The concept of this project was to create a group of autonomous,
self-contained physical objects that could sing to each other and
hear each other singing and respond according to computational al-
gorithms. The idea was that simple behavior could be programmed
into each agent and that together, complex group behavior would
arise. Agents would switch between a state where they are listening,
and another state where they respond. During the listening state
they would determine the loudest note that they heard and use
that as the input to a first order Markov model. With all the agents
separately playing single notes from the same Markov model, the
idea was that the whole group behavior would be similar to that of
a single entity playing a single Markov-model. We considered this
to be a sort of “distributed Markov model.” Additionally the average
loudness that each agent is hearing at any given time will have an
effect on the way it plays the notes. When the average loudness is
higher, they will shorten their listening period, playing notes more
often and moving through the markov model more quickly, and
they will also play shorter notes creating a frantic feel. Additionally
there would be a few ways that a user could interact and guide the
behavior. The primary method of interaction is simply picking up
the agents and moving them, if they are further apart from each
other they will not hear each other as loudly, decreasing the av-
erage loudness and when far enough away, making some agents
not interact at all. Another way that the users interact is through
a special agent called the “Leader Bird.” This agent plays a special
command pitch when a button on it is pressed. All of the follower
agents have four separate Markov models built on different pitch-
class sets and they all start on the same one. When an agent hears
this command pitch they will move to the next pitch-class set. Since
they might not all hear the command pitch depending on where
they are placed, they will likely begin to drift into groups playing
different pitch-class sets moving from a state of consonance in the
beginning of the piece to a polytonal cloud as time goes on.