Thursday, August 21, 2008

Intuition

There are some decisions that we make and we know how we made them; we can fully describe the process by which we arrived at a particular decision. There are other decisions that we make (and feel confident it's the necessary or right decision) without being able to describe systematically the process by which we arrived at that decision. And there are decisions that we make arbitrarily, either because all possible choices seem equally valid, or because the decision is too trivial to spend time thinking about, or perhaps because at some very low level there truly is an element of randomness or chance that plays a role in the course of events.

We can name these three modes of decision making systematic, intuitive, and arbitrary. System is a prescribed set of procedures to be followed to arrive at a result. Intuition is a way of knowing that is not formulated as a system but is nonetheless effective. Perhaps it is a system that has not yet been formulated, or perhaps it is a wholly different way of knowing. Arbitrary decision making needs no special system because all results are equally acceptable. In reality we probably make decisions using some complicated combination of the three modes.

There is a lot that goes on in the working process of most composers that they would not be able to formalize as a rule-based or procedure-based system. Most composers use a combination of systematized or quasi-systematized knowledge, intuition, and probably at low, trivial levels, arbitrary decision making.

This set of essays is concerned primarily with systematic decision making, which is the type that lends itself most readily to computer programming. But let's take a look at how system relates to intuition and chance.

I have proposed here a use of the word "intuition" that might not be agreeable to everyone, but will be the working definition for the purposes of this essay: a means of decision making that is intentional and in which we have confidence, but for which we have not formalized a system. In short, things we know but don't know how we know them. It might be that we are in fact using a system of which we are not fully conscious, or it might be that it's a different sort of knowledge that is not encompassed by rationalist logic.

To teach a computer to use this sort of intuition would seem to be inherently impossible if, by definition, we don't know how to explain intuition. And indeed, for a computer to make autonomous decisions that aren't fully pre-determined by the system that is its software requires that at some level its algorithm must use a form of randomness, which is to say arbitrariness. A computer can readily enact a fully described system of procedures, and it can also enact arbitrary decisions using pseudo-random processes. But how can a computer enact or emulate intuition?

If we accept the premise that intuition is "things we know but don't know how we know them", then we could investigate intuition by trying to figure out how we know the things we know intuitively. To the extent that we can describe or emulate intuition by a formal system, we can gain insight into the nature of intuition. If one could imitate intuition with increasingly probing systems, until we arrive at a level where an arbitrary decision can be shown to be at the lowest level, we can show that valid compositional methods might be totally systematized, with unpredictability and variety provided by pseudo-randomness.

Why does a composer or any artist choose one thing and not another? I propose that that "why" is ultimately reducible to a complex algorithm of "hows". That is to say, we may consider the explanation of why something is the way it is (Why do I like chocolate ice cream better than strawberry?) to be equal to the explanation of how that state was achieved. (By what mental process do I arrive at the discernment that chocolate is preferable?) The idea that decisions can be explained algorithmically is at the very heart of the field of algorithmic composition. Computers only know how to do things. They carry out instructions with no inkling or concern as to why they are doing them. Therefore, the business of programmers of artificial creativity is to turn whys into hows.

Let's take the example of a composer selecting a pitch to write on the page. Assuming that the composer has already decided to use only the 88 possibilities presented by the piano (or 89 if we include the "null" note, silence), some criteria for decision making are obviously necessary. A number of aesthetic criteria may be used by the composer in choosing a pitch: melodic contour, harmonic implications, etc. But the choice need not necessarily be based on aesthetic criteria. The composer may have a pre-established system (an algorithm, a list, etc.) or the choice may be made arbitrarily (by aleatoric means). In these instances the composer would simply be following established rules of decision making, which is something that computers do better and faster than humans. But the existence of those rules implies some prior aesthetic decision, either of commission or omission. An algorithm is being used because the composer decided at some earlier time that that algorithm would lead to a desired aesthetic result. How did the composer arrive at that decision? That previous aesthetic decision was presumably made using one of those same three means: systematic (using a system that is itself based on earlier aesthetic decisions), intuitive (using a system that has not yet been fully and consciously formalized), or arbitrary (using some unknown criteria or no criteria). So we see that rule-based decision making can always be traced back to some prior choice, either systematic or arbitrary.


When we try to trace aesthetic criteria themselves back to prior choices (By what criteria did we decide to use those criteria?) we may finally arrive at some seemingly banal conclusion such as "I don't know" ("I made the decision intuitively") or "It didn't matter" ("I made the decision arbitrarily"). The type of conclusion we reach in this genetic reconstruction of a compositional decision has implications of how to proceed to enact a similar decision by computer. A seemingly intuitive decision might be elucidated by further analysis of the underlying system. A seemingly arbitrary decision implies that randomness can be the source of desirable aesthetic results.

If we justify an intuitive aesthetic decision with "I just like it that way" we invoke an attribute called taste. Taste is a much-used term to describe a trait or criterion of aesthetic decision making, but no conclusive definition of taste has really been established. So far we don't know of a way for a computer to exercise genuine human taste or intuition, but randomness (or a very good facsimile thereof) and procedural systems are no problem at all for a computer.

Almost all computer programs that make autonomous decisions employ randomness on some level. Total randomness--also known as "white noise"--is rarely of aesthetic interest to most of us for very long. We tend to desire some manifestation of an ordering force that alters the predictably unpredictable nature of white noise. To produce anything other than white noise, a computer program must contain some non-arbitrary choices made by the programmer. Therefore, no decision-making program can be free of the knowledge and, yes, taste and intuition of the programmer.

For these reasons, we can analyze algorithmic composition--the process of programming a computer to make music using systematic and arbitrary procedures--as a complete and potentially fruitful method of composing music and/or other arts.

2 comments:

Jeff said...

"How did the composer arrive at that decision? That previous aesthetic decision was presumably made using one of those same three means: systematic (using a system that is itself based on earlier aesthetic decisions), intuitive (using a system that has not yet been fully and consciously formalized), or arbitrary (using some unknown criteria or no criteria). So we see that rule-based decision making can always be traced back to some prior choice, either systematic or arbitrary."

I guess for rule based decision making this could be the case, but I wonder if another form of knowledge might have far more influence on these processes: the pattern matching function of the mind. Watching FMRI video of thought processes in flight, it seems like decisions which are being attributed to either intuitive or arbitrary means could instead be bubbling up from our highly associative memory structures. Processes of perception and free association merge to generate a stream of concepts that compete for our consciousness. If a process like this is at the heart of decision making, then I wonder what the limits are for with a purely rule based system that doesn't incorporate an associative memory?

loveslap said...

Are there any generative composers who completely eschew randomness?