Two Perspectives on Randomness

Two Perspectives on Randomness

To understand true randomness, we need to distinguish it from pseudorandomness.

A pseudorandom generator produces a sequence through a deterministic procedure that can nevertheless have useful statistical properties. In everyday speech, people also call a choice random when they simply do not notice the pattern behind it. These are different uses of the word.

A random outcome can be unpredictable on a particular trial while its probabilities are known. Physical sources of randomness also require a model and checks; calling an outcome random does not establish that it has no causes.

Our tendency to create a narrative to order and organize the world makes us predictable. We are also highly suggestible, remembering the most recent things we were exposed to. Thus, humans often behave in a pseudorandom manner, a fact that can be exploited.

Professional magicians exploit our availability bias and narrative tendencies in some of their tricks. If you’re asked to think of a random number or pick out a random card from a pack, you might not realize there is an order to your choices. Your choice is not truly random, merely pseudorandom.

Chananel Mifelew, better known by his stage name, Chan Canasta, became famous for his magic tricks in the 1950s and ’60s. His tricks were typically simple, performed using little more than a pack of cards or a book, but with a flair that earned him the nickname “a remarkable man.” Canasta’s tricks had an experimental air, as they tended to rely on taking chances. It was not unusual for them to go wrong during live performances.

Failure was part of Canasta’s willingness to take chances. Sometimes he took wild chances if he believed it was worth the risk. In one trick, Canasta asked a panel to each come up with a random word and then combine them into a sentence that he would predict in advance. Canasta was completely wrong, but he explained that it was worth trying. The chance of him getting it right was higher than you might imagine because of the way we misinterpret pseudorandomness as randomness. When someone picks a “random” word, the chances of them picking any particular word in the dictionary are not equal. In reality, certain words are much more likely to come up than others.

Unlike many modern magicians, Canasta didn’t pretend he was performing magic. Instead, he took advantage of his impressive memory and his ability to give the impression of randomness.[16] For instance, in one trick, he would ask a volunteer to pick three random cards from a pack and then place each in a different pocket, again at random. He was able to subtly influence which cards they took and where they placed them. But the volunteer felt it was random and was unaware of his influence. Even if it didn’t always work perfectly, it was impressive when it did.[17]

Canasta went on to inspire other mind readers, who took advantage of the same psychological trick—making people think their choices were random when they were pseudorandom.

When we’re asked to make a random choice, especially if we’re under pressure, we tend to fall into certain patterns.[18] Asked to name a random vegetable, most people say carrot. After George H. W. Bush declared his hatred of broccoli, that briefly became the more popular choice. Asked to name a shape, mentalist Banachek writes that most people will opt for a square. For a flower, they will usually say a rose. For a number between 1 and 5, most will name 3, and for 1 to 10, the usual choice is 7. The typical “random” color people name is blue, and the piece of furniture is a chair.[19] Performed with enough flair, simple tricks like this can seem like mind reading. It works because we don’t recognize we’re not making random choices.

Our inability to predict in Canasta’s type of magic tricks comes from seeing only one instance of the trick instead of seeing hundreds. The magician takes advantage of the seeming randomness of a single trial, while few people would think it was purely random if they watched it performed a thousand times with different audiences. That’s because it’s not truly random; it’s only pseudorandom.

Ask someone to produce a string of random numbers and patterns tend to creep in. Physical sources such as radioactive decay, atmospheric noise, and lava-lamp motion can provide unpredictable input.[20] Software can also expand suitable seed material with a cryptographically secure pseudorandom generator. The method should fit the task: a simulation, a game, and an encryption key have different requirements.

Divination rituals offer a historical example of using an external process to vary choices. The account below interprets a hunting ritual in this way. Its practical value would come from changing predictable habits, without requiring a supernatural explanation.

The Naskapi foragers in Labrador, Canada, needed a way to come up with hunting paths at random so the caribou wouldn’t learn to anticipate their routes and avoid them.[21] They achieved this through a ritual involving heating the shoulder bone of a caribou over hot coals until the surface was covered in cracks. Then they used the cracks as a map to tell them where to hunt. Beaver pelvises, skinned otters, and fish jaws also served similar purposes.[22] In this interpretation, the ritual helped vary hunting routes beyond the hunters’ familiar choices.

The practical point is that the hunters could use an unpredictable pattern to vary their routes. That does not require the cracks to be uncaused or prove that the routes were perfectly random.