This article is at an early stage of development.
Representativeness bias is actually two different cognitive biases. That is, it’s our cognitive bias towards concluding that:
- the quantity of one attribute of something is representative of the quantity of another attribute of that thing
- the more representative something is of a particular category, the more likely it will be a member of that category.
An example of the first representativeness bias is our bias towards concluding that the higher price of one bottle of wine over a bottle of another type implies that the first bottle contains higher quality wine. This bias occurs even when the price difference is quite small. That is, we’re biased towards assuming that the price of wine is representative of quality. Of course, although there’s a strong positive correlation between the price and quality of wine, we also know that it isn’t exact. The price of wine is also determined by the running costs, and pricing strategies, of the wine producers and wine companies, and the suppliers and retailers, which can vary between different types of wine. Therefore, a higher price doesn’t guarantee higher quality.
This first representativeness bias is actually a form of availability bias, and can therefore be explained by credulism, the certainty of belief, and the speed of the brain. Whereas the prices of bottles of wines are normally readily available to us, the running costs, and pricing strategies, of the wine producers and wine companies, and the suppliers and retailers, aren't. And even if all of the latter information was just as available to us, our processing of it would be much more complex, and therefore time-consuming, than simply noticing that one bottle is more expensive than another. And the ready availability of both the price difference between two bottles of wine, and our knowledge of the strong positive correlation between the price and quality of wine, mean that the possibility that the more expensive bottle contains higher quality wine will likely enter our thought processes, via our reasoning or imagination, well before any information about the origin of the prices, or the quality of the wines, has a chance to, especially given the speed of the brain. And, given credulism, this possibility enters our thought processes as a certainty, and, as explained in Confirmation bias - in the section Belief is self-preserving - the certainty of belief will then effectively provide a degree of protection to our belief, directly and via the confirmation bias. Specifically, while our certainty exists, we’re unlikely to question our judgement, and are therefore biased against thinking about the inexactness of the positive correlation between the price and quality of wine. Thus, we’re biased, due to availability bias, towards concluding that the quantity of one attribute of the bottles of wine - price - is representative of the quantity of another attribute - quality. And the logic of this explanation means that this representativeness bias will occur regardless of how small the price difference is, even though the smaller the difference the greater the probability that it could be due solely to factors other than quality.
Of course, our belief of a correlation between two attributes, which leads us to jump to conclusions, via availability bias, about the quantity of one attribute on the basis of the quantity of the other, may itself have been influenced by availability bias. That is, our belief of that correlation may be the result of us jumping to conclusions on the basis of quite a small number of co-occurrences of these attributes, or even just one such co-occurrence.
Regarding the second representativeness bias - our cognitive bias towards concluding that the more representative something is of a particular category, the more likely it will be a member of that category - consider the following classic experiment. Subjects were first given the following description:
- Linda is 31 years old, single, outspoken and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations.
They were then asked to rank the following statements about Linda by their probability, with the order that they were presented varying from subject to subject:
- Linda is a teacher in elementary school.
- Linda works in a bookstore and takes Yoga classes.
- Linda is active in the feminist movement.
- Linda is a psychiatric social worker.
- Linda is a member of the League of Women Voters.
- Linda is a bank teller.
- Linda is an insurance salesperson.
- Linda is a bank teller and is active in the feminist movement.
Most subjects thought that 3 was more likely than 6. However, most also thought that 8 - a combination of 3 and 6 - was more likely than 6, even though this is mathematically impossible. The category of bank tellers who are active in the feminist movement is a sub-category of the category of bank tellers. And if category X is a sub-category of category Y, then the probability of something being a member of X must be less, not greater, than the probability of it being a member of Y, given that the latter probability is the first probability plus the probability of the thing being a member of the rest of Y. Put another way, the probability of two particular possibilities both being true - as described in 8 - must be less, not greater, than the probability of one of them being true - as described in 6 - given that, in the first scenario, the uncertainty of one of the possibilities is combined with the uncertainty of the other. The subjects judged the probability of Linda being a member of categories 3, 6 and 8 on the basis of how representative she is of them. That is, based on her description, she’s more representative of active feminists than bank tellers, and therefore more representative of bank tellers who are active feminists than of bank tellers in general.
This representativeness bias is also actually a form of availability bias, and can therefore also be explained by credulism, the certainty of belief, and the speed of the brain. Something that’s wholly representative of a particular category is, by definition, certainly a member of that category. And the obviousness of this fact means that it’ll tend to be readily available to our thought processes as we're trying to judge the probability of something being a member of a particular category. And, given this fact, an obvious possibility is that the more representative the thing is of the category, the more likely it will be a member of it. And, given the speed of the brain, and the obviousness of this possibility, it tends to enters our thought processes, via our reasoning or imagination, within moments of us trying to form our judgement. And, given credulism, this possibility will enter our thought processes as reality, with the certainty of belief will then effectively providing a degree of protection to our belief, directly and via the confirmation bias. Specifically, while our certainty exists, we’re unlikely to question our conclusion, and are therefore biased against realising the mathematical flaw - as explained above - in it. Thus, we’re biased, due to availability bias, towards concluding that the more representative something is of a particular category, the more likely it will be a member of that category.
Another example of this representativeness bias is the ‘gambler’s fallacy’, which is the belief that there’s a negative correlation between the probability of a particular random outcome occurring in the future and the frequency of its occurrence in the past. For example, if someone is repeatedly flipping a fair coin, and they get four heads a row, then we’re biased towards concluding that the next flip is more likely to produce tails than heads. That is, we’re biased towards reasoning that four heads in a row is more likely to be followed by tails because such a sequence is more representative of the possible outcomes of five sequential coin flips than five heads in a row, given that there are many more possible sequences consisting of both heads and tails than only heads or tails. However, as with the previous example, our conclusion is mathematically impossible. Each flip of the coin is a completely independent event from the other flips, and so the outcome of the previous four flips is irrelevant to that of the fifth flip. Therefore, either possible outcome of the fifth flip is equally likely. Any specific sequence of heads and tails, from five coin flips, is actually equally likely, because each is one of the 32 possible sequences, and so each has a 1 in 32 chance of occurring. That is, although there are many more possible sequences consisting of both heads and tails than only heads or tails, the specific sequence HHHHT is just as unique, and therefore as likely, as HHHHH. In sum, we conclude that HHHHT is more random than HHHHH because it’s more disordered, and we believe disorder to be representative of randomness, but the strong correlation between randomness and disorder isn’t exact, and HHHHH is actually just as random as HHHHT.
- Tversky, A., and Kahneman, D., 1983, ‘Extensional versus intuitive reasoning: The conjunction fallacy in probability judgment’, ‘’Psychological Review’’, volume 90, number 4, pages 293–315.