The target article by Elqayam & Evans (E&E) reminds us of a plausible methodological norm: Do not confuse normative with descriptive theories! In many areas of psychology this distinction seems trivial. Models of memory or perception try to elucidate how these systems actually function, not how they should function. But what about theories of reasoning and acting, the classical domains of theoretical and practical rationality? Is it prudent to cleanly separate normative and descriptive theories here as well? My claim is that in psychological theories of rationality there is an intrinsically constitutive relation between normative and descriptive theories (Spohn Reference Spohn2002). I exemplify this claim in two different psychological research areas, one from the domain of theoretical, the other from practical rationality. I also argue that we need to resist the temptation to collapse descriptive theories with any specific normative theory.
Theories of theoretical rationality are concerned with how we achieve the goal of acquiring correct beliefs about the world. The close ties between normative and descriptive theories can easily be seen when we consider a scientist conducting an experiment. To explain the observed behavior, for example, randomization of subjects, it seems natural to invoke normative theories of how experiments should be conducted. The reason why normative theories work as descriptive explanations here is that we understand the scientist as a rational, goal-oriented person. Everyday reasoning is, in contrast, rarely guided by explicit methodological knowledge. Nevertheless, people's beliefs also have a normative force. People view themselves as motivated by reasons, and strive for optimality to achieve their goals. They distinguish between true and false beliefs; knowing they may err, they revise their beliefs and accept corrections. We would not say that a person believes that A causes B without simultaneously assuming that this person considers this proposition to be true. In sum, people conceive of themselves as rational agents, which make normative theories the natural candidate to explain their behavior.
Recent research about causal reasoning has obtained a wealth of evidence showing that we try to go beyond observations to obtain knowledge about causal relations in the world. People are rarely aware of how they acquire causal knowledge, but they understand what it means to respond to a causal query. Thus, a natural place to look for candidate explanations is normative theories of causal inference. In fact, virtually all currently competing theories of causal reasoning can be structured according to the preferred normative account motivating the theory (Waldmann & Hagmayer, in press). Causal reasoning cannot be modeled without some normative theory that tells us what causal judgments are.
Moral reasoning, which belongs to the domain of practical rationality, is another example of the constitutive relation between normative and descriptive theories. When we request moral judgments from subjects we do not want to learn about their preferences, wants, or inhibitions; rather, we want them to provide a normative evaluation of whether an act is right or wrong. Again, we conceive of subjects as rational agents who offer us responses to normative requests. Therefore, it is not surprising that an overview of research on moral judgment reveals that psychological theories use concepts from normative theories of morality (Waldmann et al., in press). For example, when interpreting responses to the famous trolley problem about whether it is permissible to sacrifice one person to save five, many theories focus on acts, outcomes, or values, concepts that have been highlighted as morally relevant in normative theories. And even if other non-moral factors are included in the explanations, we still interpret subjects' responses as driven by the motive to provide a normative assessment. Otherwise, we could not say anymore that we are studying moral judgments.
Although I claim that in theories of rationality, normative theories are constitutive for the theoretical description of the target phenomena, I agree with E&E that there is a danger of overstating the empirical adequacy of specific theories. Normative theories come in many variants, which compete. Their development is governed by factors such as coherence and consistency, which are less influential in everyday reasoning. Moreover, normative theories may restrict themselves to aspects of the target domain that are less relevant in everyday reasoning. Hence, it is unlikely that any specific normative position can be directly used as a descriptive theory. Responses in causal reasoning tasks have been interpreted as evidence for Bayesian causal network models, although it may often be possible to provide a more parsimonious account for individual phenomena by stripping away unnecessary untested assumptions implicit in these models. Similarly, the interpretation of responses as consequentialist or deontological in the trolley problem can often be more parsimoniously explained without attributing these global philosophical positions to subjects.
A sensible research strategy, therefore, seems to be to use one of the competing normative theories as a starting point, but to then ask whether all the assumptions inherent in these theories are empirically validated and necessary to explain the target behavior. Later, we can even go outside the realm of rationally relevant explanatory concepts. Rationality is certainly an idealization of thinking; many of our thoughts and actions are influenced by factors that, on reflection, nobody would consider as legitimate. For example, Eskine et al. (Reference Eskine, Kacinik and Prinz2011) found that moral judgments about transgressions (e.g., stealing; taking a bribe) tend to be harsher when subjects were drinking a shot of a bitter beverage than when they were given water or a sweet beverage. The taste of a beverage certainly does not constitute an acceptable argument for a moral judgment.
In short, it may be necessary to rebuild the selected normative framework, delete components, sacrifice coherence and consistency, and even add non-normative factors. Some core components will stay invariant, to guarantee that we still model the target competency, such as causal or moral reasoning. The end result of the revision process may be a caricature of any accepted normative theory, but its normative foundation will still be discernible. Or as the philosopher Otto Neurath famously claimed: “We are like sailors who have to rebuild their ship on the open sea, without ever being able to dismount it in dry-dock and reconstruct it from the best components” (see Quine Reference Quine1960, p. vii).
The target article by Elqayam & Evans (E&E) reminds us of a plausible methodological norm: Do not confuse normative with descriptive theories! In many areas of psychology this distinction seems trivial. Models of memory or perception try to elucidate how these systems actually function, not how they should function. But what about theories of reasoning and acting, the classical domains of theoretical and practical rationality? Is it prudent to cleanly separate normative and descriptive theories here as well? My claim is that in psychological theories of rationality there is an intrinsically constitutive relation between normative and descriptive theories (Spohn Reference Spohn2002). I exemplify this claim in two different psychological research areas, one from the domain of theoretical, the other from practical rationality. I also argue that we need to resist the temptation to collapse descriptive theories with any specific normative theory.
Theories of theoretical rationality are concerned with how we achieve the goal of acquiring correct beliefs about the world. The close ties between normative and descriptive theories can easily be seen when we consider a scientist conducting an experiment. To explain the observed behavior, for example, randomization of subjects, it seems natural to invoke normative theories of how experiments should be conducted. The reason why normative theories work as descriptive explanations here is that we understand the scientist as a rational, goal-oriented person. Everyday reasoning is, in contrast, rarely guided by explicit methodological knowledge. Nevertheless, people's beliefs also have a normative force. People view themselves as motivated by reasons, and strive for optimality to achieve their goals. They distinguish between true and false beliefs; knowing they may err, they revise their beliefs and accept corrections. We would not say that a person believes that A causes B without simultaneously assuming that this person considers this proposition to be true. In sum, people conceive of themselves as rational agents, which make normative theories the natural candidate to explain their behavior.
Recent research about causal reasoning has obtained a wealth of evidence showing that we try to go beyond observations to obtain knowledge about causal relations in the world. People are rarely aware of how they acquire causal knowledge, but they understand what it means to respond to a causal query. Thus, a natural place to look for candidate explanations is normative theories of causal inference. In fact, virtually all currently competing theories of causal reasoning can be structured according to the preferred normative account motivating the theory (Waldmann & Hagmayer, in press). Causal reasoning cannot be modeled without some normative theory that tells us what causal judgments are.
Moral reasoning, which belongs to the domain of practical rationality, is another example of the constitutive relation between normative and descriptive theories. When we request moral judgments from subjects we do not want to learn about their preferences, wants, or inhibitions; rather, we want them to provide a normative evaluation of whether an act is right or wrong. Again, we conceive of subjects as rational agents who offer us responses to normative requests. Therefore, it is not surprising that an overview of research on moral judgment reveals that psychological theories use concepts from normative theories of morality (Waldmann et al., in press). For example, when interpreting responses to the famous trolley problem about whether it is permissible to sacrifice one person to save five, many theories focus on acts, outcomes, or values, concepts that have been highlighted as morally relevant in normative theories. And even if other non-moral factors are included in the explanations, we still interpret subjects' responses as driven by the motive to provide a normative assessment. Otherwise, we could not say anymore that we are studying moral judgments.
Although I claim that in theories of rationality, normative theories are constitutive for the theoretical description of the target phenomena, I agree with E&E that there is a danger of overstating the empirical adequacy of specific theories. Normative theories come in many variants, which compete. Their development is governed by factors such as coherence and consistency, which are less influential in everyday reasoning. Moreover, normative theories may restrict themselves to aspects of the target domain that are less relevant in everyday reasoning. Hence, it is unlikely that any specific normative position can be directly used as a descriptive theory. Responses in causal reasoning tasks have been interpreted as evidence for Bayesian causal network models, although it may often be possible to provide a more parsimonious account for individual phenomena by stripping away unnecessary untested assumptions implicit in these models. Similarly, the interpretation of responses as consequentialist or deontological in the trolley problem can often be more parsimoniously explained without attributing these global philosophical positions to subjects.
A sensible research strategy, therefore, seems to be to use one of the competing normative theories as a starting point, but to then ask whether all the assumptions inherent in these theories are empirically validated and necessary to explain the target behavior. Later, we can even go outside the realm of rationally relevant explanatory concepts. Rationality is certainly an idealization of thinking; many of our thoughts and actions are influenced by factors that, on reflection, nobody would consider as legitimate. For example, Eskine et al. (Reference Eskine, Kacinik and Prinz2011) found that moral judgments about transgressions (e.g., stealing; taking a bribe) tend to be harsher when subjects were drinking a shot of a bitter beverage than when they were given water or a sweet beverage. The taste of a beverage certainly does not constitute an acceptable argument for a moral judgment.
In short, it may be necessary to rebuild the selected normative framework, delete components, sacrifice coherence and consistency, and even add non-normative factors. Some core components will stay invariant, to guarantee that we still model the target competency, such as causal or moral reasoning. The end result of the revision process may be a caricature of any accepted normative theory, but its normative foundation will still be discernible. Or as the philosopher Otto Neurath famously claimed: “We are like sailors who have to rebuild their ship on the open sea, without ever being able to dismount it in dry-dock and reconstruct it from the best components” (see Quine Reference Quine1960, p. vii).