Notes on the Tarot

I'm intrigued by the Tarot. Actually by divination magic in general, such as the I Ching, and tea leaf reading and such, but Tarot is my favorite. I don't like astrology as much, though.

The question of whether divination magic works in more subtle than it first appears. It is tempting to dismiss it out of hand: it is obviously impossible to see the future, so how could this stuff possibly work. Well, that objection is completely wrong. It is not impossible to see the future. We do it every day. In fact that's most of what science does to make itself useful. Astronomers predict where the stars and planets will be in the future. Engineers use classical mechanics to tell you how a machine will work before building it. Weather forcasters predict the future all the time (with varying success, but with enough success that they aren't considered deluded fools by the scientific establishment) Seeing the future is an important and achievable goal, at least sometimes.

It is still tempting to dismiss divination, because it tries to use methods that are random: it is obviously impossible to get results from a method that relies on randomness. Again, this objection is wrong. There are whole subspecialties of mathematical modeling devoted to the use of randomness to make predictions. The Monte-Carlo method, statistical sampling, synthetic annealing, genetic algorithms, and many other techniques for developing models with predictive power rely heavily on randomness. One of my favorite quotes on the subject comes from a pop-science book, Reinventing Man: the Robot Becomes Reality by Aleksander and Burnett:

...it cannot be assumed that randomness, or lack of organization, in natural structures is some kind of `second best' solution that would be improved on by the kind of orderly architecture that we are, by training, predisposed to consider superior. On the contrary, the lesson would seem to be that randomness is something that nature finds easy to handle and puts to very careful use.

So if the goals of divination are not impossible and the methods are not obviously ridiculous, it becomes worthwhile to look seriously at what's going on. There are two approaches to take: experiment and theory. Can the methods be tested and shown to work, and can a mechanism be found to describe why they might work. I'm a theoretician by inclination, so I'll try to find a mechanism first.

In the parlance of artificial intellegence research, randomness is often described as "temperature". When searching for the solution to a problem an algorithm will begin looking with a high temperature, meaning that the search has a large amount of ramdomness. As the algorithm collects evidence that it is close to a solution, the temperature becomes lower meaning that the search becomes less random and more directed. The use of randomness is essential because a pure directed search will often find a solution that is plausible but wrong and be trapped by it, unable to get away and search for a totally different answer. Randomness is the part of the process which provides insight.

The behavior of a computer search algorithm with too low of a temperature appears much like the experience of a person who is stuck working on a difficult problem. It finds a plausible direction to try, makes a little progress, then discovers that the direction is wrong. It then keeps trying the same direction over and over looking for other ways to make it work by minor variations. It continues to cover the same failed ground over and over because it can't find a way to come up with the bold new direction that would actually solve the problem.

Suppose that people who experience the feeling of being unable to solve a problem because they are going over the same failed ideas over and over are actually in a state analogous to the state of the search algorithm. It would be plausible to suggest that adding randomness moght provide the clue needed to find a solution. What would be needed in this case is a method of randomly generating concepts and ideas. A Tarot deck is an example of such a method.

There are some problems with this model of what a Tarot deck does. First, Artificial Intelligence is far from being the most successful of sciences -- actual implementations of the search algorithms described above typically aren't able to solve the complicated real-world problems that their designers hope. Second, this use of the Tarot doesn't really match the descriptions that real Tarot users give of what they're doing. Although some Tarot users speak of the art as an enhancement of creativity (which is a plausible rephrasing of my model) many more describe the experience in terms of passively receiving answers from the cards, or some force that controls the cards. There are frequently rituals such as having the person receiving the reading handle the cards, or the notion that a deck can become spiritually damaged by its environment. (Interesting that one suggested cure for a damaged deck is to sort the cards into order and then shuffle them again. Is there a real effect here where some bad order of cards forms and is preserved by inadequate shuffling?) There is also some disagreement about what a Tarot reading is trying to accomplish. Some people claim that the cards are predictors of future events without any additional help, such that a good Tarot reader can predict the future of a person about whom they know nothing at all. Others seem to believe that the cards are a tool which is used as part of an extended exploration of a problem, with the person receiving the reading answering questions and providing interpretations of the cards. The second of these is much more in agreement with my model of what the cards might actually be accomplishing.

Supposing that we ignore the problems and propose a hypothesis: the Tarot (and other similar divination techniques) is a tool for allowing a skilled reader to more easily find good solutions for problems where the available data is ambiguous and difficult to interpret. This hypothesis has some interesting features. First, it says that using the Tarot is an acquired skill -- some people may be better at it than others. Second, Tarot reading only interprets existing information to make solving problems easier; it doesn't solve problems where there really isn't enough information available. An analogy I use for this hypothesis is that a Tarot deck is like an abacus. An abacus makes calculation easier for a skilled user, but does not provide answers that couldn't have been found another way.

The implications for experimentally testing whether there's anything to this are sort of frustrating. It's hard to design an experiment that evaluates this hypothesis. You can't choose subjects at random because the Tarot requires some skill to be useful. You can't test using simple clear-cut problems, because for simple problems the Tarot isn't necessary. Likewise, you can't test using problems where there is probably not enough information to solve the problem, because the Tarot doesn't invent missing information. Also, the experiment is measuring a matter of degree; Tarot readers aren't able to solve problems that are impossible to solve other ways, they are just on average better at finding the solutions.

Designing A Deck

Having come to these conclusions, I decided a few years ago to try some things out. One consequence of my thinking was that the most important property a Tarot deck can have is that it partitiions the space of concepts that the reader wants to think about in a way that they find useful. This meant that I decided to invent a deck for myself rather than learning a deck with the traditional cards. Another advantage of making my own is that even if I conclude that the Tarot is actually useless as a tool, the act of creating a deck is still a nice bit of artistic expression.

The deck I ended up with has 32 cards, and is in theory a part of a larger 64 card deck which I never developed fully. It is clearly the deck of someone who enjoys numbers. Instead of suits and numbers, each card has two numbers. The numbers from 0 to 7 are used, resulting in a full deck with 64 cards, but the deck I actually drew has 0 through 7 combined with 0, 4, 5, and 6 for only 32 cards. Each of the numbers has a symbolic meaning.

I started with zero instead of one, because I have spent so much time doing mathematics that zero has many important connotations for me. There is zero as absence (so the cards where a number is combined with zero represent a single number in its pure form). There is zero as the wild number which must be treated as a special case, which represents the notion of boundaries and the thing that has been forgotten but can change everything. There is zero as the partner of infinity.

Three and Four mark me as a physicist. Traditionally, Three has represented time: past, present, future. Four has represented place: North, South, East, West. I use them with the opposite meanings. Three represents place because space has three dimensions, and Four represents time because events happen is a four dimensional space-time. This is my deck, and I'm not tied to any particular desire to live in the past, so it embraces the modern. Card (5,6) is the computer.

The deck is not symmetric (for example (4,6) is Justice while (6,4) is Death) but there are some interesting connections between the cards in the square from (4,4) to (6,6). This is a deck of patterns.