Tuesday, 25 April 2017

Revisiting Cybernetic Musical Analysis

I had a nice email yesterday from a composer who had seen a video of an analysis of music by Helmut Lachenmann which I did in 2009 using cybernetic modelling. I'd forgotten about it - partly embarrassed by what I thought of as a crude attempt to make sense of difficult music, but also because it was closely related to my PhD which I'm also slightly embarrassed by. In the intervening time, I became dissatisfied with some of the cybernetic underpinnings, and became more interested in critical aspects of theory. Being embarrassed about stuff can be a block to taking things further: I have so many almost-finished unpublished papers - often discouraged from publishing them because of the the frequent nastiness of peer review. So it's nice to receive an appreciative comment 8 years after something was done.

It's made me want to collate the set of music analysis videos that I made in 2009. They are on Schumann, Haydn, Ravel and Lachenmann. In each I pursue the same basic theory about "prolongation" - basically, what is it that creates the sense of coherence and continuity of experience in the listener. The basic theory was inspired by Beer's Viable System Model - that coherence and continuity is a combination of different kinds of manipulation of the sound as "disruptive" - sound that interrupts and surprises; "coercive" - sound that reinforces and confirms expectations; "exhortational" - sound that transforms one thing into another.

What do I think about this 8 years later?

First of all, what do I now think about the Viable System Model which was the foundation for this music analysis? There is a tendency in the VSM to refer to the different regulating layers allegorically: this is kind-of what I have done with coercion, exhortation, etc. But now I think the VSM is more basic than this: it is simply a way in which a system might organise itself so as to maintain a critical level of diversity in the distinctions it makes. So it not that there is coercion, or disruption or exhortation per se... it is that the system can distinguish between them and can maintain the possibility that any of them might occur.

Furthermore, each distinction (coercion, exhortation, etc) results from a transduction. That it, the conversion of one set of signals into another. Particular transductions attenuate descriptions on one side to a particular type which the transduced system can deal with: so the environment is attenuated by the skin. But equally, any transducer is held in place by the descriptions which arise from the existence of the transducer on its other side. It's a bit like this:

Each transducer attenuates complexity from the left and generates it to the right. This is where Beer's regulating levels come from in the Viable System Model. 

The trick of a viable system - and any "viable music"  (if it makes sense to talk of that) - is to ensure that the richness of possible transductions and descriptions is maintained. In my music videos I call this richness "Coercion", "exhortation", "disruption" - but the point is not what each is, but that each is different from the others, and that they are maintained together.

Understanding transduction in this way gives scope for saying more about analysis. The precursor to a transducer forming is an emerging coherence between different descriptions of the world. A way of measuring this coherence is to use the Information Theoretical calculation of relative entropy.  I've become very curious about relative entropy since I learnt that it is the measure used in quantum mechanisms for measuring the entanglements between subatomic particles. Given that quantum computers are programmed using a kind of musical score (see IBM's Quantum Experience interface), this coherence between descriptions expressed as relative entropy makes a lot of sense to me. 

So in making the distinctions that I make in these music videos, I would now put more emphasis on the degrees of emerging relative entropy between descriptions. Effectively coherences can be seen in what I called "coercive" moments as repetition, and this produces descriptions of what isn't repeated - or what is surprising. Surprises on larger structural layers such as harmony or tonality amount to transformations - but this is also a higher-level transduction.  

The viable system which makes these distinctions is, of course, the listener (I was right about this in the Lachenmann analysis). The listener's system has to continually recalibrate itself in the light of events. It performs this recalibration so as to maintain the richness of the possible descriptions which it can generate. 

The world is fucked at the moment because our institutions cannot do this. They cannot recalibrate effectively and they lose overall complexity and variety - and consequently they lose the ability to adapt to changing environments. 

Here are the videos:





No comments: