What should a listening machine be able to do?

This week we'll have a kick-off meeting preparing for our project Zonder luisteraar geen muziek [No music without a listener] that has been nominated for the Dutch Academic Year Prize.

One of the questions the UvA-team (consisting of Vivienne Aerts, Shane Burmania, Olivia Ladinig, and me) will brainstorm about is: Imagine what a listening machine would look like, a machine that is able to listen and react in a human and musical way. What should such a machine know, what should it listen for, how could it respond, and how can we compare and evaluate such machines? One of the challenges is how to turn such a question into a compelling and fun problem to think about, aimed at students that want to combine their interest in music with a liberal arts and sciences education.

While it might look like a simple question, the design of a 'listening machine' that embodies the musical and listening skills common to most humans turns out to be a full-fledged research program, and it is part of the scientific enterprise generally known as music cognition. In this field of research computational modeling (formalizing a theory in the form of a computer program and relating it to human behavior) is an influential methodology that has contributed to a further understanding of music as a process in which the performer and the listener play a central role.

While for a long time music was a topic hidden away under subject headings like 'pitch' and 'time perception' in scientific reference books, in recent years several disciplines, ranging from the humanities to the social and natural sciences, show a growing interest in the scientific study of music. A recent example is Robert Zatorre who promotes music as “the food of neuroscience” (see Nature). It looks like the beginning of something ...