Skip to 0 minutes and 7 seconds Lets explore the kinds of errors people make when they reason inductively. Consider the following example. My next door neighbour is a Professor at the local university. She likes to write poetry, is rather shy and reserved and is small in stature. Is she more likely to be a Professor of Oriental Studies or a Professor of Psychology? Take as long as you like…. And still most people give the wrong answer. No she is not more likely to be a professor of Oriental Studies. She is more likely to be a Professor of Psychology. It just is objectively true that there are considerably more Professors of Psychology than there are Professors of Oriental studies.
Skip to 0 minutes and 52 seconds Remember the question asked about the likelihood of this person being a professor of a particular type. So take any Professor at random and they are way more likely to be a Professor of Psychology than Oriental Studies. Of course not everyone answers incorrectly but a considerable number do and what this reflects is that people have certain biases about the world and these are reflected in their thinking. Even System 2 comes up with the wrong answer! Both thinking fast and thinking slow can give rise to errors in reasoning and once we accept this we need to try to make sense of the errors we make. Consider our next door neighbour.
Skip to 1 minute and 37 seconds In this case we might suggest that there is a stereotypical bias that is leading us to make an incorrect inference. The very brief personal details that were provided in the description - She likes to write poetry, is rather shy and reserved and is small in stature. Fit well with our preconceptions of a stereotypical Professor of Oriental Studies. Given this we are drawn to conclude that it is highly likely that she is a Professor of Oriental studies. More generally this kind of behaviour has been studied in terms of what is known as the representativeness bias – instead of making a judgment about how likely something is, people tend to make a judgment of how representative or typical something is.
Skip to 2 minutes and 25 seconds It seems that these biases are so influential that they can and do distort our thinking so that we end up drawing incorrect inferences. Cognitive psychologists have uncovered many more of these sorts of reasoning biases. Indeed remember the confirmation bias? And as a consequence we now have a much better understanding of why and when our reasoning fails.
Slow reasoning can result in errors too
By now you will have probably realised that the representativeness bias has also been held responsible for the conjunction fallacy – remember Shelia our bank teller?
According to some theorists, the reason why people are seduced into making the conjunction error is because the brief description of our bank teller accords with our stereotype of someone who quite likely would be active in the feminist movement. It is this stereotype that then guides our thinking so as to lead us to conclude that being a bank teller and being active in the feminist movement are most likely true. Shelia fits this stereotype, hence she surely is very likely to be active in the feminist movement.
Again we need to issue a ‘health warning’ because not everyone agrees with this. An alternative account is that people are not so much seduced by the representativeness bias but they are reasoning in a very particular way when they are asked about what is most probable. To address this, we can ask, ‘Do people really not understand how the world operates or is it something more peculiar about understanding what ‘probable’ means?’
Well, one way in which this has been addressed has been to ask about Shelia, and then ask a very similar question about Walter. When we ask about Shelia, we will stick to using the term ‘probable’ but when we ask about Walter we will ask that the participant considers ‘how many’ other people fit Walter’s description.
In broad terms, the evidence shows that people are far less likely to make the conjunction error when probed with the “how many” question than the corresponding “how probable” question.
Nonetheless, we can now go onto to test the reliability of this finding and in the next step we will do this via our next Class Exercise.
© University of York