Knobe presents two kinds of theories that compete with his own: motivational bias theories and conversational pragmatic theories. He presents his own theory as a competence account. While we agree with his criticisms of the other accounts, we think his taxonomy is incomplete. We would like to suggest a different form of competence account, one that does not take morality as such to play a crucial role in these effects. (In this regard, we agree with Phelan and Sarkissian [Reference Phelan and Sarkissian2008], Machery [Reference Machery2008], and indeed even Knobe and Mendlow [2004].) On our account, the effects of morality are a piece of a larger puzzle: Morality affects judgments of intentionality and related concepts only in virtue of its effects on expectations. Consequently, we think that anything affecting expectations will produce effects similar to those produced by moral norms. In fact, Knobe's own account points to a similar conclusion, although he doesn't acknowledge this.
According to Knobe's competence theory, people's moral norms influence their default expectations of others' intentions, beliefs, values, causal roles, and so on, and these default expectations in turn affect participants' judgments. Thus, he concludes, morality plays a deep role in explaining judgments in these various domains. But in this explanation, expectations are doing all the work; moral expectations have their effects only because they are expectations, not because they are moral. Thus, if Knobe's theory is right, we should find effects similar to the effects cited here in cases that have nothing to do with morality, but instead involve participants' non-moral expectations in parallel ways. And if this is right, it suggests that there is nothing specifically moral going on in the cases Knobe cites. These effects are, rather, effects of expectation, and expectations can be affected by both moral and non-moral factors (e.g., we expect people to have con-attitudes towards losing a game, although losing a game is not, normally at least, moral in any way).
Consequently, we think Knobe encounters a dilemma: Either his explanation of the effects he cites is correct, and then there is nothing especially moral at play here, but only an effect of expectations in general; or else his explanation of the effects is incorrect (in which case there may still be room for morality to play a distinctive role). Either way, Knobe finds himself in an awkward position; it doesn't seem that his explanation of morality's effects is compatible with the conclusion that moral considerations as such figure in our folk-psychological competence.
But we do not merely mean to present the dilemma. We take sides. We think Knobe's explanation is substantially correct, and that the effects Knobe finds would follow from any expectations participants hold firmly enough, whether or not those expectations have a moral character. To see whether this is indeed the case, it is not enough to look at cases that involve moral factors. Similar cases involving non-moral norms must be constructed and tested.
As a step in this direction, we have conducted some preliminary studies involving variations on the CEO cases that involve non-moral norms. These studies were conducted using participants on Amazon.com's Mechanical Turk website. We ran multiple studies attempting to measure possible non-moral effects on judgments of intention. In one such study, we used the following vignettesFootnote
1
:
Normal case:
Two people are playing chess.
One of them considers moving her queen to square A6. She thinks, “If I move my queen to square A6, I will capture my opponent's rook. But I don't care at all if I capture my opponent's rook; I think moving my queen to square A6 will allow me to checkmate in three moves.”
She moves her queen to square A6. Sure enough, she captures her opponent's rook.
Did she “intentionally” capture her opponent's rook?
Abnormal case:
The abnormal case was identical, except that the player allowed her own rook to be captured, instead of capturing her opponent's rook. We expected that participants reading the abnormal case would be more likely to judge that the side-effect of the player's move was intentional, when compared to participants reading the normal case. After all, it is normal to want to capture an opponent's rook, and normal to want one's own rook to remain uncaptured.
Although our results almost invariably trend in the expected direction, none actually reaches significance. (The closest result to significance arose from the vignettes given above; here, χ2(1, N=124)=3.03, p=.08.) For comparison, we also reproduced the original CEO cases using Mechanical Turk participants. Here, the results were highly significant: χ2(1, N=33)=14.73, p<.001. One possibility is that moral norms have a stronger effect on participants' expectations than do non-moral norms (or at least the non-moral norms we tested). Another possibility is that Knobe's explanation, which depends entirely on expectations, needs revision. Of course, either way, more systematic research is needed.
Our main point: One cannot only examine moral norms when judging whether Knobe's data show an effect of morality. We must look at non-moral norms as well, to find just how broad the phenomenon is. In fact, Knobe has, in the past, thought similar things. In Knobe and Mendlow (Reference Knobe and Mendlow2004), the authors propose that the kind of badness that affects intentional action judgments extends beyond just moral badness. They propose this in light of studies that seem to show similar effects involving clearly non-moral factors.
These theoretical possibilities matter for further work involving this effect. If indeed the effects Knobe finds are not specific to moral norms, then we must be careful not to interpret the effects as telling us about participants' moral norms. For example, Inbar et al. (Reference Inbar, Pizarro, Knobe and Bloom2009) use participants' judgments of intentionality as a way to measure implicit moral norms. This is risky; although judgments of intentionality might tell us something about participants' expectations in general, they cannot tell us which of those expectations are particularly moral and which are not. Use of intentionality judgments to measure implicit moral norms thus runs the risk of seeing moral norms where there are none.
You have
Access
Knobe presents two kinds of theories that compete with his own: motivational bias theories and conversational pragmatic theories. He presents his own theory as a competence account. While we agree with his criticisms of the other accounts, we think his taxonomy is incomplete. We would like to suggest a different form of competence account, one that does not take morality as such to play a crucial role in these effects. (In this regard, we agree with Phelan and Sarkissian [Reference Phelan and Sarkissian2008], Machery [Reference Machery2008], and indeed even Knobe and Mendlow [2004].) On our account, the effects of morality are a piece of a larger puzzle: Morality affects judgments of intentionality and related concepts only in virtue of its effects on expectations. Consequently, we think that anything affecting expectations will produce effects similar to those produced by moral norms. In fact, Knobe's own account points to a similar conclusion, although he doesn't acknowledge this.
According to Knobe's competence theory, people's moral norms influence their default expectations of others' intentions, beliefs, values, causal roles, and so on, and these default expectations in turn affect participants' judgments. Thus, he concludes, morality plays a deep role in explaining judgments in these various domains. But in this explanation, expectations are doing all the work; moral expectations have their effects only because they are expectations, not because they are moral. Thus, if Knobe's theory is right, we should find effects similar to the effects cited here in cases that have nothing to do with morality, but instead involve participants' non-moral expectations in parallel ways. And if this is right, it suggests that there is nothing specifically moral going on in the cases Knobe cites. These effects are, rather, effects of expectation, and expectations can be affected by both moral and non-moral factors (e.g., we expect people to have con-attitudes towards losing a game, although losing a game is not, normally at least, moral in any way).
Consequently, we think Knobe encounters a dilemma: Either his explanation of the effects he cites is correct, and then there is nothing especially moral at play here, but only an effect of expectations in general; or else his explanation of the effects is incorrect (in which case there may still be room for morality to play a distinctive role). Either way, Knobe finds himself in an awkward position; it doesn't seem that his explanation of morality's effects is compatible with the conclusion that moral considerations as such figure in our folk-psychological competence.
But we do not merely mean to present the dilemma. We take sides. We think Knobe's explanation is substantially correct, and that the effects Knobe finds would follow from any expectations participants hold firmly enough, whether or not those expectations have a moral character. To see whether this is indeed the case, it is not enough to look at cases that involve moral factors. Similar cases involving non-moral norms must be constructed and tested.
As a step in this direction, we have conducted some preliminary studies involving variations on the CEO cases that involve non-moral norms. These studies were conducted using participants on Amazon.com's Mechanical Turk website. We ran multiple studies attempting to measure possible non-moral effects on judgments of intention. In one such study, we used the following vignettesFootnote 1 :
Normal case:
Two people are playing chess.
One of them considers moving her queen to square A6. She thinks, “If I move my queen to square A6, I will capture my opponent's rook. But I don't care at all if I capture my opponent's rook; I think moving my queen to square A6 will allow me to checkmate in three moves.”
She moves her queen to square A6. Sure enough, she captures her opponent's rook.
Did she “intentionally” capture her opponent's rook?
Abnormal case:
The abnormal case was identical, except that the player allowed her own rook to be captured, instead of capturing her opponent's rook. We expected that participants reading the abnormal case would be more likely to judge that the side-effect of the player's move was intentional, when compared to participants reading the normal case. After all, it is normal to want to capture an opponent's rook, and normal to want one's own rook to remain uncaptured.
Although our results almost invariably trend in the expected direction, none actually reaches significance. (The closest result to significance arose from the vignettes given above; here, χ2(1, N=124)=3.03, p=.08.) For comparison, we also reproduced the original CEO cases using Mechanical Turk participants. Here, the results were highly significant: χ2(1, N=33)=14.73, p<.001. One possibility is that moral norms have a stronger effect on participants' expectations than do non-moral norms (or at least the non-moral norms we tested). Another possibility is that Knobe's explanation, which depends entirely on expectations, needs revision. Of course, either way, more systematic research is needed.
Our main point: One cannot only examine moral norms when judging whether Knobe's data show an effect of morality. We must look at non-moral norms as well, to find just how broad the phenomenon is. In fact, Knobe has, in the past, thought similar things. In Knobe and Mendlow (Reference Knobe and Mendlow2004), the authors propose that the kind of badness that affects intentional action judgments extends beyond just moral badness. They propose this in light of studies that seem to show similar effects involving clearly non-moral factors.
These theoretical possibilities matter for further work involving this effect. If indeed the effects Knobe finds are not specific to moral norms, then we must be careful not to interpret the effects as telling us about participants' moral norms. For example, Inbar et al. (Reference Inbar, Pizarro, Knobe and Bloom2009) use participants' judgments of intentionality as a way to measure implicit moral norms. This is risky; although judgments of intentionality might tell us something about participants' expectations in general, they cannot tell us which of those expectations are particularly moral and which are not. Use of intentionality judgments to measure implicit moral norms thus runs the risk of seeing moral norms where there are none.