In my research, I have tended to ask the questions „What makes randomness feel and look random?“ and „Can randomness be seen as not random?“ In this blog entry, I will explore the clustering illusion and the illusion of randomness.
The Clustering Illusion
When looking at clouds, it becomes clear how easily recognizable shapes such as people, animals, or objects appear. Called the „clustering illusion,“ this is the human tendency to perceive patterns in random data. Our evolutionary development has honed our ability to identify specific objects, including faces, potential threats in shadows, and familiar objects. Once the brain becomes adept at recognizing certain patterns, it tends to see them everywhere.1
The clustering illusion extends beyond earthly observations to illustrate the human tendency to see random events as more orderly or uniform than they really are. A pertinent example is the perception of stars in the night sky, where certain areas appear densely populated while others appear empty. This illusion arises from the tendency to assign a physical explanation to this non-random distribution, despite the fact that the positions of the stars are inherently random.
The left image below shows an example of a truly random star field, while the right image shows a manually generated star field that appears too uniform to be realistic, but is more in line with human expectations of randomness. In the example on the right, each star is randomly positioned within its own 20 × 20 pixel square, resulting in a more uniform but unrealistic result.2
Image source: https://tomroelandts.com/articles/the-clustering-illusion
The Illusion of Randomness
Richard A. Muller illustrates the illusion of randomness by generating two sets of star patterns. Using a computer, he first creates a plot with completely random locations, resulting in a seemingly uneven distribution of stars. To counteract this perception, Muller then divides the space into 100 smaller boxes and randomly places a star in each box, creating a more uniform appearance. Surprisingly, when asked which pattern looks more random, most people choose the second plot, which is actually the non-random one. The uniform distribution makes it look more random, challenging our intuitive understanding of randomness in seemingly clustered patterns. Muller emphasizes that truly random patterns can appear clustered, leading to the need to make them more uniform to convey a sense of randomness, as in many depictions of stars in artwork.3
Image source: https://muller.lbl.gov/teaching/Physics10/old physics 10/chapters (old)/4-Randomness.htm
What makes randomness feel and look random?
Understanding randomness can be challenging because of the limitations of our intuition. Asan example, consider the image below, with three sets of 132 points representing the nests of Patagonian seabirds, the nest sites of ant colony, and randomly generated coordinates.
Image source: https://behavioralscientist.org/yates-expect-unexpected-why-randomness-doesnt-feel-random-sense-patterns/
Identifying the truly uniform random distribution can be difficult, as our perception of randomness tends to lean toward well-spaced arrangements. Surprisingly, the leftmost image with uniformly distributed points is often perceived as less random than the other two images, which represent the locations of ant and seabird nests. This cognitive bias led to research into metrics that objectively determine spatial randomness, removing human perception from the equation. While our brains may struggle with randomness, recognizing it is crucial in various situations. This finding not only challenges our preconceived notions, but also underscores the importance of using objective metrics to reveal the true nature of randomness in spatial patterns.4
Sources
- https://medium.com/@nexy.io/clustering-illusion-1055023fec59 ↩︎
- https://tomroelandts.com/articles/the-clustering-illusion ↩︎
- https://muller.lbl.gov/teaching/Physics10/old physics 10/chapters (old)/4-Randomness.htm ↩︎
- https://behavioralscientist.org/yates-expect-unexpected-why-randomness-doesnt-feel-random-sense-patterns/ ↩︎
