Most events have general probabilities, or base rates. From the percentage of men who cheat on their wives, to the percentage of dogs with cancer, to the probability of it raining in your location, it is possible to know how likely these events are given no other information. When trying to predict specific probabilities – what is the chance my husband is cheating; what is the likelihood my dog has cancer; what is the likelihood it will rain today, we have two categories of information. One is the general probability, and the other is event-specific information – lipstick on his shirt, or a limp in the dog’s walk, or dark clouds. We tend to give more weight to the event-specific information than we should, and sometimes even ignore base rates entirely. An important causal bias is the representativeness heuristic, which states that when asked about likelihood, we instead answer the question of how much this event resembles other events. So when we see clouds, we don’t think about how rarely or often it rains. We only consider how well this situation resembles other situations in which it rained, and assign probabilities based on that.
One classic example involves a town with two cab companies, Green and Blue. Blue cabs make up 85% of the cab population. There was a hit-and-run accident involving a cab, and the witness believed that the cab was green. Witnesses are 80% accurate at discerning between blue and green cabs. What is the chance that the cab is green? Many people would say 80%, or something close to it, because they ignore the base rate. The chance it was green is actually 41%.