Knobe has shown that people are far more likely to judge that an agent intentionally caused (or had the desire to cause) a negative side-effect than a positive one (e.g., harming vs. helping the environment). In his target article, he argues that such asymmetries are good evidence for a “moralist” (as opposed to a “scientific”) picture, according to which the naïve human capacity to ascribe psychological states to others for the purpose of describing, explaining, and predicting their actions presupposes the moral cognitive capacity to evaluate and judge others. He also offers an interesting semantics of psychological predicates such as intention, deciding, desiring. We think that neither Knobe's evidence nor his semantic analysis supports the moralist picture.
Knobe's semantics for psychological predicates can be seen as an extension of the semantics of gradable predicates such as cold. Following Pettit and Knobe (Reference Pettit and Knobe2009), suppose a beer and a coffee are both at the temperature of 20°C. Application of “cold” might plausibly yield a true statement in the coffee case and a false statement in the beer case. People rate each liquid relative to a default value that specifies what it is supposed to be like for it to be cold. In other words, the concepts respectively expressed by the words “coffee” and “beer” generate different standards of comparison for the application of the predicate “cold.” Similarly, the concepts expressed, respectively, by “harm” and “help” generate different comparison classes for the application of psychological predicates (e.g., “desire,” “intention”). The threshold generated by the concept harm is significantly lower than the threshold generated by the concept help. As a result, people are more likely to judge, for example, that the chairman had the desire to harm than the desire to help the environment.
Now, the fact that the semantics of gradable predicates can be extended to psychological predicates is not convincing evidence for the moralist picture of naïve psychology. Does the fact that the concepts harm and help generate different moral standards for the application of psychological predicates show that our understanding of psychological states itself is driven by moral evaluations? Consider the standards involved in the application of the quantifier many. Suppose that five children died in a fire and five children survived. The concepts expressed by “die” and “survive” generate different standards for the application of one and the same quantifier “many.” When asked, most people were inclined to accept that many children died, but to deny that many survived. But it would be odd to conclude, on this basis, that our mastery of numerical concepts expressed, for example, by the quantifier “many” (our numerical cognition) is shaped by moral evaluation.
In fact, Knobe's own semantics for psychological predicates is not consistent with the assumption that normative standards and moral evaluation directly shape our mastery of the relevant concepts of psychological states. On his account, a speaker's assumption about an agent's “pro-attitude” towards either a negative or a positive outcome will change the speaker's willingness to apply a psychological predicate (e.g., "intentional") to the agent's action. But if so, then the psychological concept of an agent's pro-attitude must be retrieved and used by the speaker before moral considerations can come into play.
We have started to address the empirical question whether moral evaluation shapes, not just our application of psychological predicates, but our very understanding of mental states themselves. Instead of testing the distinct conditions in which participants are willing to apply the verb “desire” for evaluating the chairman's action, we asked them to use their psychological concept in order to predict the chairman's decision. We designed such an experiment and ran the following study on 40 participants. After receiving either the HARM or HELP scenario, participants were then given the following text:
Imagine that, before the program is started, the VP comes back to the chairman and tells him: “It appears that we have to choose between three programs. All three will generate the same amount of benefits: hundreds of millions of dollars. The difference is that program A will have no impact on the environment, while program B will harm the environment and program C will help the environment. Anyway, it will be impossible to prove our responsibility in anything that would happen to the environment. So, if we harm the environment, no one will know of our responsibility. But, if we help the environment, that won't benefit our image. Starting program B or C will cost 10 dollars more than starting program A.” If the chairman had to make this choice, what program would he choose? A, B, or C?
In this case, if participants think that the chairman has the desire to harm the environment, then they should select answer B. If they think that he has the desire to help the environment, then they should select C. If they take the chairman to be indifferent to the environment, then they should select A. Now, if we assume that their inclination to apply the verb “desire” is a reliable guide to their prediction of the chairman's choice, then we should make the following prediction: Reading the HARM scenario should cause participants to select B more than reading the HELP case should cause them to select C.
Among the participants who received the HARM case, 90% answered A, 0% answered B, and 10% answered C. Among those who received the HELP case, 80% answered A, 10% answered B, and 10% answered C. Clearly, the participants' predictions show that they do not think that the chairman's desire to harm the environment, in the HARM case, is stronger than the chairman's desire to help the environment, in the HELP case. Arguably, the moral standards triggered, respectively, by the concepts expressed by “harm” and “help” generate different comparison classes for the application of the verb “desire,” which might enable people to convey to others their moral opinion of the chairman. But even so, these moral standards did not affect participants' use of the psychological concept desire in the prediction of the chairman's choice.
You have
Access
Knobe has shown that people are far more likely to judge that an agent intentionally caused (or had the desire to cause) a negative side-effect than a positive one (e.g., harming vs. helping the environment). In his target article, he argues that such asymmetries are good evidence for a “moralist” (as opposed to a “scientific”) picture, according to which the naïve human capacity to ascribe psychological states to others for the purpose of describing, explaining, and predicting their actions presupposes the moral cognitive capacity to evaluate and judge others. He also offers an interesting semantics of psychological predicates such as intention, deciding, desiring. We think that neither Knobe's evidence nor his semantic analysis supports the moralist picture.
Knobe's semantics for psychological predicates can be seen as an extension of the semantics of gradable predicates such as cold. Following Pettit and Knobe (Reference Pettit and Knobe2009), suppose a beer and a coffee are both at the temperature of 20°C. Application of “cold” might plausibly yield a true statement in the coffee case and a false statement in the beer case. People rate each liquid relative to a default value that specifies what it is supposed to be like for it to be cold. In other words, the concepts respectively expressed by the words “coffee” and “beer” generate different standards of comparison for the application of the predicate “cold.” Similarly, the concepts expressed, respectively, by “harm” and “help” generate different comparison classes for the application of psychological predicates (e.g., “desire,” “intention”). The threshold generated by the concept harm is significantly lower than the threshold generated by the concept help. As a result, people are more likely to judge, for example, that the chairman had the desire to harm than the desire to help the environment.
Now, the fact that the semantics of gradable predicates can be extended to psychological predicates is not convincing evidence for the moralist picture of naïve psychology. Does the fact that the concepts harm and help generate different moral standards for the application of psychological predicates show that our understanding of psychological states itself is driven by moral evaluations? Consider the standards involved in the application of the quantifier many. Suppose that five children died in a fire and five children survived. The concepts expressed by “die” and “survive” generate different standards for the application of one and the same quantifier “many.” When asked, most people were inclined to accept that many children died, but to deny that many survived. But it would be odd to conclude, on this basis, that our mastery of numerical concepts expressed, for example, by the quantifier “many” (our numerical cognition) is shaped by moral evaluation.
In fact, Knobe's own semantics for psychological predicates is not consistent with the assumption that normative standards and moral evaluation directly shape our mastery of the relevant concepts of psychological states. On his account, a speaker's assumption about an agent's “pro-attitude” towards either a negative or a positive outcome will change the speaker's willingness to apply a psychological predicate (e.g., "intentional") to the agent's action. But if so, then the psychological concept of an agent's pro-attitude must be retrieved and used by the speaker before moral considerations can come into play.
We have started to address the empirical question whether moral evaluation shapes, not just our application of psychological predicates, but our very understanding of mental states themselves. Instead of testing the distinct conditions in which participants are willing to apply the verb “desire” for evaluating the chairman's action, we asked them to use their psychological concept in order to predict the chairman's decision. We designed such an experiment and ran the following study on 40 participants. After receiving either the HARM or HELP scenario, participants were then given the following text:
Imagine that, before the program is started, the VP comes back to the chairman and tells him: “It appears that we have to choose between three programs. All three will generate the same amount of benefits: hundreds of millions of dollars. The difference is that program A will have no impact on the environment, while program B will harm the environment and program C will help the environment. Anyway, it will be impossible to prove our responsibility in anything that would happen to the environment. So, if we harm the environment, no one will know of our responsibility. But, if we help the environment, that won't benefit our image. Starting program B or C will cost 10 dollars more than starting program A.” If the chairman had to make this choice, what program would he choose? A, B, or C?
In this case, if participants think that the chairman has the desire to harm the environment, then they should select answer B. If they think that he has the desire to help the environment, then they should select C. If they take the chairman to be indifferent to the environment, then they should select A. Now, if we assume that their inclination to apply the verb “desire” is a reliable guide to their prediction of the chairman's choice, then we should make the following prediction: Reading the HARM scenario should cause participants to select B more than reading the HELP case should cause them to select C.
Among the participants who received the HARM case, 90% answered A, 0% answered B, and 10% answered C. Among those who received the HELP case, 80% answered A, 10% answered B, and 10% answered C. Clearly, the participants' predictions show that they do not think that the chairman's desire to harm the environment, in the HARM case, is stronger than the chairman's desire to help the environment, in the HELP case. Arguably, the moral standards triggered, respectively, by the concepts expressed by “harm” and “help” generate different comparison classes for the application of the verb “desire,” which might enable people to convey to others their moral opinion of the chairman. But even so, these moral standards did not affect participants' use of the psychological concept desire in the prediction of the chairman's choice.