Mea culpa: I hear your downvotes loud and clear. Asking for funding to implement these ideas was in bad taste and I apologise. I guess this is my version of AI psychosis, I have removed it from the post. Still, I think the 4 main ideas in my post have merit and I genuinely would like feedback on it and would love to see it implemented for the good of humanity:
using passive haptic feedback learning combined with the recently rediscovered method of teaching the historical style of improvisation to speed up the acquisition of musical ability,
restoring the practice of communal singing as a means of generating shared sacredness experiences for secular people to increase the psychological resilience of their communities in the face of AI erasing stable sources of meaning like identity tied to one's vocation,
that the first idea can enable the second to train the large number of amateur musicians needed to organise, lead and accompany these communal singing sessions
and that a commercial organisation reaping the benefits of the first idea in order to sustainably fund the roll out of the second is the most direct means to achieving the greatest societal impact in terms of improving resilience.
I will happily read and engage with any suggestions or counter arguments to any of the above. I will try and build my own first prototype over the winter holidays and report back on the results.
TLDR:
I'm seeing more and more highly skilled professionals becoming despondent from AI erasing their source of meaning because they feel it is displacing them in their vocation. The newest cohort is mathematicians after visual artists, musicians and programmers have already gone through different waves of this.
I think we will need to reformulate/reintroduce shared sacredness practices into communities to build resilience against the looming psychological fallout.
Based on my experience and research my suggested "social technology" for triggering shared sacredness experiences in non-religious people is communal singing.
In order to enable communal singing in a way that's sustainble long term and reliably triggers shared sacredness experiences I think we need to repurpose the approach of the church which was to accompany singing with improvised music. Professional organists to this day are the only classical musicians who are still expected to be able to improvise in the historical style.
If this approach is to work en masse, 1000s of amateur musicians need to be trained in historical improvisation so that they can organise, lead and accompany communal singing session in their own communities.
I propose historical improvisation (Bach to Mozart) as the starting point for 2 reasons:
the musical style has the necessary cultural gravitas to more reliably trigger shared sacredness experiences (I base this on revelead preferences and existing cultural practices),
recent scholarship has uncovered the fully fledged method called partimento that was used to teach church musicians in the historical style - the method entails learning about 500 - 1000 "stock phrases" of 4-8 notes in length (bass and melody pairs and 4 voice combinations)
In order to speed up acquisition of the phrases and reduce the friction of learning to improvise music on a keyboard (and playing music in general) I propose combining 2 things:
Passive haptic feedback learning research has shown that keyboard fingering order for simple melodies can be learned without paying attention by stimulating the fingers using gloves with actuators in them.
Research, anecdotes from successful jazz musicians and teachers and the behaviour of musical prodigies like Jacob Collier point to the need for audiation (musical visualisation) in order to become "fluent" in music as a language - which (oversimplifying) can be achieved by learning to sing all the parts in a piece of music.
In my opinion the mechanical skills and high number of repetitionts needed to learn to play an instrument presents the biggest barrier to becoming musical as an adult. By directing the learners attention entirely to audiation while absorbing the correct fingerings passively, many more people would be able to become proficient in improvising and playing because the need for boring repetition needed acquiring motor skills can be minimised.
My best guess as to why passive haptic feedback learning (PHFL) tech hasn't been commercialised yet even though it was first presented in 2010 by Thad Starner is because the approaches explored so far ignored the need for audiation and learning music like as language. Too much focus was played on the mechanical aspects learning music which are insufficient by themselves, akin to learning to speak English by focusing on the mechanics of touch typing.
The partimento method uncovered in the past 2 decades of scholarship presents complete manuscripts as to how to learn historical improvisation in a manner analogous to acquiring a language.
Learning this can enable people to become proficient in "speaking" music at the keyboard and give them the necessary foundation for learning and exploring other styles of improvisation and learning classical or modern pieces of piano music because they will have already internalised many of the patterns already underpining the harmonic motions most of music.
The PHFL interface would be comprised of 2 haptic feedback gloves, a music teaching software and the learner would need a laptop with a working microphone and a midi capable keyboard to study along.
Being able to rapidly teach other styles of improvisation (such as romantic era harmony) and popular classical pieces is likely to be commercially lucrative.
If the interface delivers and is commercially successful the profits can be used to train many more amateur musicians who organise, lead and accompany communal singing events.
If you know anyone, like Thad Starner, working on commercialising passive haptic learning interface, I would really appreciate if you could share these ideas with them. I may work up the courage to write to Starner myself once I've received more feedback here.
Epistemic status: Mix of exploratory and my best guess. I do try to substantiate my claims with relevant research where possible.
I’ve structured my idea into the following 3 segments and an appendix:
Shared sacredness as antidote: Why I think we need to revitalise shared sacredness practices like communal singing in order to deal with the coming meaning crisis from AI encroaching on one of the last sources of meaning - identity tied to one’s vocation.
Passive haptic learning and partimento: How training thousands of amateur musicians capable enough to improvise keyboard accompaniment to communal singing can be made feasible through teaching partimento combined with singing and passive haptic feedback learning.
Commercialisation to support the second goal: I outline a few very basic ideas on how a commercial entreprise could be built around the passive haptic feedback learning and partimento idea to support the popularisation of shared sacredness experiences for secular people.
Shared sacredness as antidote
Seeing a Professor of Applied Mathematics cry and quote Ecclesiastes in an NYT interview [1, Interview clip] in response to the newest developments in AI driven maths proofs moved me so much that I couldn’t shake the feeling that something has to be done to prepare for the coming psychological fallout. I had already read many accounts on reddit and elsewhere of people talking about how AI is eroding their sense of purpose and meaning in their job as it takes over their role, and I can sense it too, that eventually my white collar job will in large part be eliminated by AI.
I have felt purposeless for a very long time, even before AI because I never found my vocation in life and stumbled around from hobby to hobby until I was finally diagnosed with ADHD and started medication. That fixed everything except the source of meaning part. I tried all the usual approaches to finding meaning before that: meditation, jhanas, learning to compose music, play the piano, draw, cook, code, judo, reading a lot about psychology, studying thinking models (thank you lesswrong), AI, etc. I’ve run the gamut of available interests and couldn’t find meaning™ in any of them long term. I also couldn’t overcome some of the hurdles in certain activities to reach advanced levels of competence, namely: composing, piano and drawing. What stuck with me is the urge to become a composer (even though I didn’t really act on it), which eventually led me to discover the partimento method. It felt like discovering the holy grail of music education: this is what all the classical greats studied. Eureka! Well, when I first found out about it I hadn’t started ADHD medication yet and doing the mind-numbingly boring amount of repetitions it would take to become proficient in anything musical prevents even most neurotypicals from making progress, let alone someone like me, so I didn’t make any headways with the method and eventually abandoned it.
In the meanwhile my search for meaning had gotten so desperate, I tried to become religious and studied the bible and thought a lot about going to church just for the communal singing. I could never bring myself to do it because every time a religious friend invited me to some special occasion like a christening, I would experience what I now know as shared sacredness while singing, despite not knowing the words or melody of the hymn or believing in anything being said. Once the singing was over that blissful experience would then be tainted by feeling like a phony outsider infiltrating the congregation because real believers do have a tendency to look at you with an earnest hope in their eyes that you too are a believer. Or at least that’s what it felt like. And of course it wasn’t everyone, and I am not trying to imply that I went to a church full of fanatics but it was common enough to be off-putting. Nevertheless, I knew from then on that my life would be better if communal singing was a regular part of it, I just didn’t know how to find the right community to do it with.
Beyond my anecdotal experience there is some research backing up the idea that shared sacredness experience (or collective effervescence if you prefer Durkheim’s original term) is a fairly universal phenomenon and reliably achieved through singing together. A meta-analysis of this phenomenon links such experiences with wellbeing, social integration and collective identity. [2]
Communal singing probably works as a “social technology” for evoking shared sacredness experience even in non-religious people because it combines several potentially relevant features:
A shared focus of attention.
Coordinated timing and breathing.
A common activity to which each person contributes.
Emotional expression that does not require anyone to formulate an argument or disclose their personal views.
An immediately perceptible result that belongs to the group.
Singing groups have shown faster initial social bonding than comparison classes, although those differences diminished over time. [4] Another study found increased social closeness and measured improvements in pain thresholds following group singing. [5] A randomised trial involving 258 older adults found improved mental health-related quality of life following a community singing programme. [6] I think all of this validates my own experience and desire for something like this and makes it worth investigating as means of improving people’s resilience to the meaning crisis that already seems to be on our doorstep.
Let’s say my anecdotes and the research point in the right direction. We still lack multiple things to start implementing this en masse or even on a small scale:
Songs and music that have the necessary gravitas to reliably trigger shared sacredness experiences in people
Trained musicians willing to lead and improvise accompaniment to communal singing experiences
Places to gather and sing together
I don’t have a globally applicable answer to all 3 but the partimento method and professional organist training offer a strong starting point:
I would posit that the top 40 hymns in rotation in various Christian faiths can work for secular people who grew up in cultures with some local variant of the Christian faith as what I would term the “shared sacredness object”. If my proposed intervention is to work on a global scale then it would have to be expanded to other cultures, which can be done if the training methods proposed in the next section prove effective at scale. I am not wedded to this idea, and would ultimately leave it up the musicians organising community singing.
Modern professional musicians who play classical music mostly can’t improvise, with the only notable exceptions being organ players (guess why), and besides are busy practicing and aren't evenly distributed amongst communities so I would propose that each community needs its own amateur musician or musicians who can fulfil the role of leader and accompanist.
Finally, I think that the accompaniment to the communal sining should be improvised in the classical style, on the spot, rather than pre-rehearsed in order to keep the experience from going stale and because the musical style holds a certain cultural gravitas, even outside of Western World (particularly Japan, China, Korea, as far as i am aware). Also this is what worked for the catholic church - Bach would have to write a new cantata every week - so it’s worth taking as a starting point with more limited scope.
I think on some level the novelty of offering historical style improvisations to accompany singing could itself help boost the initial interest in such events. This leads me to the tech solution part of my proposal.
Passive haptic learning and partimento
As I mentioned above only classical organists learn to improvise as part of their education and at least in the German music schools, the rediscovered methods of partimento are enjoying a revival of sorts. In French organ schools they have stuck with their original methods as far as I am aware but in essence the French approach was already close enough to partimento because its origins can also be traced to the methodology originating from Naples.
What is partimento? Partimento is the keyboard pedagogy part of the education that orphaned boys received in Neapolitan conservatories. Originating from somewhere in the 1600s the method sought to teach them the basics of harmony in order to provide them with stable employment as church musicians. Robert Gjerdingen and colleagues have uncovered and translated an entire treasure trove of manuscripts on what they were taught, practically covering the entire beginner to advanced development levels. Many of them are freely available on partimenti.org. The partimento method essentially teaches a set of bass motions and sequences (rule of the octave, circle of fifths progressions, etc), cadences (the punctuation points in this style of music), common paired bass-melody patterns called schemata (that are so ubiquitous in the historical style that scholars have given them names) and methods of introducing and treating dissonances. All of this can be thought of as a set of stock phrases that can be memorised much like the way one would memorise many example sentences when learning a language. [6], [7]
In order to become fluent in this keyboard improvisation style, about 500 - 1000 stock phrases of about 4-8 notes in length need to be learned and internalised before they can be used as interchangeable modules of musical phrases to be used in improvisation and composition. The end result sounds very exciting initially doesn’t it? The hard truth is that the amount of repetition at the keyboard needed to get there is mind-numbingly boring. Not something you would do unless you were an orphan desperate for employment. Not to mention you have to both be able to sing them and play them at the keyboard if you want employment at a fine establishment in Naples (more on the singing part soon).
Passive haptic feedback learning(PHFL) to the rescue!
The video shows one of the CNN anchors learned to play the melody of Ode to Joy on the piano in half an hour while browsing his emails and doing his usual work related tasks. All the glove did was buzz the correct note sequence into his fingers. It's really fun to see how astounded the anchor is that he can play it all. Not very well of course but he does press the buttons in the right order. Which is exactly what PHFL learning can accomplish: teach the button presses in the right order.
In an early Georgia Tech experiment also led by Thad Starner, five musical novices required approximately three keyboard test attempts on average following passive haptic training, compared with seven during active training to be able to play with 90% accuracy. [8]
Source: Huang et al. [8]
More recent work combined passive rehearsal with active piano practice. A two-week study recruited 36 participants; its main analysis used 20 participants with complete, valid data. Passive rehearsal increased the modelled learning rate for note accuracy by 49.7%, but did not significantly improve rhythm learning. [9]
Two-handed learning has also been investigated directly. A 2015 study explored two-handed piano material and reported better passive-learning results when both hands were trained simultaneously rather than sequentially. [10]
So why don’t we have this tech commercially available yet?
My best guess is because, as jazz teacher Hal Galper likes to say “the instrument is an illusion”. PHFL only teaches the correct button presses which for music is only half the battle of learning. The other half is audiation and singing. In other words learning to think (audiate) and speak (sing) in the language of music.
I know this from first hand experience, anything I can sing (even badly) I can play more or less, depending on how much I practiced the motor skills needed. I am not saying it's not possible to overcome the need for audiation, I am guilty of learning patterns I can’t audiate and perform them purely mechanically and I suspect many amateur shred guitarists are guilty of this too (sorry, couldn’t resist).
Experimental work also shows that auditory and motor imagery abilities influence musical learning and recall. [11] Research has also found an association between auditory imagery vividness and singing accuracy. [12] So while I am fairly convinced of this being the missing element based on anecdote and research, it is only fair to acknowledge that this would only be substantiated if the haptic learning interface and the suggested curriculum is tested on multiple cohorts of untrained or amateur musicians.
So, where does the 2x learning speed come from exactly?
Well let’s sum up, what’ve explored so far:
There are approximately 500 - 1000 stock phrases of 4-8 notes in length needed to learn improvising in the classical style to reach the level of competent accompanist (my best guess).
PHFL can teach the correct order of keys on both hands without the need for direct attention (validated by studies for short musical phrases for both hands playing together and separately).
So that frees up the learner to focus on learning to sing the melodies and basslines while the PHFL gloves vibrate the correct fingerings into them (my best guess but there is nothing in the research which suggests that this would cause confusion rather than reinforce the two activities).
Therefore I think that using a structured approach derived from the partimento method: any average person with musical inclinations, in about 6-12 months of 30 minutes of structured daily practice can become a competent improviser and accompanist for the relatively simple songs that people sing during communal singing. I am not at all claiming that anyone can become Mozart or Jacob Collier through this method, just as learning a second language will not guarantee becoming a renowned poet in the language. If you search for partimento realisation on youtube you can find people who were able to internalise the skills needed to improvise competent little pieces in the historical style - which for shared sacredness is about the level to aim for. If you want to see what it looks like at the professional level and the same methods applied to the romantic idiom, check out En Blanc et Noir. He uses the partimento method to improvise waltzes in the style of Chopin for instance. https://www.youtube.com/watch?v=GrlyKMXYLG8
The structured approach and continuous assessment could be provided by a bespoke music learning software developed with AI.
All in all, the user would need a midi capable keyboard, a laptop with a working microphone and the PHFL gloves to learn the method.
What I have outlined so far gives us 3 memory anchors for memorising these musical phrases:
The sung musical line itself (though they may not be the catchiest tunes in the world they can stick somewhat, especially after you begin to recognise them in other music)
Physical sensations from the haptic feedback associated with the notes
The bass notes providing the context for the melodies or inner lines being sung or vice versa
There is one additional memory anchor that is worth taking advantage of and that comes from the other part of the Neapolitan method: Hexachordal Solfége.
Inspired the partimento work of Robert Gjerdingen, Nicholas Baragwanath spent a great deal of time reconstructing the solfeggio tradition of the Neapolitan conservatories and through his book showed how they complement partimento study, teach melody and aid in memorisation by turning music into “words” in context. [13] The hexachordal approach is different from modern movable do in that the typical melodic paths to key changes are better accounted for by the hexachordal system by better preserving the intervallic relationships between melodic fragments. For example, within the appropriate hexachords, C–D–E and G–A–B can both be sung as do–re–mi. The same syllables can therefore identify a recurring interval pattern in different tonal locations.
By adding hexachordal solfege syllables to the stock patterns, a fourth memory anchor is added to the learning process. The Neapolitan orphans themselves actually first had to be able to read and speak the appropriate solfege syllables before they could do anything else and were able to do so in up to 9 different clefs (IIRC)!
So to summarise the learning method would like this:
The software takes the learner through the partimento method, gradually teaching them stock phrases, starting with the phrases most used elements: segments of the rule of the octave, cadences
The fragments are initially taught in 2 voice versions and once they demonstrate that they have learned the 2 voice version they move on to 4 voice versions
Here is possible way each phrase could be learned:
Listening to the fragment with bass only first, with the haptic gloves teaching the fingering, and showing the appropriate solfege syllables on the screen.
Listening to the fragment with melody and the bass played at a lower volume and with the gloves teaching the fingering of the melody, while showing the appropriate syllables on the screen.
Listening to the fragment with melody and bass play at equal volume and both gloves teaching the sensation of simultaneous playing of the fragment
Then the user repeats the above but now sings the melody or bass line while the gloves still teach the fingering through haptic feedback - initially leading the user - not waiting for them.
Then the user repeats the singing exercise for both melody and bassline but this time the gloves only trigger when they sing the right note.
They go through enough fragments that can fit into the first 15 minutes of that day’s lesson. Any more would be too intensive based on my experience.
Once they are through all these repetitions away from the keyboard, they take off the gloves and attempt to play what they had just learned with incomplete prompts (no bass, no melody, etc) at the keyboard, checked by the program through a midi connection. Emphasis will be on playing the correct notes rather than perfect rhythm initially - this would be the other 15 minutes of their daily practice routine
Additional modules for teaching basic audiation (the ability to sing back a heard note) reading sheet music (using flash cards), basic lessons explaining the theory behind partimento and appropriate keyboard technique would be added once the prototype is advanced enough to justify spending time on these modules.
Once the learners have mastered the 2 voice versions of a particular phrase, they relearn it in a 4 voice version.
The user is then free to improvise as much as they like with what they have learned so far.
The goal is to get the user over the initial hurdle of music making as fast as possible so that they start producing “output” as soon as possible. The hope of course is that the ability is so compelling that they keep coming back to learn more.
Once they have mastered enough of the basics to complete/realise a partimento (phrases in C major, F major and G major for instance), they would then be introduced to the task of realising a historical partimento (given bass line) and accompanying the task of accompanying a simple melody.
They would progress through partimentos of increasing complexity as they learn and more of the phrases needed to realise them.
My best guess is that this approach could take almost anyone to the level of a competent improviser and accompanist in the historical style if they put in the 30 minutes a day required, at least 5 times a week.
There is one caveat to all this: some situations may require alternative fingerings depending on the hand size of the learner. This is something that can only be worked out as the prototype evolves and more and more people test it.
This the cheapest haptic feedback glove (350 USD per pair + shipping) that may be feasible for this:
Finger actuators would cue fingering patterns. Palm and wrist actuators could potentially indicate hand movements and leaps, although the usefulness of those signals would need further testing and would be more relevant for the commercialisation part of the project: teaching popular piano pieces and the stock phrases of the romantic era composers (akin to what En Blanc et Noir shows with improvising waltzes).
Path to commercialisation
If the interface works, a commercial enterprise could be built around teaching partimento as a the musical "language" foundation and further modules can be added to teach people the patterns of romantic era harmony or other popular styles like jazz and pop. Additionally, popular piano pieces (in the public domain) that most learners are interested in could be sold as individual elements - a bit like sheet music sales, but instead of notes the learner buys the PHFL patterns and the chunking of the piece into simple elements to learn the music.
The baseline approach of singing the musical lines and passively acquiring the fingering would be the same. As mentioned previously, I think the main issue with PHFL and why it hasn't become commercially successful is because the technology is being pointed at the wrong problem to solve. It should be used to make the acquisition of musicality through removing as much of the friction involved in learning the mechanical skill of learning to play the keyboard as possible rather than focusing on purely teaching the correct order of the "button presses". This is why I am convinced all the learning elements need to focus on singing.
My best guess of what the revenue model could look like:
Sale of the haptic feedback gloves (and any subsequent accessories)
Sale of stock phrase bundles for learning improvisation in other styles (romantic, jazz, pop, etc)
Sale of individual pieces or bundles of pieces by composers for hobbyists who want an accelerated means of learning those pieces (eg.: Mozart Rondo Alla Turca for 1-5 USD, Bach WTC Book 1 and 2 for 40 USD).
Or all of this could be offered in the form of subscription tiers.
US PATENT 10,121,388 - the original on PHFL
I know PHFL is patented in the US and as far as I understand it, there is no equivalent patent outside of the US and no one else can claim one either because of the Prior Art clause. So, based on a cursory assessment there is no legal barrier to testing and commercialising this outside the US.
Therefore the best candidate for creating an Enterprise in service of this is either Thad Starner's company, particularly if they are still working on using PHFL for music or an organisation of the outside of the US, if work on this from Thad Starner has stalled or he is not interested.
References
[1] The New York Times. When A.I. Comes for Your Purpose in Life. The Daily. Interview clip.
[2] Pizarro, J. J., et al. (2022). Emotional processes, collective behavior, and social movements: A meta-analytic review of collective effervescence outcomes during collective gatherings and demonstrations. Frontiers in Psychology, 13, 974683. Article.
[3] Pearce, E., Launay, J., & Dunbar, R. I. M. (2015). The ice-breaker effect: Singing mediates fast social bonding. Royal Society Open Science, 2, 150221. Article.
[4] Weinstein, D., Launay, J., Pearce, E., Dunbar, R. I. M., & Stewart, L. (2016). Group music performance causes elevated pain thresholds and social bonding in small and large groups of singers. Evolution and Human Behavior, 37, 152–158. Article.
[5] Coulton, S., Clift, S., Skingley, A., & Rodriguez, J. (2015). Effectiveness and cost-effectiveness of community singing on mental health-related quality of life of older people: Randomised controlled trial. The British Journal of Psychiatry, 207, 250–255. Article.
[7] Mortensen, J. J. (2023). Improvising fugue: A method for keyboard artists. Oxford University Press. Book.
[8] Huang, K., et al. (2010). Mobile music touch: Mobile tactile stimulation for passive learning. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. Article. Full text.
[9] Gemicioglu, T., et al. (2024). Passive haptic rehearsal for augmented piano learning in the wild. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 8(4), Article 187. Article.
[10] Seim, C., Estes, T., & Starner, T. (2015). Towards passive haptic learning of piano songs. IEEE World Haptics Conference, 445–450. Full text.
[11] Brown, R. M., & Palmer, C. (2013). Auditory and motor imagery modulate learning in music performance. Frontiers in Human Neuroscience, 7, 320. Article.
[12] Pfordresher, P. Q., & Halpern, A. R. (2013). Auditory imagery and the poor-pitch singer. Psychonomic Bulletin & Review. Article.
[13] Baragwanath, N. (2020). The solfeggio tradition: A forgotten art of melody in the long eighteenth century. Oxford University Press. Book.
And the TL;DR is great! Now I am actually interested.
I don't have time to read the whole thing or engage deeply. But: what about doing this, at least as a demo and test of concept, without new haptic hardware? I think you could do this with a laptop app with a visual keyboard. A very little practice mapping from visual to piano keyboard strikes should suffice, then you can run pretty much whatever training scheme you want. You could vibe code it in a couple of days.
Sorry if I'm missing something. Just trying to be helpful on a narrow time budget.
I think the TLDR should contain a bit more. People's time is limited, so giving them maximum information right at the top is a standard way to say you won't waste their time on a long post where they don't know if they find the central ideas important or interesting enough to be worth their time.
Thanks, happy to hear you found it interesting! I will try to make the TLDR much better when I have time later today.
Regarding your suggestion: the goal of the tech is to combine the act of singing the parts and the passive haptic feedback learning stimulus from the gloves to avoid the need for manual repetition of fingering patterns and shift all of the learners focus on audiation and internalising the music.
Your visual keyboard suggestion is something I have tried before using synesthesia. It is certainly helpful but still means you have to do the manual repetitions AND spend time separately learning to sing the phrases. This way are not reducing the learning time by minimising the need for repetitions at the keyboard, you are only adding a visual aid to the mechanical part of the work.
Mea culpa: I hear your downvotes loud and clear. Asking for funding to implement these ideas was in bad taste and I apologise. I guess this is my version of AI psychosis, I have removed it from the post. Still, I think the 4 main ideas in my post have merit and I genuinely would like feedback on it and would love to see it implemented for the good of humanity:
I will happily read and engage with any suggestions or counter arguments to any of the above. I will try and build my own first prototype over the winter holidays and report back on the results.
TLDR:
Epistemic status: Mix of exploratory and my best guess. I do try to substantiate my claims with relevant research where possible.
I’ve structured my idea into the following 3 segments and an appendix:
Shared sacredness as antidote
Seeing a Professor of Applied Mathematics cry and quote Ecclesiastes in an NYT interview [1, Interview clip] in response to the newest developments in AI driven maths proofs moved me so much that I couldn’t shake the feeling that something has to be done to prepare for the coming psychological fallout. I had already read many accounts on reddit and elsewhere of people talking about how AI is eroding their sense of purpose and meaning in their job as it takes over their role, and I can sense it too, that eventually my white collar job will in large part be eliminated by AI.
I have felt purposeless for a very long time, even before AI because I never found my vocation in life and stumbled around from hobby to hobby until I was finally diagnosed with ADHD and started medication. That fixed everything except the source of meaning part. I tried all the usual approaches to finding meaning before that: meditation, jhanas, learning to compose music, play the piano, draw, cook, code, judo, reading a lot about psychology, studying thinking models (thank you lesswrong), AI, etc. I’ve run the gamut of available interests and couldn’t find meaning™ in any of them long term. I also couldn’t overcome some of the hurdles in certain activities to reach advanced levels of competence, namely: composing, piano and drawing. What stuck with me is the urge to become a composer (even though I didn’t really act on it), which eventually led me to discover the partimento method. It felt like discovering the holy grail of music education: this is what all the classical greats studied. Eureka! Well, when I first found out about it I hadn’t started ADHD medication yet and doing the mind-numbingly boring amount of repetitions it would take to become proficient in anything musical prevents even most neurotypicals from making progress, let alone someone like me, so I didn’t make any headways with the method and eventually abandoned it.
In the meanwhile my search for meaning had gotten so desperate, I tried to become religious and studied the bible and thought a lot about going to church just for the communal singing. I could never bring myself to do it because every time a religious friend invited me to some special occasion like a christening, I would experience what I now know as shared sacredness while singing, despite not knowing the words or melody of the hymn or believing in anything being said. Once the singing was over that blissful experience would then be tainted by feeling like a phony outsider infiltrating the congregation because real believers do have a tendency to look at you with an earnest hope in their eyes that you too are a believer. Or at least that’s what it felt like. And of course it wasn’t everyone, and I am not trying to imply that I went to a church full of fanatics but it was common enough to be off-putting. Nevertheless, I knew from then on that my life would be better if communal singing was a regular part of it, I just didn’t know how to find the right community to do it with.
Beyond my anecdotal experience there is some research backing up the idea that shared sacredness experience (or collective effervescence if you prefer Durkheim’s original term) is a fairly universal phenomenon and reliably achieved through singing together. A meta-analysis of this phenomenon links such experiences with wellbeing, social integration and collective identity. [2]
Communal singing probably works as a “social technology” for evoking shared sacredness experience even in non-religious people because it combines several potentially relevant features:
Singing groups have shown faster initial social bonding than comparison classes, although those differences diminished over time. [4] Another study found increased social closeness and measured improvements in pain thresholds following group singing. [5] A randomised trial involving 258 older adults found improved mental health-related quality of life following a community singing programme. [6] I think all of this validates my own experience and desire for something like this and makes it worth investigating as means of improving people’s resilience to the meaning crisis that already seems to be on our doorstep.
Let’s say my anecdotes and the research point in the right direction. We still lack multiple things to start implementing this en masse or even on a small scale:
I don’t have a globally applicable answer to all 3 but the partimento method and professional organist training offer a strong starting point:
I think on some level the novelty of offering historical style improvisations to accompany singing could itself help boost the initial interest in such events. This leads me to the tech solution part of my proposal.
Passive haptic learning and partimento
As I mentioned above only classical organists learn to improvise as part of their education and at least in the German music schools, the rediscovered methods of partimento are enjoying a revival of sorts. In French organ schools they have stuck with their original methods as far as I am aware but in essence the French approach was already close enough to partimento because its origins can also be traced to the methodology originating from Naples.
What is partimento? Partimento is the keyboard pedagogy part of the education that orphaned boys received in Neapolitan conservatories. Originating from somewhere in the 1600s the method sought to teach them the basics of harmony in order to provide them with stable employment as church musicians. Robert Gjerdingen and colleagues have uncovered and translated an entire treasure trove of manuscripts on what they were taught, practically covering the entire beginner to advanced development levels. Many of them are freely available on partimenti.org. The partimento method essentially teaches a set of bass motions and sequences (rule of the octave, circle of fifths progressions, etc), cadences (the punctuation points in this style of music), common paired bass-melody patterns called schemata (that are so ubiquitous in the historical style that scholars have given them names) and methods of introducing and treating dissonances. All of this can be thought of as a set of stock phrases that can be memorised much like the way one would memorise many example sentences when learning a language. [6], [7]
In order to become fluent in this keyboard improvisation style, about 500 - 1000 stock phrases of about 4-8 notes in length need to be learned and internalised before they can be used as interchangeable modules of musical phrases to be used in improvisation and composition. The end result sounds very exciting initially doesn’t it? The hard truth is that the amount of repetition at the keyboard needed to get there is mind-numbingly boring. Not something you would do unless you were an orphan desperate for employment. Not to mention you have to both be able to sing them and play them at the keyboard if you want employment at a fine establishment in Naples (more on the singing part soon).
Passive haptic feedback learning(PHFL) to the rescue!
Here is a clip from Thad Starner on how it works: https://www.youtube.com/watch?v=uEdr6iY6F-w
The video shows one of the CNN anchors learned to play the melody of Ode to Joy on the piano in half an hour while browsing his emails and doing his usual work related tasks. All the glove did was buzz the correct note sequence into his fingers. It's really fun to see how astounded the anchor is that he can play it all. Not very well of course but he does press the buttons in the right order. Which is exactly what PHFL learning can accomplish: teach the button presses in the right order.
In an early Georgia Tech experiment also led by Thad Starner, five musical novices required approximately three keyboard test attempts on average following passive haptic training, compared with seven during active training to be able to play with 90% accuracy. [8]
Source: Huang et al. [8]
More recent work combined passive rehearsal with active piano practice. A two-week study recruited 36 participants; its main analysis used 20 participants with complete, valid data. Passive rehearsal increased the modelled learning rate for note accuracy by 49.7%, but did not significantly improve rhythm learning. [9]
Two-handed learning has also been investigated directly. A 2015 study explored two-handed piano material and reported better passive-learning results when both hands were trained simultaneously rather than sequentially. [10]
So why don’t we have this tech commercially available yet?
My best guess is because, as jazz teacher Hal Galper likes to say “the instrument is an illusion”. PHFL only teaches the correct button presses which for music is only half the battle of learning. The other half is audiation and singing. In other words learning to think (audiate) and speak (sing) in the language of music.
I know this from first hand experience, anything I can sing (even badly) I can play more or less, depending on how much I practiced the motor skills needed. I am not saying it's not possible to overcome the need for audiation, I am guilty of learning patterns I can’t audiate and perform them purely mechanically and I suspect many amateur shred guitarists are guilty of this too (sorry, couldn’t resist).
Experimental work also shows that auditory and motor imagery abilities influence musical learning and recall. [11] Research has also found an association between auditory imagery vividness and singing accuracy. [12] So while I am fairly convinced of this being the missing element based on anecdote and research, it is only fair to acknowledge that this would only be substantiated if the haptic learning interface and the suggested curriculum is tested on multiple cohorts of untrained or amateur musicians.
So, where does the 2x learning speed come from exactly?
Well let’s sum up, what’ve explored so far:
What I have outlined so far gives us 3 memory anchors for memorising these musical phrases:
There is one additional memory anchor that is worth taking advantage of and that comes from the other part of the Neapolitan method: Hexachordal Solfége.
Inspired the partimento work of Robert Gjerdingen, Nicholas Baragwanath spent a great deal of time reconstructing the solfeggio tradition of the Neapolitan conservatories and through his book showed how they complement partimento study, teach melody and aid in memorisation by turning music into “words” in context. [13] The hexachordal approach is different from modern movable do in that the typical melodic paths to key changes are better accounted for by the hexachordal system by better preserving the intervallic relationships between melodic fragments. For example, within the appropriate hexachords, C–D–E and G–A–B can both be sung as do–re–mi. The same syllables can therefore identify a recurring interval pattern in different tonal locations.
By adding hexachordal solfege syllables to the stock patterns, a fourth memory anchor is added to the learning process. The Neapolitan orphans themselves actually first had to be able to read and speak the appropriate solfege syllables before they could do anything else and were able to do so in up to 9 different clefs (IIRC)!
So to summarise the learning method would like this:
This the cheapest haptic feedback glove (350 USD per pair + shipping) that may be feasible for this:
Image source: https://www.bhaptics.com/en/shop/buy-tactsuit/tactglove-dk3/
Finger actuators would cue fingering patterns. Palm and wrist actuators could potentially indicate hand movements and leaps, although the usefulness of those signals would need further testing and would be more relevant for the commercialisation part of the project: teaching popular piano pieces and the stock phrases of the romantic era composers (akin to what En Blanc et Noir shows with improvising waltzes).
Path to commercialisation
If the interface works, a commercial enterprise could be built around teaching partimento as a the musical "language" foundation and further modules can be added to teach people the patterns of romantic era harmony or other popular styles like jazz and pop. Additionally, popular piano pieces (in the public domain) that most learners are interested in could be sold as individual elements - a bit like sheet music sales, but instead of notes the learner buys the PHFL patterns and the chunking of the piece into simple elements to learn the music.
The baseline approach of singing the musical lines and passively acquiring the fingering would be the same. As mentioned previously, I think the main issue with PHFL and why it hasn't become commercially successful is because the technology is being pointed at the wrong problem to solve. It should be used to make the acquisition of musicality through removing as much of the friction involved in learning the mechanical skill of learning to play the keyboard as possible rather than focusing on purely teaching the correct order of the "button presses". This is why I am convinced all the learning elements need to focus on singing.
My best guess of what the revenue model could look like:
US PATENT 10,121,388 - the original on PHFL
I know PHFL is patented in the US and as far as I understand it, there is no equivalent patent outside of the US and no one else can claim one either because of the Prior Art clause. So, based on a cursory assessment there is no legal barrier to testing and commercialising this outside the US.
Therefore the best candidate for creating an Enterprise in service of this is either Thad Starner's company, particularly if they are still working on using PHFL for music or an organisation of the outside of the US, if work on this from Thad Starner has stalled or he is not interested.
References
[1] The New York Times. When A.I. Comes for Your Purpose in Life. The Daily. Interview clip.
[2] Pizarro, J. J., et al. (2022). Emotional processes, collective behavior, and social movements: A meta-analytic review of collective effervescence outcomes during collective gatherings and demonstrations. Frontiers in Psychology, 13, 974683. Article.
[3] Pearce, E., Launay, J., & Dunbar, R. I. M. (2015). The ice-breaker effect: Singing mediates fast social bonding. Royal Society Open Science, 2, 150221. Article.
[4] Weinstein, D., Launay, J., Pearce, E., Dunbar, R. I. M., & Stewart, L. (2016). Group music performance causes elevated pain thresholds and social bonding in small and large groups of singers. Evolution and Human Behavior, 37, 152–158. Article.
[5] Coulton, S., Clift, S., Skingley, A., & Rodriguez, J. (2015). Effectiveness and cost-effectiveness of community singing on mental health-related quality of life of older people: Randomised controlled trial. The British Journal of Psychiatry, 207, 250–255. Article.
[6] Gjerdingen, R. O. (Ed.). Monuments of Partimenti and Solfeggi. Collection and editorial statement. See also Fenaroli’s Book III.
[7] Mortensen, J. J. (2023). Improvising fugue: A method for keyboard artists. Oxford University Press. Book.
[8] Huang, K., et al. (2010). Mobile music touch: Mobile tactile stimulation for passive learning. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. Article. Full text.
[9] Gemicioglu, T., et al. (2024). Passive haptic rehearsal for augmented piano learning in the wild. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 8(4), Article 187. Article.
[10] Seim, C., Estes, T., & Starner, T. (2015). Towards passive haptic learning of piano songs. IEEE World Haptics Conference, 445–450. Full text.
[11] Brown, R. M., & Palmer, C. (2013). Auditory and motor imagery modulate learning in music performance. Frontiers in Human Neuroscience, 7, 320. Article.
[12] Pfordresher, P. Q., & Halpern, A. R. (2013). Auditory imagery and the poor-pitch singer. Psychonomic Bulletin & Review. Article.
[13] Baragwanath, N. (2020). The solfeggio tradition: A forgotten art of melody in the long eighteenth century. Oxford University Press. Book.