Elqayam & Evans (E&E) present descriptivism as an alternative to normativism and as a way to avoid problematic inferences and research biases. Their proposal entails “identifying which terms are descriptive and which are deontic, and concentrating on the former” (see target article, Appendix: “Descriptivism”). They contend that “evaluative considerations need only be invoked […] where the object is to improve human thinking and performance” (sect. 8, para. 5, our emphasis). Moreover, evaluative considerations are fine so long as norms precede research rather than follow from it. In line with this, E&E prefer to entirely avoid inferences from descriptive to normative terms (“is-ought” inferences).
While recognizing that E&E principally address the cognitive sciences, broader applications of their perspective could have detrimental consequences. In this regard, we want to call attention to well-supported interactions between normative and descriptive theories: Descriptive theories can evaluate normative theories, including the evaluative terms. In such cases, the focus in psychology is on normative terms while the aim is not to improve human thinking. We give two examples from naturalistic ethics, where scientific methods are employed to evaluate normative theories. We show how these examples impact fallacies and biases, most importantly the “is-ought” inference. Although the practice of naturalizing the normative is not exclusive to ethics, ethics is nonetheless well-suited to discuss the nuances of the “is-ought” fallacy.
Consider the first example: If we presuppose a principle that links normative with descriptive terms, then we can use descriptive data to reject normative theories. Although this might be uncontroversial, it has had, and still has, far-reaching consequences for specific normative theories. For instance, in the moral sciences, “cannot” is often taken to imply “ought not.” Findings from experimental psychology, neuroscience, child development, and psychopathology suggest that it is most likely impossible to perform moral actions without being emotionally motivated (e.g., Prinz Reference Prinz2007). Therefore, philosophers reject theories that hold that good actions are not motivated by emotions (e.g., a purely Kantian morality). In a similar vein, neuropsychological findings about consciousness and decision making (among others), problematize the existence of free will. To the extent that normative moral concepts such as responsibility, blame, and guilt presuppose the existence of free will, normative theories might need improvement.
This interaction has consequences for fallacies and research biases. As E&E illustrate, psychologists use empirical findings to argue against normative theories – an “is-ought” inference – and interpret responses in terms of their normative correlates – an interpretation bias. If a psychologist's aim is to evaluate a descriptive theory, then these research practices are indeed fallacious and biased. However, evaluating normative theories opens up research questions that are interesting in their own right (i.e., above and beyond meliorism), and this goal may require interpreting responses in terms of normative correctness. Moreover, normative theories can then be rejected on the basis of empirical data. Importantly, all this requires that researchers state their commitment to a principle that links normative terms with descriptive terms.
Second, depending on one's epistemological or meta-ethical views, the meaning of a normative term may, at least partly, depend on descriptive data, such as how lay people use or understand the term. For instance, in a classical defense of moral relativism, Harman uses the following argument: “If we learn that a band of cannibals has captured and eaten the sole survivor of a shipwreck, we will speak of the primitive morality of the cannibals and may call them savages, but we will not say that they ought not to have eaten their captive” (Harman Reference Harman1975; our emphasis). Provided we know who “we” is in this quote, we have an empirically testable claim: Do we really not say that they ought not to have eaten their captive? Whether the meaning of “ought” is relative or not may depend on such descriptive facts, constituting another way in which descriptive theories can evaluate normative theories. Indeed, the persistent tendency in philosophy to refer to lay people's linguistic behavior is now a major impetus for experimental philosophy. Again, this practice provides a bridge from “is” to “ought.” Automatically rejecting such inferences, as E&E suggest, would then preclude naturalized and experimental philosophy.
Then again, one should not conclude that “is-ought” inferences are by default sound. When confronted with an “is-ought” inference, the investigator must sort out and reject wrong inferences. One should check if the authors beforehand explicitly stated a principle or philosophy linking normative with descriptive terms, as illustrated in our examples. Inadvertently inferring an “ought” from an “is” is wrong. However, if a reason is specified, the relevant question becomes whether one can agree with the principle or the philosophical rationale. For instance, 19th century evolutionary ethicists asserted that what is more evolved or developed is also morally better. This is, in the first place, wrong because evolution does not have a direction or purpose. In the second place, Moore (Reference Moore1903) interpreted these principles as attempts to state an analytic definition, entailing that the meaning of “good” (a normative term) was entirely cashed out in natural (descriptive) terms. Such attempts are also wrong: Philosophers and lay people use and understand normative terms as meaning that one cannot entirely cash them out in descriptive terms. On the other hand, we see no reason to reject an “is-ought” inference if one agrees with the stated reason, and therefore with the relevance of empirical data for normative theories.
Indeed, at least for naturalized philosophy, we relocate the burden of proof to descriptivism. Contrary to common wisdom, the classical arguments against “is-ought” inferences are not that far-reaching. Neither Moore nor Hume stated that all normative terms are entirely devoid of descriptive influence. Even Moore did not oppose an empirical research program that attempts to find what we ought to do. Hume (Reference Hume1739–1740) uses his famous “is-ought” passage to argue that it is not merely by reason or observation that we discover vice and virtue, but by sentiment. Precluding all “is-ought” inferences is a very common overreaction. In light of this, caution is in order when E&E's position is read with an eye toward other areas of psychological inquiry.
Elqayam & Evans (E&E) present descriptivism as an alternative to normativism and as a way to avoid problematic inferences and research biases. Their proposal entails “identifying which terms are descriptive and which are deontic, and concentrating on the former” (see target article, Appendix: “Descriptivism”). They contend that “evaluative considerations need only be invoked […] where the object is to improve human thinking and performance” (sect. 8, para. 5, our emphasis). Moreover, evaluative considerations are fine so long as norms precede research rather than follow from it. In line with this, E&E prefer to entirely avoid inferences from descriptive to normative terms (“is-ought” inferences).
While recognizing that E&E principally address the cognitive sciences, broader applications of their perspective could have detrimental consequences. In this regard, we want to call attention to well-supported interactions between normative and descriptive theories: Descriptive theories can evaluate normative theories, including the evaluative terms. In such cases, the focus in psychology is on normative terms while the aim is not to improve human thinking. We give two examples from naturalistic ethics, where scientific methods are employed to evaluate normative theories. We show how these examples impact fallacies and biases, most importantly the “is-ought” inference. Although the practice of naturalizing the normative is not exclusive to ethics, ethics is nonetheless well-suited to discuss the nuances of the “is-ought” fallacy.
Consider the first example: If we presuppose a principle that links normative with descriptive terms, then we can use descriptive data to reject normative theories. Although this might be uncontroversial, it has had, and still has, far-reaching consequences for specific normative theories. For instance, in the moral sciences, “cannot” is often taken to imply “ought not.” Findings from experimental psychology, neuroscience, child development, and psychopathology suggest that it is most likely impossible to perform moral actions without being emotionally motivated (e.g., Prinz Reference Prinz2007). Therefore, philosophers reject theories that hold that good actions are not motivated by emotions (e.g., a purely Kantian morality). In a similar vein, neuropsychological findings about consciousness and decision making (among others), problematize the existence of free will. To the extent that normative moral concepts such as responsibility, blame, and guilt presuppose the existence of free will, normative theories might need improvement.
This interaction has consequences for fallacies and research biases. As E&E illustrate, psychologists use empirical findings to argue against normative theories – an “is-ought” inference – and interpret responses in terms of their normative correlates – an interpretation bias. If a psychologist's aim is to evaluate a descriptive theory, then these research practices are indeed fallacious and biased. However, evaluating normative theories opens up research questions that are interesting in their own right (i.e., above and beyond meliorism), and this goal may require interpreting responses in terms of normative correctness. Moreover, normative theories can then be rejected on the basis of empirical data. Importantly, all this requires that researchers state their commitment to a principle that links normative terms with descriptive terms.
Second, depending on one's epistemological or meta-ethical views, the meaning of a normative term may, at least partly, depend on descriptive data, such as how lay people use or understand the term. For instance, in a classical defense of moral relativism, Harman uses the following argument: “If we learn that a band of cannibals has captured and eaten the sole survivor of a shipwreck, we will speak of the primitive morality of the cannibals and may call them savages, but we will not say that they ought not to have eaten their captive” (Harman Reference Harman1975; our emphasis). Provided we know who “we” is in this quote, we have an empirically testable claim: Do we really not say that they ought not to have eaten their captive? Whether the meaning of “ought” is relative or not may depend on such descriptive facts, constituting another way in which descriptive theories can evaluate normative theories. Indeed, the persistent tendency in philosophy to refer to lay people's linguistic behavior is now a major impetus for experimental philosophy. Again, this practice provides a bridge from “is” to “ought.” Automatically rejecting such inferences, as E&E suggest, would then preclude naturalized and experimental philosophy.
Then again, one should not conclude that “is-ought” inferences are by default sound. When confronted with an “is-ought” inference, the investigator must sort out and reject wrong inferences. One should check if the authors beforehand explicitly stated a principle or philosophy linking normative with descriptive terms, as illustrated in our examples. Inadvertently inferring an “ought” from an “is” is wrong. However, if a reason is specified, the relevant question becomes whether one can agree with the principle or the philosophical rationale. For instance, 19th century evolutionary ethicists asserted that what is more evolved or developed is also morally better. This is, in the first place, wrong because evolution does not have a direction or purpose. In the second place, Moore (Reference Moore1903) interpreted these principles as attempts to state an analytic definition, entailing that the meaning of “good” (a normative term) was entirely cashed out in natural (descriptive) terms. Such attempts are also wrong: Philosophers and lay people use and understand normative terms as meaning that one cannot entirely cash them out in descriptive terms. On the other hand, we see no reason to reject an “is-ought” inference if one agrees with the stated reason, and therefore with the relevance of empirical data for normative theories.
Indeed, at least for naturalized philosophy, we relocate the burden of proof to descriptivism. Contrary to common wisdom, the classical arguments against “is-ought” inferences are not that far-reaching. Neither Moore nor Hume stated that all normative terms are entirely devoid of descriptive influence. Even Moore did not oppose an empirical research program that attempts to find what we ought to do. Hume (Reference Hume1739–1740) uses his famous “is-ought” passage to argue that it is not merely by reason or observation that we discover vice and virtue, but by sentiment. Precluding all “is-ought” inferences is a very common overreaction. In light of this, caution is in order when E&E's position is read with an eye toward other areas of psychological inquiry.