In artificial life research, the target of study is life and its salient properties—evolution, reproduction, autopoiesis, cognition, etc. To study these, we use computer simulations, mathematical models, philosophical arguments and we employ diverse methods of analysis. Each choice of methodology influences the kinds of phenomena we can uncover. For example, information theoretical anaylsis and dynamical systems analysis each reveal and emphasize different features of a dynamical neural network (Beer and Williams,2014).
The metaphors that we use also shape our research. If we consider cognition to be a form of computation, we end up studying problems that are readily presented to a computer, like chess (Risi and Preuss, 2020). If, on the other hand, we consider cognition to be the result of dynamical interplay between coupled brain, body and world, then we find ourselves considering cognitive tasks that can be readily formalized as fitness functions—e. g. investigating categorical perception in an embodied robot that can use whisker-like sensors to distinguish between circle and diamond shaped objects (Beer, 2003).
What we choose as an exemplar of the target of study also biases our research. If human cognition is the target, which of its many remarkable abilities should we focus on to understand it? Our ability to solve mazes or puzzles? Our ability learn a new sporting ability? Our ability to detect liars? Many examples of human cognition focus on our ability to solve problems that have well defined and pregiven criteria of success. Human problem solving ability in these contexts can indeed be remarkable, but so too are some of our abilities where success is not so easily defined or pre-given.
Particularly interesting examples of this include expressions of creativity and improvisation. These abilities, creativity and improvisation, are key parts of human playfulness, and musical creativity is a place where play persists a bit longer than those other forms of play that we associate with children. What can we learn about human cognition by studying musical improvisation? For about 30 years I have played the drums. Recently, af-
ter a long break, I have returned to practicing and performing at regular ‘jam nights’ hosted at our local bar. This is the first time that I have been regularly practicing and performing since becoming a scientist and I find that I think about music in a differently now thanks to things that I’ve learned along the way. For example, when I might have taken it for granted before, I now find it quite amazing, from an adaptive systems perspective, that strangers can come together, and spontaneously perform an unplanned creative musical performance that coheres—a performance that did not exist in anyone’s ‘head’ before the performance and that
cannot be reduced to any individual, nor even to the entire band, as the band is not just responding to itself, but also to dancers, other audience members, and occasionally the political events of the day. In a way, this kind of performance, creates itself, and while it lasts it maintains itself. It adapts when a guitarist’s string breaks, the music carries on. As the performance develops, so to develop norms for what kinds of musical expressions would fit, and which would be errors. Sometimes mistakes are transformed into key themes in the improvised performance. These observations remind me of the self-constructing, and self-maintaining, adaptive,
and self-defining autopoietic nature of living systems (Maturana and Varela, 1980).
The A-Life and enactivist communities have taught me about new frameworks and ways to try to understand, or at
least think about improvisation and drumming and other remarkable human behaviours—concepts like the dynamical
hypothesis in cognitive science (van Gelder, 1998), sensorimotor feedback (Braitenberg, 1986); sensorimotor contingency theory (O’Regan and Noë, 2001); autopoiesis (Varela et al., 1974); autonomy (Barandiaran, 2017); etc. And this paper represents the start of my efforts to bring these ideas to bear on improvisation and on drumming—inspired, in part by the efforts of Torrance and Schumann (2019) to relate jazz improvisation to embodied and enactive cognitive science.
In my talk, I will present a collection of thoughts that have emerged while thinking about the interface between drumming and these different A-Life or A-Life related concepts. My goal will be to stimulate discussion, and a central themein what I will explore will be taking a cyclical or symmetric perspective, rather than one that is linear or asymmetric (Pickering, 2010). In other words: instead of the linear view
where drumming is something that is done by a drummer, I am interested in what comes to light when we instead see drumming as an interaction between drummer and drums. In the first section of my talk, I will explain how skillful drumming is situated, embodied and dynamical (Beer, 2000). In other words how it depends not just upon the drummer’s brain making smart decisions about when to move which muscles, but also bodily and environmental dynamics. To support this point, I’ll describe the difference between a novice’s stiff performance of a two-stroke roll, and an expert’s performance of the same rudiment, which takes advantages of the dynamical properties of the drumstick, drum head and mechanical properties of the drummer’s hands and arms. I’ll also talk about ways that technological advancements in electronic drums have actually failed to fully recognise
the roles played by the drums in drumming. Instead of given the full dynamical complexity that is possible in an acoustic drum, electronic drums often work by using a trigger to either start the playback for a short recorded sound (a ‘sample’) or to excite some simulated model that is used to generate the drum sound. This excessive simplification of an acoustic drum’s complexity is apparent to any intermediate drummer who has had the experience of digital vs. acoustic drums—and it is noteworthy that it is rare to see electronic drums being used in place of acoustic drums by professional musicians. Generally speaking, the variety of tones
one can get using just a stick and an acoustic snare drum is surprisingly diverse. Electronic snare drums by comparison is capable of much less sonic diversity. I find it interesting that the electronic drums fail to compete successfully with acoustic drums. I’ll also connect this dismissal of acoustic drum complexity to GOFAI’s (Good Old-Fashioned AI)’s dismissal of the important roles played by the body the environment, and time.
In the third section of my talk, I will raise and discuss the question: What determines the next note that a musician plays? The answer to this question is arguably relatively straight-forward when the music is composed in advance. In that case, the next note is largely determined by the composer and what they wrote down when the composed the song. When performance is improvised, the answer is much more complicated, and ‘messy’. By ‘messy’ here, I mean that there answer involves many non-linear and interconnected factors that influence what the next performed note will be. In essence, the answer is largely irreducible. An incomplete list of such factors includes: the constraints agreed upon before the song starts—“Let’s jam a blues shuffle in E.”; the technical ability of the musician; the motifs or scales or rudiments that they have practiced; the music they have
listened to; the other musicians (e. g. how the drummer is comping the solo); the audience and how they are dancing (or not!) in response to the performance; and perhaps most interestingly: performer’s own recent performance in the few seconds leading up to that note. I’ll spend some time elaborating on this last point, and the reflexive idea that what counts as a good in a performance is a product of the performance itself. I’ll relate this ‘self-defining’ structure to that proposed in the enactivist ideas of autopoiesis (the self-
constructing nature of living systems (Varela et al., 1974)), where what is good for an autopoietic system is defined by (i. e. emerges from) the way that that system is organized (Barandiaran and Egbert, 2013). In both cases, the system’s norms (what is good or bad) are defined by (the result of) how it produces itself.
If time allows, I will close by providing an overview of a project we have started at The University of REDACTED FOR DOUBLE BLIND REVIEW PROCESS, where we are exploring ways to augment acoustic drums electroni-
cally. This work relates to that of Lupone and Seno (2006), Eldridge and Kiefer (2017), Morreale Morreale et al. (2019), and others, who avoid reducing acoustic instruments to triggers and samples, but instead work to augment instruments, using rich and continuous feedback to enrichen the kinds of sounds that acoustic instruments can produce.