scott, but just so we can identify his voice, do you want to say hi, Scott?
00:02:17
Speaker
Uh, this is Scott Miles. Yeah. Um, so, uh, this is what my voice sounds like. Um, I'm also a musician and, uh, and a music cognition, uh, researcher and neuroscientist. Uh, yeah, the story about finding Dave, uh, since he, he, uh, he's that a little bit, I was looking for someone to find a arousal and valence in the emotional content of music that I was looking at for a proposed project that I didn't have funding for yet.
00:02:41
Speaker
And I was trying to find somebody who exactly complemented my skills as a music cognition researcher. And I was looking for someone with a little bit more engineering chops. And so I went to Drexel. And through the serendipitous nature of the internet, I found somebody who perfectly complemented my skills. And then that guy went to work for Pandora in California, and they stuck me with Dave.
00:03:08
Speaker
Well, sorry to hear about that. Yeah. He ended up having exactly the same skill set as I did. Great to have you guys on the show. Thanks for having us. Thanks for having us.
Researching Harmonic Surprise in Music
00:03:20
Speaker
So the paper that we're going to talk about today, or we're going to start our conversation about today, is one that you have written together. The paper is from 2017. It's called a statistical analysis of the relationship between harmonic surprise and preference in popular music. And that's by the two of you. And then the third author, I'm not going to try to pronounce his last name. It's Norberto. Grivitch. Grivitch. OK, thank you. So in this paper, you're looking at
00:03:47
Speaker
I mean, the basic question here is, does surprise in music cause people to prefer it? Or are more surprising things in music, do they cause greater preference for music? And I'm butchering the bottom line of this, but maybe you want to describe how you ran the study and what kinds of findings you came out of it with.
00:04:08
Speaker
Sure. Well, this was actually to be a little bit more statistically rigorous. This was an association and the next study we did was a behavioral study to decide if there was causation. We were looking for that, but this doesn't really show that in our findings. What we did was we looked at, there's a corpus that was transcribed by Ashley Burgoyne at McGill in Montreal.
00:04:32
Speaker
And what he did for his PhD, it's called the McGill Billboard Corpus. It's about 600 songs. And he transcribed all the different sections and all the different, you know, labeled all the sections and named all the chords in the corpus. And so what we did was we said, okay, well, what is different about the top quartile of the Billboard charts in this representative sample of the Billboard charts from 58 to 91, from Johnny B. Goode to Smells Like Teen Spirit.
00:05:01
Speaker
and what is different between a particular statistical measurement of the top and the bottom of the Billboard charts. And what we did is we looked at rarity of chords, which is like a measure of Shannon entropy, really, of information, of surprise, in the zeroth order transcriptions of the chords themselves.
00:05:23
Speaker
And what we found to jump to the findings, we found that the top songs on the Billboard charts were higher in absolute surprise and surprise throughout the songs.
Measuring Surprise and Its Effects
00:05:35
Speaker
And that most of this higher surprise was concentrated in the sections that immediately preceded a chorus, the verses that came before a chorus. So as musicians, we looked at this and said, hmm, this must be some sort of a cognitive basis of
00:05:53
Speaker
of the dopamine release in music has been posited to have something to do with expectation. And so what Dave and I drew as an inference from these findings was that there was this kind of dopamine rush happening when you have this tension and release between the verses and the choruses.
00:06:11
Speaker
All right, so you mentioned that surprise is determined by information theory. So, you know, if you're thinking about the, you know, you say something like, we like music, that's more surprising, but you have to operationalize that some way, you have to be very specific about it.
00:06:27
Speaker
So what exactly is a surprise? So so you're looking at all of the total chords in this corpus of, you know, these popular songs on Billboard. And you take you. So you've got every single chord that's in there and then check the frequency that a single chord is. And so one that would be played less would be more surprising. Correct. Right. And we normalize by key. So normalized by key. So everything is standardized to be a single key. So we work off of chord functions in that way.
00:06:57
Speaker
Okay. I am definitely, so you have to speak to me like a non-musical expert. Sure. Sure. So basically, you know, there's, there's 12 different notes that you could, that you could play and could be the ultimately the root or the chord, the name of that chord. It could be an A major or a C major or an F major. And so what we did was with all of the transcriptions of these songs, which basically were hand annotated chord by chord for every beat in a song, we were able to look at all those 600 songs and normalize them.
00:07:27
Speaker
standardized them to all be one key so that we could make those statistical calculations of rarity across the entire set in a normalized fashion. Makes sense. Yeah, so in that context then you're looking for how the surprise or information in a particular chord is related to the inverse of the probability of that chord
00:07:52
Speaker
in the corpus, basically. Exactly. Exactly. So you can imagine, right. So you can imagine, right. That's in that Shannon entropy model. That's like a zeroth order entropy of just like a pure frequency calculation. You could imagine you could expand beyond that and look at, you know, higher order dependencies and things like chord progressions, like a famous one in rock music, like a one, four, five chords, progressions, like the blues that we see.
00:08:18
Speaker
certain series of chords you can move and think about how that happens at various levels of time within a song and across songs. Right, exactly. And that makes a lot of sense and that was
00:08:32
Speaker
I imagine that would be where you would be going with it. But in the paper, you saw that even with just the zeroth order, just the absolute surprise or information of the chords that were played, and especially the verses before the chorus, that that was predictive or associated with, I guess is probably the best way to say it, Scott, to your point. A higher amount of surprise there would be more likely in a,
00:08:59
Speaker
first quartile popularity song, one that's closer to the top of the charts versus the fourth quartile songs, which are still popular songs, but closer to the bottom of the charts. Yep. That's correct.
00:09:15
Speaker
So from, from that perspective, then, you know, uh, there was the idea that information that, that basically surprise then is positively associated with, um, with preference. And so you guys sort of mentioned that, you know, this might be related to, um, to the release of dopamine. Ralph and I were talking about this last night. We were, we were thinking about the fact that like, you know, surprise can be.
00:09:43
Speaker
in this sense of surprise, where you have some reward expectation, which is really what dopamine is encoding. It's not necessarily just a positive reward, but some difference between the expectation for a reward and the actual reward that you receive. So that deviation from the reward expectation, that could be positive or negative. Right.
00:10:07
Speaker
In this sense, how do you see that playing out in music? When is this surprise positive and when is it not so? Sure. I like to tell a little anecdote about our two memory systems to really talk about. I think there's a great analogy to music here. I think it's around the idea that dopamine, so we always talk about dopamine, is the learning molecule.
00:10:35
Speaker
So in that gap in terms of what we're coding neurally and those predictions, there's something rewarding about learning about something that we care very much about in our lives like music that has such an impact. So when we talk about
00:10:52
Speaker
There's these two memory systems. We
Familiarity vs. Surprise in Musical Experience
00:10:55
Speaker
have, outlined by Kahneman and Sversky, we have system one and system two, where system one is more unconscious and implicit. And what we see is responding to this surprise that occurs in music. We're not really aware or conscious of it necessarily happening.
00:11:13
Speaker
However, we are constantly tracking in our environments this surprise that's happening around us to reduce error. And that's an adaptive property of humans. So that's one system. And the second system is more of that explicit memory system of declarative memory, right? When we sing back the lyrics to our favorite song that we've heard a million times.
00:11:35
Speaker
And so it's like that surprise on the first system and now this explicit familiarity in the second system. So I like to tell a story about walking down steps. I think it really, a weird set of stairs, because it really exemplifies this phenomenon. So if I take either of you, Joe or Rolf, and I put you at the top of a staircase, and there's a hundred steps, and you start to walk down those steps like you would in any building.
00:12:00
Speaker
And after a while, you might say, OK, these are steps and I'm going to look at my phone and I'm going to text or do whatever because I understand how steps work based on what I've experienced before in the world. I went down a ton of different stairwells. And what I know is that usually after a few steps, I get the hang of how far the steps are apart from one another. And I can now automate that behavior and just continue to walk down the steps.
00:12:23
Speaker
Now, if I say you've gone through 97 of these steps and you're smooth sailing and you're thinking you're great at going down these steps and you're gonna make it all the way, but on that 98 step, that 98 step is one foot larger of a gap than every other step before. And so the odds are is that when that step comes up, whoever's going down that stairwell for the first time is likely gonna have a nasty spill, especially if they're not paying close attention to what's happening. And so when you have that nasty spill,
00:12:53
Speaker
That's a red flag, right? That sets off your explicit system two, right? Because you were in this automatic unconscious mode of system one, just kind of using what you've learned before, these very broad rules to the world. And now in this specific situation, because something drastic is happening, caught your attention, you're going to encode that, right? And you're going to encode that into your now knowledge of this specific set of stairs. And there's learning happening there.
00:13:22
Speaker
And then if I take you
AI and Musical Expectations
00:13:23
Speaker
for this next time and put Joe or Ralph back at the top of the steps and you go down these steps for the second time, when you approach this step, this 98th step, there's going to be something that kicks off in you where you feel good.
00:13:37
Speaker
Not only you feel good cuz you're not gonna take a spill but you feel good because you learned this very surprising example of steps and now you know it and you get through those steps and you skip over that ninety eight and take that extra one foot distance down you get to the bottom successfully.
00:13:55
Speaker
I think that's really interesting parallel to when we hear things unconsciously and make these learning associations and become familiar with the things that are statistically irregular in our environment and that's kind of how it plays out in music as well. Add to more directly relate this to the music and how it can be bad surprising.
00:14:16
Speaker
Um, I liked the story of, uh, back to the future, you know, when, uh, when Marty McFly plays the Van Halen, the Eddie Van Halen solo at the end of Johnny be good. And it goes crazy. He says, you guys aren't ready for this yet, but your kids will love it. You know? And so the people who aren't ready for it yet, when they go down those stairs, maybe they're like two feet tall and that one foot difference.
00:14:41
Speaker
Is going to kill them, you know, and so you have to be, that's why we, we like the interplay between familiarity and surprise. It's like, you have to be kind of. Led towards the, you know, that, that staircase can't be too jarring for you or else. If you break your neck, you're never going to go down the staircase again.
00:15:00
Speaker
Right. Because these are also, these are also songs that people are listening to over and over again, right? If it's a, you know, if it's a top billboard hat and it's on the charts for 10 weeks, then this isn't, it's not necessarily shocking each time they hear that different chord progression. It's like, uh, what you're saying is it's that feeling of familiarity together with, uh, sort of. Right. And if, and if, and if the, yeah, if the Beatles never did the one six four or five, you know, she loves you and want to hold your hand.
00:15:29
Speaker
nobody would ever give them a chance when they started playing the sitar and doing Tomorrow Never Knows and all the crazy stuff on Rubber Soul and Revolver. So it's kind of like boiling a frog. Yeah, so I think that was what was going through my mind as I was reading the paper, actually, is exactly this. This seems to be like the meat of the problem, right? Like, how do you right size the amount of surprise and innovation in a piece of music?
00:15:57
Speaker
that makes you want to listen, that makes more people want to listen to it. It's fascinating because it relates to culture and musical history and people's patterns of behavior, what they're listening to. How do you guys think about that?
00:16:12
Speaker
Well, culture is, culture is expectation, right? So we're not cockroaches. You know, I like to tell the big, make the analogy of a cockroach that has that's DNA, it's all set and it's wherever it goes around the world, it's going to be the same, right? We're very sophisticated animals. We adapt to our circumstances. We've been nomadic and kind of spread around the world. So when we're born, you know, we have this like 10, 20 years where we're kind of adapting to where we are. You know, it's kind of like,
00:16:40
Speaker
You know, firmware and software on a computer and it's not all hard coded. And so because we learned this stuff, we learned it for through exposure and you know, and that's culture. And that's, that's what bands us together to the place where you are, you know, like.
00:16:54
Speaker
You know, Dave, for some reason, he's in Philadelphia and he likes the Eagles, you know? And so, you know, um, we forgive him because he lives in Philadelphia, you know? And so, um, and we have our, we have our Eagles chant that we do, like when they score a touchdown, right? And that's, that's all of our culture. And so yeah, it's all part of this.
00:17:11
Speaker
So because expectation is culture, and I don't wanna get too scary about the artificial, the rise of the machines thing about it is, but we try to do this in a good thing, is because expectation is culture, and because it's based on exposure, if you have a sophisticated enough artificial intelligence algorithm, and you understand the properties that you're looking for, which is not just raw information and audio signal, but actual expectation information about what you're listening to,
00:17:40
Speaker
then you're able to in some ways to some extent model that expectation and find out where that sweet spot is. And that's kind of at the core of what we do in our software company.
00:17:53
Speaker
OK, let me follow up on that and think about next kinds of steps. And we like science fiction stuff, too. So we know that expectations can be modeled and measured fairly well with EEG, right? So we know that there are some waveforms that can indicate a surprise or sort of a shift of mismatch between an expectation and a stimulus, right?
00:18:21
Speaker
I wonder if what you're thinking about potential for getting individual differences in surprise from different populations, say if you wanted to, just for instance, you wanted to figure out what the teenagers are listening to these days. So you bring a whole bunch of teenagers who are kind of in tune with popular culture in and measure what sorts of
00:18:48
Speaker
What sorts of individual differences they're showing between groups for surprise? And is this something that you can use to then predict back to a demographic? And so instead of using information that's common to the culture, taking individual differences and looking at how people are surprised in very different ways.
Cultural and Generational Music Preferences
00:19:10
Speaker
Yeah.
00:19:10
Speaker
Yes, exactly. Exactly. You just described an SBIR application that we're actually working on right now. Small Business Innovation Research Grant.
00:19:21
Speaker
Thank you, thank you, Scott. Yeah, to look at, right, because you pointed out, you know, one of the biggest, one of the weaknesses of that paper, the statistical analysis of of harmonic surprise there is that one, it's calculated purely statistically, and there's absolutely a difference between the way that our brains register, you know, a surprising event and how we can track that among a group of, you know, a population of pop listeners who are
00:19:49
Speaker
you know savvy and up to date and have been exposed to the recent pop music to make certain predictions and so that's we want to then track and see like what is the difference you know where are we missing the mark among different of different individual differences you know within pop listeners between the actual surprise that is being signaled in the brain and then this is what is the difference between that and our statistical calculation of surprise and really linking it all back to
00:20:17
Speaker
of preference and enjoyment to further understand, um, the relationship there. So I'm wondering, do you have any intuitions about, about findings you might get out of that? Well, well, one of the things that, uh, so, so we're looking at individual differences. One of the things that's great about looking at the brain is that we're going to have so much data, um, hopefully that we're going to be able to do a lot of different post-hoc analysis. Um, when we, when we look at, so, so one of the things that you touched on originally,
00:20:47
Speaker
uh, the Dave's already responded to us, which is the individual differences and the actual surprise reaction in the brains of the listeners. Another thing is just the basic thing of like, do we get it right? You know, are we, are we, when we say this is surprising, you know, is, is there actually a mismatch happening in the brains and the brain response and the lecture, the electrical potential response in the brain?
00:21:10
Speaker
You know, that's, that's in this, you know, early right anterior negativity that has to do with this frontal like negativity that, that, that, that, uh, goes off when you miss, have a mismatch of cords and expectation. And so that's just another thing is, okay. So if you have that mismatch and it happens in the first 20 seconds of a pop song, what happens when it happens again? Because, you know, Wittgenstein, uh, the philosopher, uh,
00:21:33
Speaker
You know, uh, one time said that, you know, there's no such thing as repetition and music because when you have us, the second note is not the note, again, it's the note that was played before and it has, you know, has a, has a characteristic of that.
00:21:45
Speaker
And so, um, so we, we, we're just going along with our, with our algorithms and we don't know what the difference is, you know, when something is played again, what the, um, you know, there's no differences, just statistics. It's negligible, but the difference in the brain of the listener might be huge. They'd be like, Oh yeah, that's playing again. Oh, when, you know, and then the third chorus is different and it's a different key. And you have a bridge, you have those things for a reason. And that's because, you know, we have this thing called accommodation, uh, and sensory perception. I know.
00:22:13
Speaker
You know, if you're familiar with vision, you know, you know that we have accommodation. If you look at something long enough, you know, the less salient parts of the, of the visual stimulus go away. Well, the same thing happens, um, and, and, and auditory stimulation. And so, um, another thing that we're looking at is, is the system one system to this explicit, um, familiarity. So half of the songs, just to give you a little bit of an idea of the design, half of the songs we're going to give to these participants, they're going to hear on an app.
00:22:44
Speaker
in the proposal where they're going to hear on an app for, for two weeks before they come in. And so you have different groups and different people hear different songs first. And so the idea is how are those expectations and the dopamine release, you know, and the, the sort of, um,
00:22:59
Speaker
preference mechanism, the enjoyment mechanism that comes from those expectations being violated, how is that different if you know them explicitly? So you're actually controlling the stimulus environment of the participants ahead of time, setting their expectations a bit. Right. And that can help us because if you know that a certain kind of surprise is terrible the first time you hear it, but then after 20 listens, it's great. That's really valuable information to give to someone in a record company.
00:23:30
Speaker
Right. Absolutely. I mean, so it sounds like that, you know, there's definitely some interesting things to discuss about how to predict what would be pleasurable. So surprise is definitely there. Um, there's other factors and we could get into those a bit as well, but I'm interested as, as you're talking about this topic strikes me that, you know, this kind of thinking is potentially quite valuable for generative music. So, you know,
00:23:59
Speaker
artificial intelligences that could actually create music. Have you guys thought about that at all? Yeah, that's, that's come up a number of times in conversations with, with music industry partners, also investors and scientists as well, like all around, because that there's a big wave of generative music, right? And where I've, you know, where I see Seeker Chord, I think stepping in and our software, Doper is and, you know,
00:24:24
Speaker
just because what we found in our work when we were making the second portion, I think this is relevant, from the statistical analysis of harmonic surprise and pop music, the next piece was a causal study where we basically modeled
00:24:38
Speaker
the surprise of the top and bottom quartile songs from the Billboard charts. And we generated, like you're saying, with generative music, we made kind of 45-second verse and chorus pairs that actually modeled the surprise from the top and the bottom of the Billboard Corpus.
00:24:54
Speaker
And there was this step in between the modeling and the delivery of these songs to actual participants where we needed human ears on these generative, generatively produced songs and harmonies because, you know, not all of them were very musical or very good.
00:25:13
Speaker
And I think that's where the preference piece and kind of where we can leverage our knowledge is to understand, okay, here's like a number of generative pieces. Given these restraints or this style of music, can you help me filter through these different examples and tell me which ones are the most promising and musical?
00:25:34
Speaker
You can think of it like natural selection, right? Cause if you have like mutations that happen in the genome and then they go out in the real world, it's a brutal process, right? But it's still a process that, that, you know, has arrived at the complexity that we have of, of, of the biosphere now. And so what, what you could see is like that, that Doper or something, you know, product of secret cord could go ahead and do is not necessarily be instrumental in the, in the actual generation of the music, which is possible. But what Dave is talking about is that.
00:26:03
Speaker
it could be involved in the curation or the sort of natural selection, survival of the fittest, you know, you, you make a, you know, a thousand different tunes and then, you know, the secret chord would be able to through its algorithms, say which one is the most musical or the most preferable.
00:26:19
Speaker
I have to jump in because it makes me remind and you have to be sort of, you probably are aware of sort of jokes about making the world's best song or the world's worst song. I remember from a few years ago, there was the world's most unwanted song, which was a song created by committee. I think it's features where it was about 25 minutes long.
00:26:43
Speaker
It was about cowboys, Walmart, Labor Day. Oh, central policy issues. Some of my favorite things. Accordions feature a lot of accordion music, opera. So it was just like a collection of things that people didn't like, put them all together. And it's funny to listen to, although it's not 23 minutes of funny, but...
00:27:05
Speaker
And then the world's best loved song I think is about, it features, it's sort of soft jazz and it's about love and it's about two minutes long. So I mean, if you think of that as like a starting point or what can go wrong with things, I mean, you're describing a much more subtle method and I think you're also talking about how to weed out things that would be kind of on the ridiculous end.
00:27:31
Speaker
Yeah, so to touch on that last point you made about the two different kinds of songs, I would say there are certain characteristics specifically in pop music. There are structural commonalities and kinds of standards and constraints.
00:27:46
Speaker
They're baked into an algorithm because they are they exist out in the world in terms of like behavior and what is preferred and decisions that are being made. So that's a piece of this. And in terms of the other part of what you were kind of just hinting at is that with best and worst songs, I guess the way that I think about that and that our conversations go is that
00:28:09
Speaker
There's no reason that when you release something to say this is the best song or the worst song, our vision is of a world where music is going to be and media more generally is going to be highly personalized.
00:28:25
Speaker
And there is no one best song, even for the fans of a single artist. I know we've done some pilot tests on a Billboard charting R&B artist who has an international following of people from the ages of 10 years old into their 60s and 70s. And if you think about these models that are based on previous exposure and culture and individual differences, then a real strength in this approach is to say,
00:28:54
Speaker
If I have an audience of people in San Francisco who are females between the ages of 25 and 35 for this artist, and I have people in Japan,
00:29:02
Speaker
in Tokyo who are between the ages of 50 and 60 and male, there shouldn't just be one song that is the hit song, that is the best song for that audience. But by having this more kind of segmented, targeted approach with these individualized expectations, I think we can deliver more optimal content out there to people of what they really want to hear based on those expectations that have been set for them.
00:29:29
Speaker
Yeah, and I guess an interesting corollary of that would be that if you're thinking about it that way, then a song can change over time too, right? The song could be right. I mean, maybe this is, I'm coming up with these obvious revelations here, but a song, I mean, that would imply that a song sort of has the right qualities for a particular time when the stage is set.
00:29:53
Speaker
Yeah, look at Christmas carols, right? I mean, you have every Christmas carol in the book recorded in every generation, right? Or the song Hallelujah, which is where we get the name of our company from, Secret Court. There's songs like that that are covered in almost every generation and everybody kind of gives it their twist to it.
00:30:12
Speaker
Yeah, I know. So I mean, one of the things that came to my mind as you guys were talking about that was, you know, even for a given person over time, that your favorite song might change even within a, you know, a small, like, for example, like I'm a Grateful Dead fan. And one of the things that will get you
00:30:32
Speaker
Uh, you get your head ripped off and and one of these in any of these online forums, uh is if you ask people what their favorite Grateful Dead song is You know a dead head is always Immediately offended by the just the fact they're asking that question because it's like you should be asking your what's your what's your favorite version of dark star ever? Yeah, exactly right exactly and so, you know, there there is like yeah, I think that that point is is important that like
00:30:58
Speaker
there's not going to be like a best song. You know, it just doesn't really even make sense. There'll be- Yeah, yeah. And one of the- This is probably obvious to David and Scott. I think you're probably like, okay, you guys came to this conclusion after thinking about it, but it is really an interesting thought. All these different reasons why there can't be just, you can't just sort of have an absolute best song. It's all about matching it to the person and the time and everything. Well, let's use this moment to take a little bit of a break
00:31:26
Speaker
And we'll come back and go over some of these other interesting findings.
00:31:48
Speaker
All right, and we are back talking to David Rosen and Scott Miles about the influence of Surprise in how we prefer popular music. So extending this idea a little bit further, so what other kinds of research exists out there that can tell us about how Surprise plays into our preference?
00:32:12
Speaker
Right. So, so we went into this, um, into our analysis and it, and it makes it sound almost like it's the static thing. Like maybe when, you know, Dave talked a little bit about individuals, uh, being different because of their different exposure. But we're really, what you can see is that this kind of effect, um, happens over time on, on, on a broad scale and an effect is kind of feeds into itself, you know, and then sort of this meta kind of way. And because, so, um, I think we have a, um,
00:32:41
Speaker
a line in our upcoming paper which is about surprise increasing over time.
00:32:47
Speaker
in popular music, in preferred music, in the top quartile, sort of, of the Billboard music and harmonic surprise. And I think it's this, it says, the expectations of yesterday, of, you know, what is it? The expectations of today once thwarted become the yesterday's, you know, expectation violations of tomorrow or something like that. Some, you know, sort of Lewis Carroll,
00:33:13
Speaker
Alice in Wonderland statement in our academic paper. And basically what it means is that you keep needing more and more. It's kind of like tolerance to a drug or to any sort of stimulus. And it's not necessarily the people who need more, but it's the population over time.
00:33:31
Speaker
that kind of ramps up and adjusts to this, because like something that's violated and when the Beatles do something and it's crazy or Elvis moves his hips and that's something not in the music itself, but culturally it's like, oh my gosh, now somebody has to do something more shocking and more shocking, more shocking to get the same result. So you think, so actually when you said,
00:33:55
Speaker
people don't do this but now that made me wonder is that is that something that uh total music heads might be you know that's something that you know as you as you adjust your preferred music you might be doing it because of tolerance for something like that that you want something more right right well there's there's an interplay here because you have the it's not linear over time right because
00:34:20
Speaker
If you, if I were to ask you, uh, Joe or Rolf, if I were to ask you what your, what is the best music of all time is, I guarantee you, you would talk about something that came out, you know, or leads to these that you were listening to when you're between like 15 and 19 years old.
00:34:36
Speaker
Yeah, that's right. That's such, and that's such an interesting effect. So the rest of the other, the music today is crazy. It's stupid, but, but the, everybody has that preference, you know? And so, and so you have that linear effect having over, over time where it gets more and more surprising, but somehow this is happening between the ages of 15 and 19 years old. This, this effect is being driven by cohorts of 15 to 19 year olds over time.
00:35:02
Speaker
who have all been living in an environment where the children's music is exactly the same throughout the decades pretty much in terms of the complexity of that's like the simple children's music sets us up and sets those kind of the regularities that we talk about what leads to surprise it's like because in those ages of of your adolescence where you're forming your closest social bonds and your identity with your friends and having these like
00:35:23
Speaker
meaningful experiences to yourself as an individual, yet living within a similar common culture, you've had the same expectations set through the music you're exposed to as a kid. Right. Well, if you could listen to some of the Disney music now, it has triplet rhyme, triplet hip-hop kind of aspects to it. The zombies show, the movie just came out, my six-year-old is really into, has some Hamilton-type aspects to it.
00:35:53
Speaker
But it's interesting also that you have this phenomenon called reminiscence bumps. And so what we do is we look at 20 years of music and it kind of ramps up every, anytime you look at it, it's almost like a fractal effect. And this is the paper we have coming up. I'm going to guess what this is already. So I know, and I was going to say, so I know that I, I'm hearing more of my kids' music now cause I have kids that are starting to be right around that age. And I'm, and I'm going to predict, you're going to say that
00:36:21
Speaker
It's a generational thing so that you may be listening to. You listen to the music your parents, that's kind of like the music your parents listen to. So like the eight, the nineties is coming back now. You see kids walking down the street with Nirvana shirts on like, I don't know, I saw Nirvana in New York City in 1993, but these kids don't know who Nirvana is, you know? It's predictive and it's predictably cyclical like that. Exactly. Exactly. Exactly.
00:36:47
Speaker
Yeah, I heard these kids on the beach. This summer, I heard some kids on the beach. I texted Scott as soon as I was listening. I was on the beach. Metallica. 18-year-olds had a Metallica there. No one knows more about Metallica. No one knows more about them or loves them more than me in the world. The kids are 18. There's some people who've been seeing Metallica for 30 years, and that's just amazing. Then us adults don't get any credit for being cool, but it doesn't transfer to us at all. No, no, no, no.
Live vs. Digital Music Experiences Post-COVID
00:37:14
Speaker
Parents never get any credit.
00:37:17
Speaker
Yeah, but I remember, you know, when I was a kid, you know, listening to, you know, like Frankie Valley and, you know, and like, and, uh, you know, all those oldies songs and my mom's record player when I was a kid, you know, and I wonder how that, you know, influenced, you know, my, my preference for music. But so you have these kind of explicit, uh, expectations that are developed that like happened every 20, 30 years. And that shows up in this, in the cyclical nature of time. And.
00:37:43
Speaker
and then of, you know, of time in popular music. And then you have, you know, you wonder how much it's kind of like then reinforced and kind of baked into the system where they kind of expect it and make it happen. Yeah. I mean, I, I definitely, you know, so I, I, I am a child of the seventies. And so, I mean, all my friends and I all listened to Led Zeppelin.
00:38:05
Speaker
and Pink Floyd and all that kind of stuff when we were kids. As teenagers, we listened to that stuff that we heard when we were kids first. So it wasn't even timely when we were listening to it. It was like when we were born. The other thing is preferences are formed in that age.
00:38:26
Speaker
Sometimes even earlier than 15, but like, for example, like for sports teams, you know, like the, you're talking, we're talking about the Eagles, the Eagles before, but like, you know, like I'm a Red Sox fan and the New England Patriots fan. And if you think about it, I haven't lived there in 20 years. Yeah.
00:38:44
Speaker
Well, that's because that's when you're getting ready to fight for your land, right? I mean, if you look evolutionarily rise, right? When you get to be like 16, 17, 18 years old, you're going to have to die for your land. And so you have to develop an attachment to what's going on around you culturally. Yeah, it's fascinating. It's fascinating. So here's a question for you.
00:39:10
Speaker
You and you would know more about what a good theory of what music is for, maybe an evolutionary explanation of music. But I know that one proposed explanation is that music can be for communal synchrony, right? So drums around the fire, right?
00:39:29
Speaker
And I wonder, this is totally speculative, but I wonder if you have thought about how musical listening is changing in the age of COVID when we don't have huge groups gathering together to experience the rhythm at the same time. You can't go to concerts. So what is music listening like now? What's your experience of music listening now?
00:39:56
Speaker
I got to say that not going to live concerts is one of the things I miss the most. I think we're talking about neural synchrony is one correlate of that experience of just like one piece of such a great phenomenon when you know what's led us all. Me and Scott here is standing in a building with thousands of people where a few other humans are on a stage with air vibrating at certain
00:40:22
Speaker
know it's certain paces and frequencies being created and hitting our ear drums and yet we like we're all melting during a common moment and experience all together so that's not happening and I know there's you know there's no webcasting and lots of shows like that but you know when I experienced those
00:40:39
Speaker
I've done the webcasting before and I've been to live shows. I'm also a fish fan and they do when I can't make it out to a show. I'm webcasting at home with my wife and friends and that's a fundamentally different experience. But I do think that to think about
00:40:54
Speaker
Those are different experiences that happened before COVID and now it's just more so the case not having a live experience with COVID. We talk all the time about one dichotomy in listening is passive versus active on the edge of your seat listening. When I close my eyes and I just want to chill out for 30 minutes and put on a cool new EP by a band that I like and really listen closely,
00:41:19
Speaker
That is like a quintessentially different type of listening experience than some ambient electronic of music that I have on the background with no vocals to help me work during the day and knock out some like emails and and so those are you know
00:41:34
Speaker
Those are really interesting topics for further exploration that I haven't seen much
Feedback in Music Creation Challenges
00:41:40
Speaker
work on. At this point, the studies that I know we designed and others have, have required this kind of active, really close listening. And then the next question I think is, well, in terms of people, tracks on Spotify and streaming on YouTube and all those things, is there a way to classify those different types of listening and have any insights into what's driving what types of music in those different contexts?
00:42:03
Speaker
Yeah, and it's interesting, the feedback that the artist is getting is different too. I mean, talk about bands like Fish, and we were talking about The Grateful Dead a little bit. You know, those bands work out their material on the stage, so they'll have a song going in that they wrote, but it isn't done, right? It's continuing to get worked out on the stage, and the feedback of what people are responding to is direct in that way.
00:42:29
Speaker
And you just don't have that without live music. And there are, I mean, versus other types of music, which are all purely pre-produced, you're getting feedback from maybe the immediate circle of people around you who are saying, Oh, that's cool. You know, add this or take that out.
00:42:44
Speaker
But, you know, you just don't have that, that feedback loop of, of that live performance right now. Right. And that's not, it's not just with COVID. I mean, you have a lot of, you know, you had like a little Nas X with, you know, the hit the, with the old town road kind of, we have a lot of our like Finneas and, uh, and Billie Eilish who are, uh, who are making music, you know, on their laptop in their bedroom and they don't have the, uh, and that's just kind of a fact of what's, what's happening because of what technology is allowing even.
00:43:12
Speaker
you know, even barring any pandemic. And that's kind of what, you know, 10 years ago, you know, this coming summer, 10 years ago, I was making an album. I was up in Cambridge and I was working for a neurologist up there in Massachusetts. And, you know, that's, that's kind of how the idea came up for us to start this research was that I didn't have any, you know, I kept sending my friends back at home, like MP3s of the different bridges that I've written for a song. And they're like,
00:43:40
Speaker
You know, I'm not playing gigs like I'm doing research. You know, I'm like, I'm looking at, you know, brain images all day long. I don't have time to go out and gig and I don't have the money and the resources to bounce this off. And so one of the impedances of secret cord and the research that we've been doing is to give someone sign up kind of that sounding board.
00:43:59
Speaker
that, you know, nothing will ever replace, of course, you know, going out and, you know, and workshopping material on a stage.
Role of AI in Music Creation
00:44:06
Speaker
But this would maybe somehow give someone a little bit of an idea as to, you know, what their tweaks and changes are doing. That's fascinating. I mean, I think that that that probably leads us directly into the into the Robopocalypse question. But you've been forewarned about so. Yeah, we've got it before.
00:44:27
Speaker
Yeah, sure. I mean, we're, we're, we're musicians, you know, we're worried about that too, you know? Yeah. Yeah. Yeah. So in, in the, in the worst case scenario, how could the, like, not necessarily like your thing that you're working on right now, but like the extension, like the logical extension of, of this line of reasoning, how could that lead to robots taking the world over or other.
00:44:49
Speaker
sort of apocalyptic outcome. Well, if you look at- I'll add one more quick one too, which is what is the, what's the theme song to the robo-apocalypse? Oh, that's a good question. Yeah, you could be mine, I guess. It depends who you are and where you live.
00:45:05
Speaker
Exactly expectation based so so we look at what what's happened in the past right and you have to the scary thing is right. In the past you've had with industrial revolution and then the information revolution and now the artificial intelligence revolution you've had.
00:45:22
Speaker
What happens is you have people's jobs up, you know, upended and it's, you know, this manual labor or, or, or repetitive tasks or anything involving, you know, what's replaced. And then those people, the idea is, okay, well, now you go get a gig and you do Uber or you, you become a graphic designer for a website and all these, all these jobs that are more creativity and more service oriented, come and replace them, replace putatively are going to help, you know, in the economy.
00:45:52
Speaker
and kind of stand in the way and replace the jobs that are displaced. And that's always sustained itself to a certain extent overall that we've been able to create more jobs than are taken away. And if there is the training for that, that's the big thing.
00:46:11
Speaker
So so there's creative and their service industries that are still kind of, you know, kind of rising up and taking the place of all these, these jobs that are replaced. But the question is, what happens if when computers can serve and computers can can be creative? And so that's the dangerous thing. That's what I think.
00:46:31
Speaker
you know, it kind of, the corner that this question gets kind of pushed into is like, when we have creative computers, or if we have, you know, computers that can be of service to, you know, elderly people or, you know, and food service and that sort of thing, that's, to me, is the question of where it gets scary.
00:46:51
Speaker
Yeah. And so, I mean, I guess my idea of the sort of utopian future is, you know, people sitting around flourishing, doing the things that they want to do, which would probably be, you know, listening to music, creating music, that sort of thing. You know, what are we going to do if we can't even make good music? I think I think I guess I think about I hear that. I mean, I spent a lot of time making music and
00:47:15
Speaker
A lot of the time that I spend making music is intrinsic. It's an intrinsic motivation. I do quite a bit of research on creativity and flow states, achieving this optimal experience where you're fully engaged and it's intrinsically rewarding because it's a sense of something you really care about and you develop a discipline and a skill around. And giving those tasks and having AI and robots have them is scary. But I think in reality,
00:47:44
Speaker
humans are going to continue whether there's AI making music or not humanity is going to continue to immerse themselves in these types of activities because what they feel and what they get out of them at its core give us meaning and give us purpose in our lives and it's not every time someone sits down at a piano or picks up a guitar it's not to get the most streams or to be predicted to make a certain you know list so that's like kind of one piece and the other I guess I'd say is
00:48:14
Speaker
When we think of algorithmically predicting how something will be enjoyed or preferred by an audience, we always talk about that as one of many data points that people are going to use.
00:48:31
Speaker
to really as a point to check their gut when there's many, right now there's many subjective opinions around the table when there's decisions to be made about what artists are to be signed or what song is going to be the single. And I think to have that data point on that side, on the business end, but then also on the production side for the artists to say, hey, I'm in a rut.
00:48:53
Speaker
And I want some feedback about how my song compares. Maybe this could help me tweak my writing style. And so the other part of this, the vision of, you know, science and technology and this singularity is that there's like this symbiotic relationship where I can actually progress my own style and understanding of myself as a composer and what things I do repeatedly and what habits I fall into and maybe use that as a way to, you know, push creativity in actuality.
00:49:22
Speaker
Well, that seems like a great place to stop it on a high note, a positive note. I would like to thank you both, David and Scott for, for being on the show. Very interesting conversation. Thanks. Thanks a lot. Thanks for having us. It was fun.