Thinking, Fast and Slow

System 2’s operations are, at least in many instances, dependent on the information that System 1 generates through its associative processes for System 2. This creates a likelihood of error regarding “merely statistical” information—information in which the rules of statistics control the outcome rather than the type of causation that System 1 excels at identifying (whether accurate or not).

Kahneman explains that we can predict statistical (or mathematical) facts, such as the percentage of random draws from a jar containing several types of marbles that will produce a specific array, but doing so lacks the sense of causation that System 1 relies on. The draws from the jar will yield predictable results only if they are carried on long enough to reach the statistical averages. Just a few draws will not yield the mathematically predictable result but will instead create a skewed impression because of insufficient sampling.

This sampling effect has routinely led to unreliable results in psychological experiments because the researchers have frequently relied on their System 1 intuition and, therefore, selected samples that were too small. This reveals the shortcomings of System 1’s intuition when dealing with statistical data. System 1 frequently produces results that are overconfident, generally through