Philosophy of Psychology. Lisa Bortolotti
with evidence instead of merely stamping their foot in a debate (Bortolotti 2014). But rationality as logicality, as we might call it, is widely accepted in philosophy, economics, and psychology. For instance, Phil Gerrans says that ‘a rational subject is one whose reasoning conforms to procedures, such as logical rules, or Bayesian decision theory, which produce inferentially consistent sets of propositions’ (Gerrans 2001, 161), and Richard Nisbett and Paul Thagard define rational behaviour as ‘what people should do given an optimal set of inferential rules’ (Thagard & Nisbett 1983, 251).
Having said that, not everybody agrees that the standard picture is the best understanding of human rationality. This controversy lies at the heart of the rationality wars between pessimists about human rationality (often appealing to the heuristics and biases programme) and optimists about human rationality (often associated with the ecological rationality programme). We will come back to this in Section 1.5.
There are some technical issues about the standard picture of rationality that we would like to mention briefly here.
First, the standard picture seems to presuppose that there is just one system of logic, one theory of probability, and one set of principles for decision-making. However, there are different formal systems of logic and different interpretations of probability; even the principles of decision-making can be disputed. Some rules of inference that are valid in standard logic (often called classical logic) are not valid in some non-classical logical systems. This raises a question: which system of logic should be adopted in evaluating the reasoning performance of agents? This is especially tricky if reasoning performance is consistent with one system of logic but not with another. Should we adopt the former and say that an agent’s performance is rational? Or should we adopt the latter and say that it is irrational? A similar issue arises when considering interpretations of probability. There are different interpretations of what probabilistic statements (e.g., there is a 80% chance that it will rain tomorrow) actually mean. There are also some probabilistic statements that make sense in some interpretations but not in others. This issue is relevant to the debate between pessimists and optimists. Gigerenzer, the most notable optimist, argues that some probabilistic questions in the heuristics and biases experiments are meaningless in light of his favourite interpretation of probability (which is known as the frequency interpretation). This issue will be discussed in Section 1.5.
Second, the standard picture assumes that our reasoning should be evaluated against the standards of logic, probability, and decision-making. This implies that, if our intuitive answer to a reasoning task is incompatible with a rule of logic, we should conclude that our intuition is at fault. But why can’t we say that it is logic, not intuition, that is at fault? In fact, the development of non-classical logic is sometimes at least partially motivated by some counter-intuitive features of classic logic.
This issue raises a further question: what should we do when facing an apparent discrepancy between logic and intuitive judgment? Should we trust logic and dismiss intuition as irrational? Or should we trust intuition and dismiss logic instead? Jonathan Cohen, another notable optimist, raises a similar issue. If the normative rules, against which our intuitive judgments are evaluated, are themselves evaluated on the basis of our intuitive judgments, then our intuition ‘sets its own standards’ (Cohen 1981, 317). But then how can our intuitive judgment be irrational? How can our intuitive judgment deviate from the standards that are set by itself? Cohen adopts a radical conclusion that human irrationality cannot be proven in principle no matter what psychology shows; ‘ordinary human reasoning – by which I mean the reasoning of adults who have not been systematically educated in any branch of logic or probability theory – cannot be held to be faultily programmed’ (Cohen 1981, 317). In making this claim, Cohen relies on the distinction between reasoning competence and reasoning performance (see Box 1B).
Third, the standard picture presupposes that what matters to rational and successful reasoning is the conformity to the rules of logic, probability, and decision-making. Not everybody agrees. For example, Keith Stanovich (1999) makes a distinction between rationality and intelligence, where rationality is a broader notion than in the standard picture, and intelligence is what the standard picture captures. In common discourse, intelligence and rationality are often conflated, or it is assumed that intelligence comprises rationality. But IQ tests do not measure the capacity for making good judgments and good choices that we commonly regard as a mark of rationality. If we take intelligence to stand for whatever IQ tests measure, then it does not tell us which behaviours are more likely to be conducive to the agent’s well-being in real life. Whereas IQ tests measure the capacity to process and manipulate information quickly and efficiently, they are not sensitive to whether the agent forms beliefs that are well supported by the evidence or whether she can critically evaluate the information she receives. To illustrate this distinction, Stanovich describes famous cases of smart people who acted foolishly, by which he means that people who have high intelligence in some domain made bad judgments and bad choices, thereby behaving irrationally. This does not mean that intelligence is not worth studying, just that there are other things that we value. Intelligence and rationality can be seen as having different domains of applications: Stanovich, for instance, suggests that intelligence maps the efficiency of cognitive functioning at an algorithmic level, whereas his more comprehensive notion of rationality tracks thinking dispositions at a higher level, governs decision-making, and takes into account the agent’s goals and values. Some notions of rationality like Stanovich’s are distinct from the standard picture, where rationality is associated with behaviour that conforms to the rules of logic, probability, and decision-making.
1.3 Systematic Biases and Errors
Here is a tentative answer to the philosophical question (with the qualifications we made at the end of the previous section): ‘rationality’ consists in reasoning in accordance with the rules of logic, probability, and decision-making.
Now, let us turn to the psychological questions. The crucial question is: ‘Do human agents actually reason in accordance with the rules of logic, probability, and decision-making?’ We will now review relevant studies, mainly from the heuristics and biases programme, which offer ample evidence that human reasoning systematically deviates from the rules of logic, probability, and decision-making. They include some of the most famous results in 20th-century psychological research.
Wason Selection Task
According to deductive logic, for a conditional statement of the form ‘If P then Q’ to be falsified, the antecedent (P) must be true and the consequent (Q) must be false. So, the statement ‘If you want to go to Brighton, then you need to catch the next train’ is false if you do want to go to Brighton but you don’t need to catch the next train.
The purpose of the selection task is to establish whether people can recognize when conditional statements are false. In the classic version of the task (Wason 1966), there is a deck of cards and each of them has a number on one side and a letter on the other. Participants can see four cards on the table, the first has a vowel (A) on the visible side, the second an odd number (7), the third a consonant (K), and the last an even number (4) (Figure 1). Participants have to say which cards they need to turn to test the following rule: ‘If a card has a vowel on one side, then it has an even number on the other side.’ Most participants in the classic version of the task said that the cards to be turned are the card with A on the visible side and the card with 4 on the visible side, or just the card with A on the visible side. However, the correct way to test the rule is to turn the card with A on the visible side and the card with 7 on the visible side, because a conditional statement is falsified when the antecedent (‘If a card has a vowel on one side’) is true and the consequent (‘then it has an even number on the other side’) is false. Only 5% of the participants solved the selection