Creativity Dial

to Cathy Woodgold's home page
Send me email

Imagine a "creativity dial": the idea that there's a continuum of types of thinking. You can set the dial high to get lots of new creative ideas, or set the dial low to stay on-task and be organized, but you can't do both at once any more than you can set your fridge to keep the milk cold and the butter warm in the same compartment.

I want to describe my experience in writing a computer program years ago which led me to this idea. Perhaps this description can convey some of the imagery and logic behind the creativity dial concept. I'll try to make it readable for those who don't get along with math well.

There was a contest in Scientific American around 1983 or 4, to create a sentence of the form "This computer-generated pangram contains six a's, zero b's, two c's, ..." and so on, with the number of letters in the sentence correctly described. The tricky part is that the letters in, for example, the word "six" must be counted too. So it's very hard to get it right.

They claimed that it was very, very hard, so hard that nobody would be able to do it on a digital computer. To get an idea of how hard it is: if you assume each number would be between 0 and 9, then trying all the possibilities would mean trying 10 to the exponent 26 possibilities (because there are 26 letters in the alphabet). I suppose that would take a fast computer trillions of years. Even if you narrow it down, restricting the letter e to be between 8 and 25, say, and y to between 0 and 2, etc., it's still a very large number of possibilities to try.

As I pondered how I might solve this problem, I imagined the set of possibilities as a multi-dimensional space. The number of occurrences of each letter is a dimension. I don't visualize multi-dimensional spaces well, so I imagined a two-dimensional space, with the third dimension representing how good the solution was. (Have I lost all the non-math-oriented people?)

I imagined myself standing on a steep mountain. All around me were many, many steep mountains and valleys of various heights. The ground went smoothly but steeply up and down and down and up in all directions. The trick was to find the one spot (actually, there may be a few) where the valley went down to the deepest possible level. I got a lonely feeling as I imagined little me (or my computer program) searching in that huge, unpredictable multidimensional space.

The height of the mountains represents how wrong the sentence is. I added up all the differences between the actual number of letters and the number claimed. For example, if my trial sentence said "ten t's" and there were actually 12 t's (or 8 t's), I would add 2 to the score. Zero, the lowest valley, is the winning score.

I remembered a method someone had described to me: Simulated Annealing. If you're in a high area and you take a step, you'll probably still be in a high area. Or if you're in a low area and take a step, you'll probably still be in a low area. But you never know, and you may need to cross over some mountains to get to the valley you want.

So you start from a trial solution and then take a small step. If the new solution is better, you keep it and use it to start from for the next step. But if the new solution is worse, you don't necessarily throw it away.

If it's a lot worse, you use random numbers (roll dice) to give yourself a very small chance of keeping the new solution. If it's only a little worse, you allow a higher chance of taking the new one. That way, if you're stuck in a small valley, eventually you'll probably accept a worse solution that allows you to climb out and find another valley.

Another image is that the whole landscape is being shaken, and a ball on it tends to go downwards into valleys, but often gets shaken over low walls and occasionally gets enough energy to jump over a high wall.

The method includes a parameter called "temperature" which controls how likely you are to accept worse solutions. It's like the speed of shaking.

In early runs, the program usually had a score of about 2 or 3, but occasionally hit 1 and stayed there for a long time. It didn't get a zero, though, after hours of running.

I then raised the temperature, increasing the chance that the program would accept worse solutions, allowing it to jump out of one valley into the next more easily (speeding up the shaking).

It jumped around, usually having a score of about 4 or 5. But after a relatively short time it hit a zero! I had solved the problem.

The same images apply to all sorts of problems in life. We deal with many real-life problems that have more dimensions than that problem. How many different ways are there to put a jigsaw puzzle together wrong? How many different ways are there to buy 3 objects with $10.00? Creativity can be seen as choosing solutions out of a very large space, a space too large for trying out every possibility.

If the problem is "which of these 4 types of flowers shall I put in front?", there are only 4 possibilities, so you can just think about each one and then choose. But if the problem is "How shall I arrange the flowers?" there are too many ways to try them all out, even in one's mind, so creativity must be used.

The low temperature is like an Organizer mind. Once it finds a fairly good system (score 1), it puts blinkers on and stays there. This is good for some types of tasks.

The high temperature is like an Explorer mind. Even if it finds something pretty good, it often throws it away, and goes on exploring. This is good for many other types of things, and is how creative ideas are formed.

It would be good to be able to control where the creativity dial is set. I'm learning more about how my thinking ability is affected by amount of sleep, time of day etc. When I'm tired I can't do much of any kind of thinking well. But perhaps sometimes I can turn the creativity dial up or down a bit when I want. Certainly when there's a brainstorming session I go wild. I'm working on knowing where it's at at any given time; then hopefully I can move on to turning it up and down. Not down too often! :-)

to Cathy Woodgold's home page
Send me email