Algorithms as Scores: Coding Live Music

Source: Thor Magnusson, “Algorithms as Scores: Coding Live Music,” Leonardo Music Journal 21 (2011), pp. 19–23. (Magnusson, 2011) Local PDF

Central claim

An algorithm can act as a musical score because it represents musical thought, prescribes machine action and generates varied realisations. Software can therefore make score, instrument, composition and performance hard to separate.

Evidence and method

Magnusson gives a historical and conceptual account of the score, algorithmic music, graphic notation and live coding, then uses his ixi lang system as a case study. The article does not report a controlled study, though it refers to user feedback about the system’s constraints.

A score can describe sound, prescribe performance, store music and let a composer externalise and develop musical thought (p. 19). Algorithms suit generative results, though their non-linear and variable nature requires a form beyond the fixed sequence (p. 20).

In live coding, code represents what happens in sound and reduces cognitive load by placing musical thought outside the performer (p. 21). A language’s constraints allow some compositional practices and prevent others, blurring the divisions between tool and work as well as composer and performer (p. 22).

Magnusson also argues that a mobile app can contain enough score, instrument, compositional system and music to count as a musical piece (p. 22). A score written or executed in real time supports generative music rather than one fixed realisation (p. 23).

Key concepts

• Algorithm as score: Rules prescribe a range of machine actions rather than one fixed sequence.
• Externalized musical thought: Code stores and exposes compositional decisions.
• Generative realization: Each execution can produce a different audible result under stable rules.
• Constraint: A language or system makes some musical actions easy and others unavailable.
• Blurred roles: Work, tool, composition and performance may occupy the same software system.

Limits

Live coding exposes and changes its score in front of an audience, while most functional-music apps hide their code and limit listener control. The claim that apps were replacing MP3s reflects a 2011 media context rather than current market evidence. The article does not study passive listening, physiological effects, platform economics or proprietary generative services.

Link to generative functional apps

The algorithm-as-score concept describes musicianship as the composition of possible behaviour rather than fixed events on a timeline. A musical work can include a sound library, rules, constraints and each runtime realisation.

Live coders expose and revise the score during performance. Endel-like systems also generate during listening, though the audience cannot inspect or change their rules. The company that owns the runtime system therefore holds more control within a distributed form of authorship.

references