The clustering illusion is a result of the human desire to see patterns in data or events even when they don't actually exist.
For instance, when studying research data it is normal to search for patterns. However, how relevant and accurate a seeming "pattern" really is is frequently a function of how large the original population sample size was. For instance, if you are looking for the prevalence of schizophrenia in a specific ethnic population it would be more reliable to look at a population sampling of a few thousand individuals than a population of two dozen. If you only sampled two dozen individuals and observed six individuals with schizophrenia you may conclude that a whopping 25% of the population has schizophrenia - this would be a clustering illusion. Whereas if you sampled thousands of individuals the true, typical 1% percentage of schizophrenia would, as is common to most human populations, most likely emerge. The larger population sample makes it easier to extrapolate accurate numbers than a small population sample.