What sentimentalists should say about emotion
Not every form of sentimentalism is plausible, and Josh May's (Reference May2018) book shows that there is reason to doubt some recent, prominent formulations. But it does not follow from this that we should be rationalists. Rather, I believe that May's criticisms help us see what a better sentimentalist metaethic should look like. More specifically, investigating what a sentimentalist should say about the nature of emotion reveals that emotions play a more significant role in moral cognition than May presumes. A sophisticated sentimentalism thus remains an important rival to rationalism.
Emotions for sentimentalists
As May sees it, sentimentalists face a dilemma. If emotions are just non-cognitive feelings, then they play no substantive role in moral cognition. By contrast, if emotions are partly cognitive (i.e., belief-like states), then the substantive work that they do in moral thought is best explained by their cognitive – not sentimentalist – features (pp. 51–52).
In response, sentimentalists should reject the picture of emotion that May's dilemma presupposes. At a gloss, emotions are intentional mental states with evaluative content. To be angry about a comment is to see that comment as an affront – as something that calls for a response; to feel compassion toward another is to see her as suffering – as someone to be helped. Pushing deeper, sentimentalists should follow emotion researchers in seeing emotions as states that involve multilevel content and processing (e.g., Griffiths Reference Griffiths, Evans and Cruse2004; Izard Reference Izard2007; Kurth Reference Kurth2018, Ch. 2; Levenson et al. Reference Levenson, Soto, Pole, Kitayama and Cohen2007).
At a low-level, emotions have course-grained, non-conceptual evaluative content that is intimately tied to feeling and action. So, for example, to feel angry is to experience the actions of another as challenge-to-standing-bad; feelings of shame convey something like social-rank-asymmetry-bad; compassion presents its target as another-suffering-bad. Here the hyphenated strings are gestures toward the distinctive, motivationally laden, evaluative dimensions of these emotions’ low-level, non-conceptual content.
At the high-level, an emotion's distinctive evaluative content is both fine-grained and conceptual in a manner that facilitates their use in reasoning. So, for example, anger toward a comment presents that comment as, roughly, an affront to one's (moral) standing. With shame, one sees oneself as having failed to live up to an ego-ideal. In both cases, the high-level conceptual content facilitates inferences about (respectively) being wronged and one's social-moral inferiority.
Importantly, a single emotional experience (e.g., a token of anger) will typically engage both types of content and both levels of processing (Griffiths Reference Griffiths, Evans and Cruse2004; Kurth Reference Kurth2018; Wringe Reference Wringe2015). Moreover, although the two channels of emotion content/processing generally preform complementary – though distinct – functions, they can come apart in ways that lend support to the above picture.
Consider, for example, experimental work on “repressors.” When these individuals are presented with a threatening stimuli, they display the attentional and physiological changes associated with fear – but they deny being afraid. What we appear to have, then, is a dissociation of low- and high-level emotion processing: while the low-level processing of repressors generates the action-oriented attentional shifts and physiological responses characteristic of fear, their high-level processing fails to categorize the situation under the relevant concept (fearsome or danger). Hence, they deny feeling the fear that they otherwise seem to be experiencing (Derakshan et al. Reference Derakshan, Eysenck and Myers2007; Kurth Reference Kurth2018, pp. 58–59).
Notice as well that emotions are not unique in being mental states with multilevel content/processing of this sort. Work in vision science, for instance, indicates that the content of visual perception is the upshot of two distinct channels: one (the ventral) that is involved in the perception of action and another (the dorsal) that is tied to memory and speech-processing. As with emotions, although visual perception typically combines these two sources of content as part of a unified visual experience, the two channels can be forced apart (Aglioti et al. Reference Aglioti, Goodale and Sousa1995; Wringe Reference Wringe2015).
In the present context, recognizing the multilevel structure of emotion is important because it opens up space for a distinctly sentimentalist thesis about the content and function of emotions. More specifically, with the above account in hand, sentimentalists can maintain that the low-level, motivationally laden, evaluative content of an emotion grounds the evaluative concept(s) distinctive of that emotion's high-level content. So, for example, shame's low-level content (i.e., social-rank-asymmetry-bad) fundamentally shapes and constrains both one's concept shameful and shame's associated high-level content (roughly, the evaluation that I have failed to live up to an ego-ideal). Similarly, compassion's low-level content (namely, another-suffering-bad) fundamentally shapes and constrains one's concept of compassion-worthy and compassion's associated high-level content (roughly, the evaluation that the target of one's compassion is enduring a serious and underserved misfortune that merits one's attention).
Crucially, the dependencies here are fundamental in a distinctly sentimentalist sense: Our understanding of the high-level evaluative concepts that are associated with emotions like shame and compassion comes by way of the motivationally laden, non-conceptual content carried by theses emotions’ low-level evaluations (D'Arms Reference D'Arms2005; Izard Reference Izard2007; Kauppinen Reference Kauppinen2013a). We find empirical support for this sentimentalist thesis in work on the evolutionary origins and development of emotion. For example, research in anthropology, psychology, and cognitive science provides evidence of high-level emotion content being shaped and constrained by low-level content for a range of emotions including shame (Fessler Reference Fessler, Tracy, Robins and Tangney2007), fear and anxiety (Kurth Reference Kurth2016; Reference Kurth2018; Öhman Reference Öhman, Lewis, Haviland-Jones and Barrett2008), and disgust (Tyber et al. Reference Tyber, Lieberman, Kurzban and DeScioli2013).
Moreover, the idea that low-level, non-conceptual content can ground high-level content is not unique to emotion. Consider color. The “unity relations” (that is, the phenomena of, e.g., reds looking more similar to oranges than greens) are thought to be non-conceptual features of color experience that shape and constrain both our color concepts and high-level, color content (e.g., red and green pick out “opposites” but red and orange do not) (Cohen Reference Cohen2003; Johnston Reference Johnston1992).
The payoff: A sophisticated sentimentalism
If emotions are states of the sort sketched above, then – contra May – sentimentalism can explain how emotions are able to both “carry morally relevant information” (p. 52) and “substantially influence moral judgment” (p. 28).
Taking these in turn, first notice that emotions are concerned with fundamental human values: compassion concerns the suffering of others, shame concerns the loss of social status, anger concerns challenges to one's standing. But notice, as well that the protection and promotion of these values is at the core of what we take morality to be. If that is right, then the above sentimentalist account of the content of emotions entails that they carry morally relevant information.
May might object that this connection between emotion and morality is too indirect – although emotions might highlight morally relevant information, they are not essential for making moral judgments (pp. 13–14). However, if the sentimentalist is right that emotions are essential to our understanding of evaluative content – grounding, for example, the distinctive badness of shameful, the special neediness of compassion-worthy – then May's objection is misplaced. Acquiring evaluative concepts is not something a “sophisticated robot” could do (p. 14). At best, a robot could approximate emotion's distinctive evaluative content by drawing on information provided by actual emoters (cf. Kauppinen Reference Kauppinen2013a).
Turn then to the question of emotions’ influence on moral judgment. The above account of emotions and their connection to moral/evaluative content, entails that emotions contribute to moral inferences insofar as they are essential sources of morally relevant content. Here too May is likely to protest that an influence of this sort is too thin to vindicate sentimentalism – though emotions “facilitate information processing,” they are not essential to moral inference in a deeper way (pp. 13, 71). But again notice that, on the above sentimentalist account, the low-level content of emotions is foundational for our understanding of the associated, high-level evaluative concepts that we use when making moral inferences. So, contra rationalists like May, moral inferences are “ultimately dependent on non-rational emotions” (p. 7).
Yet one might still worry that even if emotions are fundamental in this sense, the role that they play is still too paltry – after all, their distinctly sentimentalist-friendly low-level content only plays an indirect role in moral inference. In light of this, it is important to recognize that emotions’ low-level content also has a direct impact on moral decision making and inference.
For example, the low-level content of emotion can block the inferences and conclusions that one is brought to via explicit reasoning. Huck Finn's deliberations told him he ought to turn Jim over to the slave hunters. But the compassion he felt for his friend interfered, preventing him from endorsing the conclusion of his reasoning (Tappolet Reference Tappolet2016, p. 180). Additionally, emotion's low-level content can also lead us to question the moral judgments we have made: Martin Luther King, Jr., for example, spoke of the anxiety he felt about his conclusion that it would be wrong to protest the Vietnam War – in particular, he saw his anxiety as central to his realization that his decision not to protest was mistaken (Kurth Reference Kurth2018, Ch. 6).
In both of these cases, the low-level content of emotion not only provides morally relevant information that was not captured via deliberation, but also directly influences these individuals’ subsequent decisions and actions.
Most significantly, emotions can be immediate, non-inferential drivers of basic moral beliefs and judgments. To draw this out, first notice that May allows that we can come to have beliefs without engaging in any (explicit or implicit) reasoning. He thinks this happens when, for instance, you immediately (i.e., non-inferentially) come to the conclusion that the door opening before you retains its rectangular shape: such a judgment is not the result of reasoning, but rather the upshot of you “simply taking your visual experience at face value” (p. 9).
But now notice that moral judgments can be formed via emotions through the same kind of immediate, non-inferential process: I immediately come to believe that I have been insulted from the anger that I feel at your comment; your judgment that the invalid needs help springs immediately from the compassion you feel on seeing her crumpled on the sidewalk. Basic moral beliefs like these need not be the upshot of (implicit) reasoning. Rather – just like May's door example – they can result from simply taking your emotional experience at face value. Moreover, although this point has been made by sentimentalists who take emotions to be perceptions (e.g., Tappolet Reference Tappolet2016), the above account of emotion indicates that it holds for sentimentalism more generally.
In short, we have a range of examples showing not only that emotions carry morally relevant information, but also that they can play a significant role in moral judgment and inference.
Emotions are not mere consequences
At this point, May might object that the sentimentalism sketched here fits poorly with empirical findings suggesting that emotions are merely a consequence of (non-emotion-based) moral inferences and beliefs, not the drivers of them (pp. 38–41). In particular, May could extend the conclusions that he draws from experiments investigating the temporal order of subjects’ judgments about the disgustingness and moral wrongness of certain actions (Yang et al. Reference Yang, Yan, Luo, Li, Zhang, Tian and Zhang2013). This work suggests that disgust judgments follow moral judgments – a conclusion that fits poorly with standard sentimentalist proposals.
However, the relevance of these experiments is questionable. First, it is unclear how much we can draw from experiments focused on just one emotion (disgust). Moreover, research on other emotions (fear and anxiety) suggests that the temporal ordering of emotion and higher cognition is more in line with the sentimentalist account sketched here (e.g., Hofmann et al. Reference Hofmann, Ellard and Siegle2012, Kurth Reference Kurth2018, pp. 52–53). Most significantly, the task used in Yang et al.’s Go/No-Go experiments was complex: subjects were asked to make a decision about what button to push based on comparisons of their assessments of the disgustingness and moral wrongness of an action. But given that this was the task, the experiment does not appear to provide insight of the sort May needs (namely, evidence about the temporal order of feelings of disgust in comparison to moral judgments). Rather, it appears to focus on something else: how we make comparative assessments about (i) our judgments regarding the disgustingness of an action and (ii) our judgments of the moral wrongness of that action.
Stepping back, we can see how a richer understanding of what emotions are provides sentimentalists with new resources that help them vindicate a central role for emotion in moral cognition.
What sentimentalists should say about emotion
Not every form of sentimentalism is plausible, and Josh May's (Reference May2018) book shows that there is reason to doubt some recent, prominent formulations. But it does not follow from this that we should be rationalists. Rather, I believe that May's criticisms help us see what a better sentimentalist metaethic should look like. More specifically, investigating what a sentimentalist should say about the nature of emotion reveals that emotions play a more significant role in moral cognition than May presumes. A sophisticated sentimentalism thus remains an important rival to rationalism.
Emotions for sentimentalists
As May sees it, sentimentalists face a dilemma. If emotions are just non-cognitive feelings, then they play no substantive role in moral cognition. By contrast, if emotions are partly cognitive (i.e., belief-like states), then the substantive work that they do in moral thought is best explained by their cognitive – not sentimentalist – features (pp. 51–52).
In response, sentimentalists should reject the picture of emotion that May's dilemma presupposes. At a gloss, emotions are intentional mental states with evaluative content. To be angry about a comment is to see that comment as an affront – as something that calls for a response; to feel compassion toward another is to see her as suffering – as someone to be helped. Pushing deeper, sentimentalists should follow emotion researchers in seeing emotions as states that involve multilevel content and processing (e.g., Griffiths Reference Griffiths, Evans and Cruse2004; Izard Reference Izard2007; Kurth Reference Kurth2018, Ch. 2; Levenson et al. Reference Levenson, Soto, Pole, Kitayama and Cohen2007).
At a low-level, emotions have course-grained, non-conceptual evaluative content that is intimately tied to feeling and action. So, for example, to feel angry is to experience the actions of another as challenge-to-standing-bad; feelings of shame convey something like social-rank-asymmetry-bad; compassion presents its target as another-suffering-bad. Here the hyphenated strings are gestures toward the distinctive, motivationally laden, evaluative dimensions of these emotions’ low-level, non-conceptual content.
At the high-level, an emotion's distinctive evaluative content is both fine-grained and conceptual in a manner that facilitates their use in reasoning. So, for example, anger toward a comment presents that comment as, roughly, an affront to one's (moral) standing. With shame, one sees oneself as having failed to live up to an ego-ideal. In both cases, the high-level conceptual content facilitates inferences about (respectively) being wronged and one's social-moral inferiority.
Importantly, a single emotional experience (e.g., a token of anger) will typically engage both types of content and both levels of processing (Griffiths Reference Griffiths, Evans and Cruse2004; Kurth Reference Kurth2018; Wringe Reference Wringe2015). Moreover, although the two channels of emotion content/processing generally preform complementary – though distinct – functions, they can come apart in ways that lend support to the above picture.
Consider, for example, experimental work on “repressors.” When these individuals are presented with a threatening stimuli, they display the attentional and physiological changes associated with fear – but they deny being afraid. What we appear to have, then, is a dissociation of low- and high-level emotion processing: while the low-level processing of repressors generates the action-oriented attentional shifts and physiological responses characteristic of fear, their high-level processing fails to categorize the situation under the relevant concept (fearsome or danger). Hence, they deny feeling the fear that they otherwise seem to be experiencing (Derakshan et al. Reference Derakshan, Eysenck and Myers2007; Kurth Reference Kurth2018, pp. 58–59).
Notice as well that emotions are not unique in being mental states with multilevel content/processing of this sort. Work in vision science, for instance, indicates that the content of visual perception is the upshot of two distinct channels: one (the ventral) that is involved in the perception of action and another (the dorsal) that is tied to memory and speech-processing. As with emotions, although visual perception typically combines these two sources of content as part of a unified visual experience, the two channels can be forced apart (Aglioti et al. Reference Aglioti, Goodale and Sousa1995; Wringe Reference Wringe2015).
In the present context, recognizing the multilevel structure of emotion is important because it opens up space for a distinctly sentimentalist thesis about the content and function of emotions. More specifically, with the above account in hand, sentimentalists can maintain that the low-level, motivationally laden, evaluative content of an emotion grounds the evaluative concept(s) distinctive of that emotion's high-level content. So, for example, shame's low-level content (i.e., social-rank-asymmetry-bad) fundamentally shapes and constrains both one's concept shameful and shame's associated high-level content (roughly, the evaluation that I have failed to live up to an ego-ideal). Similarly, compassion's low-level content (namely, another-suffering-bad) fundamentally shapes and constrains one's concept of compassion-worthy and compassion's associated high-level content (roughly, the evaluation that the target of one's compassion is enduring a serious and underserved misfortune that merits one's attention).
Crucially, the dependencies here are fundamental in a distinctly sentimentalist sense: Our understanding of the high-level evaluative concepts that are associated with emotions like shame and compassion comes by way of the motivationally laden, non-conceptual content carried by theses emotions’ low-level evaluations (D'Arms Reference D'Arms2005; Izard Reference Izard2007; Kauppinen Reference Kauppinen2013a). We find empirical support for this sentimentalist thesis in work on the evolutionary origins and development of emotion. For example, research in anthropology, psychology, and cognitive science provides evidence of high-level emotion content being shaped and constrained by low-level content for a range of emotions including shame (Fessler Reference Fessler, Tracy, Robins and Tangney2007), fear and anxiety (Kurth Reference Kurth2016; Reference Kurth2018; Öhman Reference Öhman, Lewis, Haviland-Jones and Barrett2008), and disgust (Tyber et al. Reference Tyber, Lieberman, Kurzban and DeScioli2013).
Moreover, the idea that low-level, non-conceptual content can ground high-level content is not unique to emotion. Consider color. The “unity relations” (that is, the phenomena of, e.g., reds looking more similar to oranges than greens) are thought to be non-conceptual features of color experience that shape and constrain both our color concepts and high-level, color content (e.g., red and green pick out “opposites” but red and orange do not) (Cohen Reference Cohen2003; Johnston Reference Johnston1992).
The payoff: A sophisticated sentimentalism
If emotions are states of the sort sketched above, then – contra May – sentimentalism can explain how emotions are able to both “carry morally relevant information” (p. 52) and “substantially influence moral judgment” (p. 28).
Taking these in turn, first notice that emotions are concerned with fundamental human values: compassion concerns the suffering of others, shame concerns the loss of social status, anger concerns challenges to one's standing. But notice, as well that the protection and promotion of these values is at the core of what we take morality to be. If that is right, then the above sentimentalist account of the content of emotions entails that they carry morally relevant information.
May might object that this connection between emotion and morality is too indirect – although emotions might highlight morally relevant information, they are not essential for making moral judgments (pp. 13–14). However, if the sentimentalist is right that emotions are essential to our understanding of evaluative content – grounding, for example, the distinctive badness of shameful, the special neediness of compassion-worthy – then May's objection is misplaced. Acquiring evaluative concepts is not something a “sophisticated robot” could do (p. 14). At best, a robot could approximate emotion's distinctive evaluative content by drawing on information provided by actual emoters (cf. Kauppinen Reference Kauppinen2013a).
Turn then to the question of emotions’ influence on moral judgment. The above account of emotions and their connection to moral/evaluative content, entails that emotions contribute to moral inferences insofar as they are essential sources of morally relevant content. Here too May is likely to protest that an influence of this sort is too thin to vindicate sentimentalism – though emotions “facilitate information processing,” they are not essential to moral inference in a deeper way (pp. 13, 71). But again notice that, on the above sentimentalist account, the low-level content of emotions is foundational for our understanding of the associated, high-level evaluative concepts that we use when making moral inferences. So, contra rationalists like May, moral inferences are “ultimately dependent on non-rational emotions” (p. 7).
Yet one might still worry that even if emotions are fundamental in this sense, the role that they play is still too paltry – after all, their distinctly sentimentalist-friendly low-level content only plays an indirect role in moral inference. In light of this, it is important to recognize that emotions’ low-level content also has a direct impact on moral decision making and inference.
For example, the low-level content of emotion can block the inferences and conclusions that one is brought to via explicit reasoning. Huck Finn's deliberations told him he ought to turn Jim over to the slave hunters. But the compassion he felt for his friend interfered, preventing him from endorsing the conclusion of his reasoning (Tappolet Reference Tappolet2016, p. 180). Additionally, emotion's low-level content can also lead us to question the moral judgments we have made: Martin Luther King, Jr., for example, spoke of the anxiety he felt about his conclusion that it would be wrong to protest the Vietnam War – in particular, he saw his anxiety as central to his realization that his decision not to protest was mistaken (Kurth Reference Kurth2018, Ch. 6).
In both of these cases, the low-level content of emotion not only provides morally relevant information that was not captured via deliberation, but also directly influences these individuals’ subsequent decisions and actions.
Most significantly, emotions can be immediate, non-inferential drivers of basic moral beliefs and judgments. To draw this out, first notice that May allows that we can come to have beliefs without engaging in any (explicit or implicit) reasoning. He thinks this happens when, for instance, you immediately (i.e., non-inferentially) come to the conclusion that the door opening before you retains its rectangular shape: such a judgment is not the result of reasoning, but rather the upshot of you “simply taking your visual experience at face value” (p. 9).
But now notice that moral judgments can be formed via emotions through the same kind of immediate, non-inferential process: I immediately come to believe that I have been insulted from the anger that I feel at your comment; your judgment that the invalid needs help springs immediately from the compassion you feel on seeing her crumpled on the sidewalk. Basic moral beliefs like these need not be the upshot of (implicit) reasoning. Rather – just like May's door example – they can result from simply taking your emotional experience at face value. Moreover, although this point has been made by sentimentalists who take emotions to be perceptions (e.g., Tappolet Reference Tappolet2016), the above account of emotion indicates that it holds for sentimentalism more generally.
In short, we have a range of examples showing not only that emotions carry morally relevant information, but also that they can play a significant role in moral judgment and inference.
Emotions are not mere consequences
At this point, May might object that the sentimentalism sketched here fits poorly with empirical findings suggesting that emotions are merely a consequence of (non-emotion-based) moral inferences and beliefs, not the drivers of them (pp. 38–41). In particular, May could extend the conclusions that he draws from experiments investigating the temporal order of subjects’ judgments about the disgustingness and moral wrongness of certain actions (Yang et al. Reference Yang, Yan, Luo, Li, Zhang, Tian and Zhang2013). This work suggests that disgust judgments follow moral judgments – a conclusion that fits poorly with standard sentimentalist proposals.
However, the relevance of these experiments is questionable. First, it is unclear how much we can draw from experiments focused on just one emotion (disgust). Moreover, research on other emotions (fear and anxiety) suggests that the temporal ordering of emotion and higher cognition is more in line with the sentimentalist account sketched here (e.g., Hofmann et al. Reference Hofmann, Ellard and Siegle2012, Kurth Reference Kurth2018, pp. 52–53). Most significantly, the task used in Yang et al.’s Go/No-Go experiments was complex: subjects were asked to make a decision about what button to push based on comparisons of their assessments of the disgustingness and moral wrongness of an action. But given that this was the task, the experiment does not appear to provide insight of the sort May needs (namely, evidence about the temporal order of feelings of disgust in comparison to moral judgments). Rather, it appears to focus on something else: how we make comparative assessments about (i) our judgments regarding the disgustingness of an action and (ii) our judgments of the moral wrongness of that action.
Stepping back, we can see how a richer understanding of what emotions are provides sentimentalists with new resources that help them vindicate a central role for emotion in moral cognition.