The heuristics theory of emotions and moderate rationalism

ABSTRACT

This paper argues that emotions can play an epistemic role as justifiers of evaluative beliefs. It also presents the heuristics theory of emotion as an empirically informed explanation of how emotions can play such a role and why they in practice usefully complement non-affective evaluative judgments. As such, the heuristics theory represents a form of moderate rationalism: it acknowledges that emotions can be epistemically valuable, even privileged in some sense, but denies that they would be uniquely privileged. I argue that judgments and emotional responses pick out different but correlated kinds of evaluative properties and therefore emotional responses and non-affective evaluative judgments play mutually complementary rather than mutually exclusive roles. It follows that emotional responses can be legitimately drawn upon to support evaluative beliefs, but they lack supreme epistemic authority.

KEYWORDS:

• Emotions
• rationalism
• heuristics
• sentimentalism

I. Introduction

This paper investigates the role emotions play in providing us with justification for evaluative claims about the world. Traditionally, we find two contrasting approaches to the philosophical significance of emotions in the evaluative and normative domain: rationalism and sentimentalism. There are different ways of distinguishing these two approaches, not all of which have to do with the epistemic role of emotions. In this paper, however, I want to focus (almost) exclusively on this role, only briefly referring to the debate between sentimentalists and rationalists in other areas (e.g., concerning the metaphysics of value). Now, as regards the epistemic contribution of emotions to evaluations, substantial variations notwithstanding, the smallest common denominator among all sentimentalist views is acceptance of the claim that the epistemology of moral, aesthetic, prudential and other values is crucially dependent on emotions because emotions are more reliably correlated with evaluative properties than other kinds of responses we might have to these evaluative properties (Deonna & Teroni, 2012; Döring, 2010; Johnston, 2001; Prinz, 2011; Tappolet, 2000, etc.). By contrast, whatever their differences otherwise, rationalists will all deny the claim that there is a particularly robust connection between emotions and evaluative properties, and they will therefore also deny that epistemic access to values would inherently depend on emotional attitudes.

This paper argues that a moderate version of rationalism best captures the epistemic role of emotions in evaluations. Specifically, I aim to carve out moderate (epistemic¹) rationalism as a distinct and stable theoretical position in two ways: by making it empirically plausible, and by distinguishing it both from sentimentalism and from more strongly rationalistic approaches. Accordingly, first, the paper presents a specific, empirically informed theory – what I will refer to as the heuristics theory – of how emotions justify our evaluations. Heuristics are mental shortcuts or rules of thumb. The principal claim of the theory is that emotions are heuristics of value. That is, emotions are employed as mental shortcuts to detect value “out there”. Feelings of guilt indicate to us that we may have done something wrong. Feelings of fear indicate to us that we may be the targets of a threat of some kind. The general picture on this account is that emotions are one way of finding out about what is valuable in the world but by no means inherently superior to other ways.

I want to claim that its empirical plausibility is one of the advantages of this theory. This feature is especially important when it comes to a comparison with rival sentimentalist theories since one of the main strengths of the latter is said to be that (unlike rationalist theories, allegedly) they succeed in accommodating morality within a naturalistic view of the world (Kauppinen, 2014). I want to show, however, that a moderately rationalist epistemology of value may fare no worse than sentimentalism, and possibly better, in terms of consistency with empirical findings about human psychology and sociality. In my view, the heuristics theory neatly ties together much empirical data about the biology and psychology of emotions in experimental psychology, evolutionary anthropology, and neuroscience. Further, the theory forges a link between philosophical debates on emotions and the growing literature on heuristic thinking, and it can also be related to dual process views of cognition. Thanks to these features, the theory proposed here can even underwrite talk about the privileged epistemic function of emotions, but not in a sense amenable to more ambitious forms of sentimentalism.

I also want to demarcate the heuristics theory from alternative versions of rationalism. Some moderate rationalists accept that emotions can play a positive epistemic role, and so they do not share radical rationalists’ (e.g., the Stoics’) blanket mistrust of emotional influences on cognition and deliberation. At the same time, they also hold that emotions cannot play any epistemic justificatory role. They argue that, while emotional responses can motivate characteristics forms of action (e.g., direct attention), they can never provide reasons for evaluative beliefs. I will distinguish between this kind of “motivational moderate rationalism,” and my own preferred version of rationalism, which I will call “justificatory moderate rationalism”. I will try to show that motivational moderate rationalism fails to recognize the implications of the fact that emotional responses are appraisals. If emotions provide access to distinct kinds of evaluative properties, then they can surely provide some epistemic support for evaluative beliefs which concur with those appraisals. I will argue that, unlike its rationalist rivals, the heuristics theory can accommodate this insight.

II. The heuristics theory

In this chapter, I will first introduce the idea of heuristic cognition highlighting the importance of the mechanism of substitution in the relevant cognitive processes. I will then go on to develop the theory of emotions as heuristics of value. The last section discusses how emotions latch onto evaluative properties.

II.1 Heuristic cognition

Heuristics at their most general are cognitive shortcuts. Cognition is heuristic whenever “a difficult question is answered by substituting an answer to an easier one” (Kahneman & Frederick, 2002, p. 50). For example, it is hypothesized that people typically estimate the risk of flying, terrorist attacks, nuclear accidents, etc. by relying on the “availability heuristic”. That is, we tend to answer the more difficult question about statistical frequency by answering a much simpler question “how easily can I recall the information?”. The hypothesis predicts that (i) since such catastrophes are reported prominently and graphically in the media, and so (ii) they come very readily to mind, people will overestimate their frequency. This is indeed the case (Sunstein, 2005).

In general, what makes heuristics special from an epistemic point of view is this mechanism of substitution. The “target attribute” is substituted by a “heuristic attribute” which is easier to handle for the human mind (Kahneman & Frederick, 2002). Why is the manipulation of the heuristic attribute less difficult for the human mind? Because heuristic cognition is “spontaneous, intuitive, effortless, and fast”, whereas reflective thinking is “deliberate, rule-governed, effortful, and slow” (Kahneman & Frederick, 2002, p. 49). Since our cognitive capacities are limited and since we may lack time and opportunity to make full use of even these limited capacities, we are often forced to rely on heuristics. It can be predicted that the less time there is for reflection and circumspection, the more likely we are to turn to heuristics.

As the example of the “availability heuristic” has shown, heuristics can yield mistaken factual conclusions. Heuristic thinking can also be made responsible for erroneous or repugnant moral views as well as objectionable political and legal decisions and policies (Kahneman, 2011; Sunstein, 2003, 2005). For example, people often rely on an “outrage heuristic” when forming their opinions about the severity of punishment, i.e., penalties are assessed in terms of the outrageousness of the act, and not in terms of actual or intended harm (Sunstein, 2005, p. 538; see also, Kahneman & Frederick, 2002, p. 49, 63). Similarly, a “revulsion heuristic” appears to guide folk opinions regarding questions of sexual morality: “if I find it disgusting, then it must be wrong” (see, Haidt, 2001, p. 814; Sunstein, 2005, p. 540).

Having said that, heuristics frequently provide the right answer to both factual and evaluative questions. In fact, since there many situations in which time and opportunity are restricted and so we cannot make use of our, in any case limited, cognitive powers, heuristics-based cognitive processes can be expected to outperform reflective judgments in at least some cases. This prediction has been amply confirmed by numerous empirical studies (see, for example, Czerlinski et al., 1999; Gigerenzer & Brighton, 2009; Gigerenzer & Goldstein, 1996).² A related claim, which has found some empirical support, is that heuristic cognition is adaptive. This is the reason why heuristics will function well in response to cognitive tasks familiar from the formative period of human evolution and perform much less well outside such natural environments (see, Sauer, 2012, p. 802).

It is quite natural to draw a link between the idea of heuristic cognition and dual process theories (Damasio, 1994; Kahneman, 2011; Sauer, 2012; De Sousa, 2010). One popular version of generalized dual process theories³ proposes that when deliberating or making decisions human beings rely on two different systems which have different evolutionary histories and neural architectures. If so, then heuristic and reflective cognition can be said to be anchored in two distinct cognitive systems, whereby System 1 is spontaneous, intuitive, effortless, and fast, while System 2 is “deliberate, rule-governed, effortful, and slow” (Kahneman & Frederick, 2002, p. 49, see also, Damasio, 1994; De Sousa, 2010; Greene et al., 2001; Kahneman, 2011).

A “generalized dual process theory,” such as the two-systems dual process theory, can be a helpful conceptual framework to explain how heuristics-based thinking is possible and how it interfaces with other cognitive strategies. At the same time, the heuristics theory proposed here does not presuppose a commitment to dual process theories. It is important to note this because dual process theories have been contested. For example, Roeser (2009) argues that (i) dual process theories illicitly reduce the manifold ways of apprehending reality into two opposed systems, and that (ii) dual process theories imply that System 2 is somehow epistemically more reliable and morally more respectable, and therefore, it is normatively superior to System 1 – but, so Roeser, this is not the case.⁴

Whatever their relevance to dual process theories may be, these criticisms do not constitute a problem for the heuristic theory developed in this paper. First, the theory does not portray heuristic and reflective cognition as being opposed to one another. Rather, it describes heuristic processes and reflective cognition as complementary, typically producing convergent inputs into the deliberative process. The convergence will only break down in a minority of cases where heuristics tend to misfire for various reasons (see, Sauer, 2012, pp. 801–802).

Second, nothing said above would be invalidated if it turned out that, statistically speaking, heuristic cognition yielded better results overall than reflective judgments. In fact, we should grant all of the following points: (i) that reflective judgments are fallible, (ii) that heuristics are often indispensable, and finally, (iii) that heuristics can be drawn upon to correct reflective judgments or used when the agent is torn. I will say more about these issues below.

I now move on to discuss emotional heuristics. I will argue that these constitute a distinct subclass of heuristics. Several of the points just discussed carry over to this group of heuristics as well. However, affective heuristics are of special interest for the purposes of this paper as they are the key to understanding the role of emotions in the epistemology of value.

II.2 Heuristic emotions in evaluation

The principal claim of the heuristics theory is that emotions are heuristics of value. That is, emotions are employed as mental shortcuts to detect evaluative properties. Feelings of guilt indicate to us that we may have done wrong. Feelings of anger indicate to us that we may be the targets of aggression or hostility. Fear indicates to us the presence of a threat or hazard. Overall, emotional experiences constitute defeasible evidence based on which we may come to hold evaluative beliefs.

Emotions are special in terms of how they deliver information about value: emotional responses are typically fast, spontaneous, and often unreflective.⁵ The heuristics theory not only recognizes these facts about the role of emotions, but it can also provide a cogent explanation for them. Take guilt, for example. In terms of the heuristics-hypothesis, guilt is a heuristic which functions as follows: if I feel guilty about what I have done I will typically infer from this that I have done wrong. This will usually happen in a rapid and spontaneous fashion. My feeling of guilt will be an important heuristic for normatively evaluating my action. The main reason for relying on this heuristic is that drawing on the emotional response is easier than to reach judgments in a more systematic fashion (Sinnott-Armstrong et al., 2010; Szigeti, 2013).

Again, a link to dual process theories suggests itself. In general, it is quite common to assign emotions to System 1 (Epstein, 1994; but cf., Roeser, 2009). Consider so-called recalcitrant emotions. These persist even though one judges that the emotion is not justified as in Hume’s celebrated example is of someone afraid of falling despite knowing that she is safe (Hume, 1974; see also, D’Arms & Jacobson, 2003; Döring, 2010, 2015). It is tempting to describe this phenomenon as involving a conflict between the outputs of the two systems: hard-wired emotional reactions (e.g., fear of heights) clash with reflective judgments (e.g., flying is safe). The same could be true of some of our moral emotions such as “survival guilt”.

More generally, the heuristics model can well explain why our emotional responses become systematically less reliable in certain situations, including morally relevant situations. Emotions have evolved to deal with what Paul Ekman has called “fundamental life-tasks” (Ekman, 1992), i.e., challenges and threats facing human beings from their natural environments (Damasio, 1995, p. 24). As such, emotional dispositions were to a large extent fixed in our distant evolutionary past (Brady, 2013, pp. 20–21; Damasio, 1995; D’Arms & Jacobson, 2003, p. 138; Tooby & Cosmides, 1990). It is no surprise, therefore, that emotions can provide quick-and-ready assessments in familiar situations when there is no time or opportunity for cognitively more demanding forms of evaluation. As phylogenetically hard-wired response-mechanisms, however, emotions may become significantly less reliable in situations not encountered in the evolutionary past of human beings. Accordingly, we can expect the guilt heuristics, for example, to be geared primarily toward one’s own actions in relatively uncomplicated, familiar situations. These tend to involve a single action of visible, proximate impact and short duration whereby a single agent or a small group of agents causes harm to one or a few victims. This would explain why, when many people contribute to bringing about some collective harm, the participants will feel no guilt at all, or their guilt will be less intense (Szigeti, 2013).

However, responses made in Section II.1 to potential criticisms of the general idea of heuristic cognition apply to emotional heuristics as well. As noted there, the heuristics theory does not presuppose a rigid dichotomy of reason and emotion. So too in the evaluative and moral domains, affective and affectless cognitive processes typically complement each other (see, Craigie, 2011; Jones, 2003; Sauer, 2012). Furthermore, the characterization of emotions as heuristics of value does not imply that they would be relegated to a subservient role in our evaluations. Heuristic cognition frequently outperforms reflective thinking, so it can be justified to rely on heuristic emotions, even against the contradictory evaluative judgment. Huckleberry Finn refused to turn his friend Jim in feeling (correctly) that this would be wrong, even though he firmly judged (mistakenly) that he should have done so (Arpaly, 2000; Bennett, 1974; Döring, 2010, 2015). In general, heuristic emotions are often available to correct non-affective evaluative judgments, and they can help when the agent is deeply conflicted (see, Jones, 2003 for similar points about emotions, esp. pp. 185–7 and p. 197).

It has also been argued that at least some emotional responses display certain features of System 2-type thinking, and more generally features associated with reflective thinking (Roeser, 2009, 2010). Unlike other cognitive outputs of System 1 processes, emotional evaluations often seem to be, among others, slow, inferential, conscious, amenable to voluntary control, and cognitively penetrable.⁶ While this characterization of certain emotional responses is probably correct, this is not an embarrassment to the heuristic model. It may well be true that emotions are often elicited by cognitive appraisals whereby the cognitive appraisal itself can be of an evaluative (and not just factual) nature.⁷ These appraisals can be the result of slow and explicit reflection as well as fast and implicit cognitive processes (Gross, 1999; Gross & Feldman, 2011). For example, it is possible that I come to regard my action as morally objectionable after thinking more carefully about it and I feel guilt only as a result of such reflection.⁸ What the heuristics theory says here is that while in such cases the emotional response is indeed not only preceded but also causally explained by the prior judgment, the heuristic emotion can have a back-up function re-enforcing the evaluative judgment. In sum, the theory does not entail that emotional responses would necessarily precede evaluative judgments and can even allow that in some cases the evaluative judgments trigger the heuristic cognitive process.

Here is an example of how this could play out in practice. Goldie (2000, pp. 59–60) tells the following story of how he felt before and after slipping on ice:

Then I only thought of the ice as dangerous; now I feel fear towards the ice. Of course, it is true that I now do still think of the ice as in some way dangerous, but my way of thinking of it as dangerous is now distinct. Now I think of its dangerousness as emotionally relevant in a special way. Coming to think of it in this new way is not to be understood as consisting of thinking of it in the old way, plus some added-on phenomenal ingredient – feeling, perhaps; rather, the whole way of experiencing (…) the world is new (…). The difference between thinking of X as Y without feeling and thinking of X as Y with feeling will not just comprise a different attitude towards the same content – a thinking which earlier was without feeling and now is with feeling. The difference also lies in the content, although it might be that this difference cannot be captured in words.

Observe how in the story told by Goldie the feeling of fear aligns with and strengthens the prior judgment that the ice is dangerous, while also providing a sui generis input, a different perspective, which is not reducible to the reflective judgment. The heuristics theory is a detailed account of what this perspective consists in (which of course significantly diverges from Goldie in its specifics), i.e., how emotions latch onto value properties, and how this perspective relates to non-affective evaluative judgments. I will now proceed to a discussion of these issues.⁹

II.3 The structure of emotional heuristics

Recall the claim that what is epistemically distinctive about heuristic thinking is the underlying mechanism of substitution. The suggestion is that this modus operandi sets heuristics apart from reflective (or System 2) cognitive processes. This means that the heuristic theory is committed to a specific view about how the emotional experience is used as a heuristic input in the process of evaluation.

One basic idea it shares with other theories of emotion is that emotional responses pick out evaluative properties of their intentional objects.¹⁰ We can call these properties “sentimental evaluative properties”. For example, the evaluative property common to all the things we fear is the fearsome. The evaluative property common to all the things we are amused by is the funny (Shoemaker, 2017). The evaluative property common to all the things we are ashamed of is the shameful.¹¹

However, it seems that there can be evaluative properties that are not directly picked out by emotional responses. For example, on various grounds I may conclude calmly and without experiencing any emotion that a spider is dangerous or breaking a promise was wrong. So, we can distinguish between sentimental evaluative properties and non-sentimental evaluative properties. But how are these two kinds of properties related to one another?

The heuristic theory characterizes the relationship between sentimental evaluative properties and non-sentimental evaluative properties as follows. Emotions as heuristics of value latch onto sentimental properties. For example, when one is fearful of a spider one becomes aware of the spider as fearsome. Now, the awareness that the object of the emotion instantiates the sentimental property feeds into the evaluative process because the emotional experience as a heuristic of value generates prima facie reasons for the subject in favor of evaluating the object of the experience in some way. At the same time, the subject can make non-affective evaluative appraisals of the same object as well. These latch on to the non-sentimental evaluative properties of the same object. For example, one may ask friends, consult experts about the spider, or read up on the subject.¹²

When the relevant object is the target of both an evaluative judgment and a heuristic emotion, these forms of assessment can yield convergent results: the spider may be experienced as fearsome as well as judged to be dangerous. Or it may not: I may be fearful of the spider, although people tell me it is not dangerous. In short, the outputs of affective heuristic processes, on the one hand, and those of non-affective, more or less reflective cognitive processes, on the other, can clash.

II.4 The heuristics theory and sentimentalism

This paper is not aimed at directly refuting specific claims made by sentimentalists. However, I do mean to put pressure on sentimentalism indirectly by attempting to show that the heuristic theory can accommodate and explain the characteristic and familiar role of emotions in evaluations within a broadly rationalist framework and without subscribing to certain key sentimentalist tenets. In particular, the heuristics theory denies that emotional responses are more reliably connected to evaluative properties than other kinds of non-affective responses, and so the sentimentalist claim is also rejected that epistemic access to values would inherently depend on emotional attitudes or that these emotional attitudes would command special epistemic authority.¹³

That said, some features of the above analysis are likely to be greeted by sentimentalists. In particular, as we have seen, the heuristics theory does not in any way imply that the epistemic outputs of non-affective, more or less reflective cognitive processes would be by default epistemically superior to emotion-based evaluative cognition. Further, the theory not only recognizes that we can experience values through emotions, but it can also provide an empirically informed explanation of salient characteristics of this experience. It can explain, that is, why emotional evaluations tend to be fast, spontaneous, and often unreflective, and why they (occasionally) persist in the face of contrary judgments. Overall, the heuristic theory is happy to take on board the sentimentalist’s claims about the special role and significance of emotions as an epistemic resource. What is more, the heuristic theory can link these claims to theories about dual process cognition and heuristic thinking in cognitive science.

It is also worth noting that the approach proposed here seems perfectly consistent with a recognition of the unique motivational role of emotions. The motivational theory of emotions, which influential sentimentalists have described as “the best account of the natural emotions” (D’Arms & Jacobson, 2017, p. 266), characterizes emotional motivation as typically impulsive, linked to specific forms of action, geared toward focusing attention and prioritizing certain goals. In short, emotional states – at least in the short run – tend to have an outsize influence on how an agent will (be disposed to) act. Tapping into empirical research on dual process cognition and heuristic thinking, the heuristic theory, I submit, is particularly well placed to explain why this should be so.

In sum, the heuristic theory seems to be able to capture many of the features of emotional experiences which, as sentimentalists rightly complain, rationalist theories have tended to ignore, or at least have had considerable difficulties with. At the same time, there will be several points of disagreement between sentimentalism and the heuristic theory. For one thing, the heuristic theory is more pronounced and more systematic in its approach than its sentimentalist rivals about the inherent limitations of emotional experiences. If it is true that emotions are affective heuristics of value, then in terms of their epistemic role they will be subject to the same constraints as heuristics in general. As noted, heuristic thinking will become systematically less reliable in many situations, which is ultimately to be explained by the evolutionary origins of heuristics as an adaptation. This is also true of emotional heuristics. These too tend to perform less well in situations which they were not, to steal a phrase from Prinz (2004, p. 54), “set up to be set off by”.

So, the heuristics theory has a leg up on sentimentalists by offering a systematic and empirically informed explanation of when to trust and when not to trust our emotional responses. Nevertheless, it may be argued that this difference does not in itself decisively distinguish the heuristic theory from sentimentalist rivals as at least some sentimentalists insist that the main claim of sentimentalism is neither that emotions “always get it right,” nor even that emotional responses are statistically more likely to be correct (D’Arms & Jacobson, 2010, p. 611). However, there is a further issue that does mark a point of fundamental disagreement between the heuristic theory and sentimentalism. This concerns the status of sentimental evaluative properties – such as the property of “fearsome” or “guilt-worthy” – tracked by emotions. As noted, at the heart of the theory of heuristic cognition is the idea of property substitution. Heuristics, and so affective heuristics too, track a heuristic property. This heuristic property (e.g., the “fearsome”) is substituted for the non-heuristic property (e.g., the “dangerous”) because it is easier to do so for the human mind. It follows from this picture of how affective heuristics work that emotional experiences provide only defeasible or prima facie evidence that some non-heuristic property is indeed instantiated. Affective heuristics can provide such evidence because in many cases when the heuristic property is instantiated, then so is the non-heuristic target property (e.g., what we perceive as “fearsome” often is dangerous). After all, heuristic cognition is remarkably successful in many cases. However, on the heuristic theory appealing to the emotional experience will not be decisive because further evidence (e.g., testimony), which can be accessed by non-affective evaluative cognition, will be required to determine whether the non-heuristic target property is in fact present or not. As noted, it is this claim that puts the heuristic theory in the camp of rationalist theories, and it is this claim sentimentalists will crucially disagree with.

III. Varieties of rationalism

The account of the epistemic role of emotions outlined above contrasts not only with sentimentalism, but also with versions of rationalism according to which emotions cannot play any epistemic justificatory role. If that were true, then of course emotions could not serve as heuristics for our evaluative beliefs either. The claim that an emotional response never provides reasons for belief will be most familiar from what we may call extreme or “Stoic” rationalism which portrays emotions as nothing but a source of bias and prejudice.¹⁴ However, interestingly, some moderate rationalists also refuse to assign any epistemic justificatory role to emotional experiences, even though they do not share the extreme rationalists’ blanket suspicion toward emotional influence on deliberation. In the present section, I will compare and contrast this variety of moderate rationalism with the one defended in the paper. I will call the former “motivational moderate rationalism” and my own version “justificatory moderate rationalism”.

III.1 Emotions: Epistemic motivators or justifiers?

Michael Brady’s account (Brady, 2013) is a good example of motivational moderate rationalism. On his view, emotional responses motivate the search for justificatory reasons by focusing our attention:

Consider one’s experience of fear, upon hearing a noise downstairs as one is trying to get to sleep … In this situation, we are typically motivated to seek out and discover additional reasons or evidence. In particular, we are motivated to seek out and discover considerations that have a bearing on whether our initial emotional “take” on the situation, namely that we are in danger, is accurate. We strain our ears to hear other anomalous noises, or rack our brains trying to think of possible non-threatening causes for the noise. (Brady, 2013, pp. 86–87)

On this account, emotions focus and capture our attention, prompting us to evaluate situations. The epistemic value of our emotional reactions does not lie in their providing support for our evaluative beliefs or in making these beliefs more credible.

Since it attributes some epistemic value to emotions, motivational moderate rationalism is less severe than the “Stoic” rationalism because for the latter the epistemic role of emotions is always pernicious. Given their agreement that emotions can play a positive epistemic role, the heuristics theory and Brady’s account have been grouped and discussed together as varieties of “shrewd rationalism” (Milona, 2016). There is, however, a crucial difference between the rationalism of the heuristics theory and Brady’s approach. While motivational moderate rationalism, as we have seen, attributes epistemic value to emotions only as epistemic motivators, the heuristics theory recognizes that emotions can also be sources of justifying reasons.

The main objection to moderate motivational rationalism is that the refusal to recognize emotions as epistemic justifiers is inconsistent with the view of emotions as appraisals. Note that Brady himself talks about emotional responses as appraisals. When I feel fear upon hearing noise downstairs, I am motivated to check what is going on precisely because thanks to my fear I appraise the situation as potentially threatening. In general, as Brady rightly says, emotions inform us about value. This makes sense as it would be hard to explain why emotions are subject to normative demands if one did not accept that they provide access to evaluative properties. More specifically, emotions are subject to normative demands of (i) intelligibility because we think there are good reasons to have them (or not to have them as the case may be), and (ii) fittingness because they can fail to be correct or incorrect. So, for example, there is reason for me to feel fear upon hearing the noise downstairs because noise downstairs, when everyone at home is supposed to be asleep, may be caused by an intruder. And my fear (while justified) may not be fitting because the noise may have been caused by the contraction of wooden furniture. Furthermore, if emotions were not directed at evaluative properties in this way, it would also be much harder to explain why fear, rather than anger, is the appropriate response to the noise downstairs.

However, if emotions are indeed appraisals, then it seems to follow that these appraisals can also provide some epistemic support for evaluative beliefs. This is particularly clear in cases in which the emotional appraisal is consonant with non-affective reasons for having some evaluative belief. Imagine that I catch sight of a dog and get scared of it. At the same time, I also recall that somebody has already warned me about that dog. It seems perfectly natural to treat my fear reaction as well as the warning as both providing prima facie reasons for my belief that it is best to give that beast a wide berth. In other words, because emotions evaluate their intentional objects, they can generate prima facie reasons for the subject in favor of evaluating the object of the experience in some way. In fact, things are no different when the emotional appraisal conflicts with non-affective considerations. I may be afraid of the dog, but then remember having been told that it was perfectly docile. When deliberating how to evaluate the situation (e.g., is there really a threat?) and what to do (e.g., run away or approach the dog?), both my emotional experience and testimonial information will be used as legitimate epistemic input generating prima facie reasons for belief and action.¹⁵

In short, the objection to Brady’s version of moderate rationalism is that it is impaled on the horns of a dilemma. If one denies that emotions are appraisals, then it remains unclear why emotional responses should generate reasons to “seek out and discover considerations” beyond “our initial emotional take on the situation”. If, on the other hand, one accepts that emotions are appraisals, then it becomes hard to deny that these appraisals can also constitute prima facie justificatory reasons.

III.2 The objection from double-counting

To clarify, what is claimed by those who think that emotions can play a justificatory and not just a motivating epistemic role is not that my experience of the emotion qua experience provides a justificatory reason, but rather that my experience qua an experience of an evaluative property provides a justificatory reason. For example, my fear does not provide a reason for the belief that the situation is threatening qua fear, but because my fear experience is the experience of something as fearsome.¹⁶ In short, emotions can provide justificatory reasons, which reasons are based on the intentional content of the emotional experience.

Since my preferred justificatory moderate rationalism grants the last point, it might be seen as vulnerable to an important objection by motivational moderate rationalists. The objection is based on the thought that taking an emotional reaction to be a reason for the evaluative belief would amount to double-counting.¹⁷ Thus, Brady (2013, see esp. pp. 130–1) argues that when we reply to some feature of a situation, then those features provide both the reason for the emotional response (e.g., fear) and the reason for the evaluative judgment (e.g., “the situation is dangerous”). Consequently, if we took the emotional response to these features as an additional reason for the evaluative judgment about those features, then we would be counting the same reason twice: “once in their own right, when we become aware of them and form the evaluative judgment on this basis, and once again via our taking the emotional response to such features as a further reason for the evaluative judgment.” That we do not in fact do so, Brady adds, is evidenced by the fact that the justificatory force of emotional experience can disappear or become otiose once the evaluative judgment has been formed. Once I judge the dog to be dangerous, my fear of it provides no additional reason for that judgment.

The double-counting objection is based on the assumption that the same evaluative property gives reasons for the emotional response and the evaluative judgment.¹⁸ I want to argue that the double-counting objection is misguided because this assumption is false, and that this is made clear by the heuristics theory (among others¹⁹). Therefore, the objection poses no threat to the claim that emotions can play a justificatory role, and more specifically, it does not undermine what the heuristic theory says about how emotional experiences can justify evaluations.

Consider (say) fear. On the view I defend here fear tracks the (heuristic) property of being “fearsome” – a sentimental evaluative property, whereas the corresponding evaluative judgment tracks the property of being dangerous – a non-sentimental evaluative property (see, Section II.3). Now, if the emotional response and the evaluative judgment do indeed track different properties (sentimental and non-sentimental evaluative properties, respectively) there is no reason why the output of the emotional heuristic could not figure as input into the evaluation complementing the non-affective, evaluative judgment. On the heuristic theory, emotional responses do in fact play exactly this epistemic role. The situation is comparable to how a testimony that a is F and my judging that a is F (or my perceiving that a is F) can both constitute reasons for my believing that a is indeed F.²⁰

Brady (2013, p. 113): “So reasons to think that something is dangerous are equally reasons to fear that thing … the fact that the bull in the farmer’s field has sharp horns and is advancing rapidly towards me is a good reason for me to be afraid and a good reason for me to judge that the bull is dangerous.” It is true that the emotional response and the evaluative judgment respond to the same descriptive features (i.e., the bull’s sharp horns and its rapid movement toward me), but these features ground two kinds of reasons, reasons for the evaluative judgment and reasons for the emotional response. The descriptive features ground two different kinds of reasons because they underlie both the sentimental evaluative property of being fearsome and the non-sentimental property of being dangerous. Further, because in this case they underlie both the sentimental evaluative property and the non-sentimental evaluative property, the emotional response and the non-affective evaluative judgment (provided there is time to make the latter) will converge. But they of course do not always do so – after all, not everything fearsome is dangerous. Some large, hairy spiders are fearsome but not dangerous: the size, speed, hairiness, etc. make the spider genuinely fearsome, but they do not necessarily make it dangerous. For reasons explained above, this argument generalizes for various other sentimental and non-sentimental property pairs both within the moral domain (e.g., my feelings of guilt track the heuristic property of “the guilt-worthy”, and function as heuristic indicators that my action may well have been wrong all-things-considered) and outside of it (e.g., my feelings of aesthetic delight track the heuristic property of “delighting”, and serve as heuristic indicators that the aesthetic object may be aesthetically valuable all-things-considered).²¹

It also follows from this that, contrary to Brady’s claim, emotional responses do not become otiose as justifiers even when a non-affective evaluative judgment has already been made. This is a welcome implication of the account. Any realistic portrayal of our affective life should recognize that emotions continue to influence our evaluations even after the subject has started to engage in reflective evaluation (Elster, 1999). Nor would it be right to say that such an influence is necessarily pernicious, or even if not pernicious, just an “optional extra”. I move on to elaborating these ideas in the next section.

III.3 Justificatory moderate rationalism

One advantage of the heuristics model defended in this paper is that it offers a detailed and empirically specific account of the complementary epistemic roles played by emotional responses and non-affective evaluative judgments in the formation of our evaluative beliefs. The claim that emotions contribute to our evaluations even after non-affective cognition has kicked in is confirmed by evidence from empirical research²² as well as introspection. Nor does reliance on emotional responses to justify one’s evaluative beliefs indicate any irrationality or lack of (epistemic) virtues. The epistemic input derived from emotions retains its sui generis importance and in a well-circumscribed sense remains a privileged source of epistemic information even after one has started to engage in evaluative reflection.

In other words, the claim I want to defend is not (just) that due to pragmatic constraints we may have to rely on emotional responses because we cannot engage in reflective thinking at all, or that we need to rely on them before we can begin to think more carefully about the situation (cf., Brady, 2013, p. 188). This claim is of course also true. In fact, one of the strengths of the heuristics model is precisely that it can explain why emotions serves us well in such cases: it is because they are a form of heuristic cognition rooted in System 1-type processes.

However, the crucial point is that emotional responses and non-affective evaluative judgments can combine to support a certain evaluative belief.²³ I used earlier Goldie’s story describing how he felt before and after slipping on ice (2000, pp. 59–60). In my analysis, the feeling of fear corroborates the prior reflective judgment that stepping on ice was dangerous. This story is also instructive because it shows how relying on the emotional input can be epistemically beneficial. Emotions provide a sui generis input, a different perspective, which is not reducible to the reflective judgment. This also explains why the emotional input is unlikely to disappear or become impotent once the reflective judgment has been formed. Anchored in a different type of cognitive and neurophysiological architecture, (affective) heuristic cognition continues to “broadcast” as a complementary but autonomous information channel. Nor is the subject irrational in “listening to” information coming from this channel. Convergent information from our emotional response can raise the credibility of the evaluative belief. To return to Goldie’s example, since most fearsome things are dangerous our fear of ice will strengthen the justification of the evaluative belief that the ice is dangerous. For the same reason, divergent information from the evaluative response can lower the credibility of the evaluative belief. Since most things we are afraid of are dangerous, if we experience (say) no fear in a certain situation we have some reason not to believe that the situation is dangerous – even if our reflective judgment of the situation is that it is dangerous. And by the same token, if we experience fear in a certain situation, then we have some reason to believe it is dangerous – even if our reflective judgment of the situation is that it is not dangerous.

So, there is a sense then in which according to the heuristics model emotional experiences are indeed a privileged source of information about value.²⁴ First, as noted above, given the exigencies of life in many cases we do not have the time, possibility and/or cognitive capacity to engage in reflective evaluation. In such situations relying on emotional heuristics is privileged simply because indispensable. Second, some believe that heuristics may outperform reflective thinking even when such practical limitations do not obtain. Third, even when heuristic emotions and non-affective evaluative judgments complement each other in our evaluations, emotional responses contribute useful information by providing access to the sentimental evaluative properties of their intentional objects.

At the end of this section, it may be worth summarizing the differences between the two varieties of rationalism discussed above. Despite important similarities, I believe the moderate epistemic rationalism represented by the heuristics model – justificatory moderate rationalism – fares somewhat better than Brady’s alternative – motivational moderate rationalism. First, justificatory moderate rationalism can take on board the intuition that emotions provide access to distinct kinds of evaluative properties, second, it offers an account of the systematic nature of the relationship between emotional responses and non-sentimental evaluative judgments, and third, based on this account justificatory moderate rationalism can explain how emotions can provide prima facie justification for evaluative beliefs.

IV. Conclusion

This paper presented an empirically informed account of the epistemic role of emotions in evaluations. I have argued that emotions serve as heuristics of value complementing and sometimes replacing non-affective, more or less reflective evaluative judgments. The heuristic theory helps to explain why emotional responses give us a sui generis perspective on value. It can also explain how this perspective interfaces with non-emotional sources of evaluative knowledge. The heuristics model thus represents a moderate form of rationalism, a version I called justificatory moderate rationalism.

The idea that emotions can play a justificatory role has been challenged by other rationalists, even those who are not opposed to attributing some epistemic (albeit non-justificatory) function to emotions. However, these rationalists fail to see that if emotional experiences constitute appraisals, then they can justify evaluations qua appraisals too. There is no problem about double counting either as heuristic emotions do not latch onto the same evaluative properties as non-affective evaluative judgments, and so they generate independent prima facie reasons for evaluative beliefs.

This is of course by no means the end of the debate. Sentimentalists will think that the theory concedes too little to them for several reasons. First, they might complain that the heuristics theory fails to capture the distinctive phenomenology of emotional experiences. And second, they are likely to find that theory fails to capture the privileged metaphysical connection between emotions and value.²⁵ A comprehensive defense of the heuristics model as a form of moderate rationalism will have to address these concerns as well. In my view, these sentimentalist criticisms can be satisfactorily answered, but attempting to do so would go well beyond the scope of this paper which has focused on the epistemic role of emotions in justifying evaluative beliefs.²⁶

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

András Szigeti

András Szigeti is Associate Professor of Philosophy at Linköping University and Associate Director and Research Fellow of the Lund-Gothenburg Responsibility Project (LGRP) at Lund University.

Notes

1. From here onwards, unless otherwise indicated, the labels “rationalism” and “sentimentalism” will stand for views concerning the epistemic role of emotions.

2. There is some debate whether heuristics perform better only given the cost in terms of time and effort of collecting additional information and performing more calculations, or whether heuristics can outperform other cognitive strategies even when “information and computation are entirely free” (see, Gigerenzer & Brighton, 2009, p. 110).

3. Gawronski and Creighton (2012) distinguish “phenomenon specific” dual process theories, “generalized dual process theories” (such as the one discussed above), and quantified dual process theories.

4. For reasons well explained by Sauer (2012), the second point is certainly a valid criticism of problematic applications of the dual process theory as in, for example, Greene et al. (2001).

5. At the same time, I do not mean to claim that all non-reflective contributions to moral deliberation are necessarily emotional. For example, there is room in the account for intuitions that are both non-affective and non-reflective. See more on this point below (esp. footnote 12).

6. The claim that emotional evaluations often seem to be inferential is especially controversial and relates to a debate about whether the content of emotional experience is best characterized as non-conceptual and hence non-inferential, see, esp. (Döring, 2010, 2015) for this claim (and see, Crane, 1992 on the general idea that non-conceptual intentional content is best characterized as non-inferential intentional content).

7. Some deny that emotions can occur in the absence of any accompanying or prior cognitive activity. See, Elster (1999) for the claim that they can and Lazarus (1984) for the claim that they cannot. Note that the latter view can but need not be coupled with judgmentalism, i.e., the nowadays somewhat less popular position that emotions are evaluative judgments (Nussbaum, 2001).

8. Damasio distinguishes between “primary emotions” (e.g., fear, anger), which involve responses to changes in the world or one’s body in a hard-wired, “pre-organized fashion”, and “secondary emotions” which necessarily involve a cognitive-evaluative component (whereby the cognitive-evaluative process precedes the secondary emotion). Further, Damasio holds that the secondary, personalized emotional dispositions are obtained under the influence of primary, that is, hard-wired emotional response mechanisms (Damasio, 1994, p. 136, 1995, p. 22). While Damasio’s work is frequently cited in support of sentimentalist claims about the inseparability of emotion and reason, his views are consistent with the heuristics model. See, for example, here: “At their best, feelings point us in the proper direction, take us to the appropriate place in a decision-making space, where we may put the instruments of logic to good use.” (Damasio, 1994), p. xiii.

9. Sinnott-Armstrong et al. (2010) defends an approach which is in several aspects similar to the model outlined above. It may be instructive here to list some of the differences. First, this model is only about emotions not moral intuitions in general. Second, Sinnott-Armstrong et al. (2010) define heuristics as predominantly subconscious influences on deliberation. We should not accept this definition in my view (Sauer, 2012 seems to agree, see esp. pp. 803–4). I can only discuss this issue briefly here, but I think the difference is quite significant. The point is that we can offer a more realistic account of the interplay of heuristic and reflective cognition if we dispense with the claim that heuristics are subconscious. Specifically, allowing heuristics to be conscious helps to explain (i) the fact that heuristics appear to be quite flexible and corrigible through learning, and (ii) the fact that heuristics do not exert a causal influence on one’s deliberation, but are better described as providing reasons for belief and reasons for action, which reasons are drawn upon by deliberating agents and may or may not conflict with their reflective conclusions. I must leave a detailed discussion of these disagreements with Sinnott-Armstrong et al. (2010) for another occasion.

10. As Echeverri (2019) notes, this idea is shared by most philosophers of emotion (though not all, as we will see, in Section III). Further, it is often said that emotions present or represent such evaluative properties, e.g., my fear (re)presents the dog as fearsome. The representationalist view is not entirely unproblematic (see, Echeverri, 2019 for discussion; Mulligan, 2009 as well as Deonna & Teroni, 2012 for criticisms). However, nothing in the following depends on whether one accepts “representationalism” or not. What matters is that one accepts the intentionality of emotions and that they provide access to the evaluative properties of their intentional objects. These claims are very widely accepted (see, Echeverri, 2019, p. 4). To make clear that I am committed to these two claims (but not necessarily to representationalism), I will say that emotions “pick out” evaluative properties.

11. This characterization entails that other affective phenomena, such as moods, are not covered by the heuristics model – provided, that is, one accepts that the intentionality of moods is (at best) diffuse (Ben-Ze’ev, 2009). The latter claim means that, in contrast to emotions, moods, while being “about” the world in some sense, are not about specific objects or situations. This is controversial. Schwarz and Clore (1983) found that moods can also play an epistemic role: they have a direct “informational function,” as they put it.

12. It bears repeating that these non-affective appraisals need not always be the result of slow and careful deliberation. They can also be instinctive, intuitive, or habitual reactions. Throughout the paper, I use the term “non-affective evaluative judgment” broadly to cover all such non-affective appraisals.

13. It may be objected here that these are not really the issues many sentimentalists are interested in. Rather, the primary concern of sentimentalism is ontological: it is about what values are and how to explain them. The metaphysical sentimentalist’s thesis is that values depend essentially upon emotional responses, and so emotions are to be understood as constitutive of value (see, for example, D’Arms & Jacobson, 2017; Kauppinen, 2014 for an overview). I believe that it is nevertheless worth spending time on the epistemic side of the debate for two reasons. First, there is a plausible case to be made that metaphysical sentimentalism entails epistemic sentimentalism. This implication is clearly recognized by D’Arms and Jacobson (2010) who write: “Our point is … that [emotional sensibilities] have internal connections to sentimental values which theoretical reflection does not. This fact renders such reflection prone to forms of error and confusion from which sensibilities are immune.” The same implication is also highlighted by Pettit (1991), Elgin (2008), and Szigeti (2021). Second, epistemic sentimentalism matters because one can accept epistemic sentimentalism without being a metaphysical sentimentalist. It may be argued that, while value is not ontologically dependent on emotions, as a matter of empirical fact evaluative properties are most reliably picked up by our emotional responses. (An analogous and not implausible position in the philosophy of perception would be this: the existence of color properties is not dependent on the dedicated visual apparatus of color perceiving creatures since color properties are reducible to physical properties “out there”. However, as a matter of empirical fact color properties are most reliably picked up by the visual apparatus of color perceiving creatures including human beings.) So, the heuristics theory puts pressure on epistemic sentimentalism whether it is coupled with metaphysical sentimentalism or not. I thank an anonymous reviewer for insisting that I clarify the relationship between metaphysical and epistemic sentimentalism.

14. It is customary to ascribe this view to the Stoics, but this may not be historically inaccurate (see, Perler, 2011, esp. pp. 120–122). Hence the quotation marks.

15. Similarly for aesthetic judgments, it could be in accordance with my professed aesthetic values as well as the rave reviews to conclude that that opera production was a masterpiece. Nevertheless, when trying to decide what I should think about it at the end of the day, would it not be perfectly appropriate, in fact epistemically virtuous, to also take into account that the piece actually bored me to death? And conversely: I know I should be finding this movie cheap and vulgar, but it made me laugh – should that not make me think that there may be some value to it?

16. So, the experience of unfocused moods (e.g., lingering anxiety) that lack an intentional object will not provide justificatory reasons for an evaluative belief. See also footnote 11 above.

17. This objection was originally put forward in Brady (2013) against the perceptual theory of emotions, but it applies equally to the heuristics model or any theory holding that emotions can play a justificatory role.

18. What makes the double-counting objection particularly important is that the assumption on which it is based – i.e., that the same evaluative property gives reasons both for the emotional response and the evaluative judgment – is even shared by authors (see, esp. Deonna & Teroni, 2012, p. 118 f.; Goldie, 2004, p. 98) who, in contrast to Brady, are willing to assign a justificatory epistemic role to emotions.

19. In fact, as will become clear, the reasons to reject this assumption are not predicated on acceptance of the heuristics theory. The assumption will be rejected by any approach which holds that emotional responses and evaluative judgments track different kinds of properties (see, for example, Schroeter, 2006). The heuristics theory is a specific account of how these different kinds of properties are accessed and used in cognitive processes.

20. Consider the following objection: the analogy does not hold because if the “fearsome” is distinct from “dangerous” as has been argued above, then we do not have two different forms of evidence for Fa. We have the judgment that Fa and the feeling that Ga. That is correct. However, since in many cases something that is Fa is also Ga (i.e., “the fearsome” is dangerous), there is a fairly reliable correlation between a being F and a being G, and so it is justified to rely on a being F as evidence that a is G. Note also that despite there being such a correlation the two sources of evidence are mutually independent. In accordance with the heuristics theory, our information regarding a being F is supplied by affective heuristics, whereas our information regarding a being G is supplied by non-heuristic cognition. More in Section III.3. below on the correlation between sentimental and non-sentimental evaluative properties.

21. It is worth noting here that this need not imply that any of the evaluative properties discussed above would necessarily be objective or mind-independent.

22. Murphy and Zajonc (1993) found that affective valence associated with objects or events will substantially influence evaluations of those objects or events. Note that affect has been shown to have such impact whether or not the affect is consciously felt, and even if the affective state in question was mere emotional arousal involving no cognitive activity (Kunst-Wilson & Zajonc, 1980; Zajonc, 1980).

23. Several examples have already been discussed in the foregoing. Additional evidence from experimental research is also available. For example, it has been found that people use affective tagging to render information meaningful. If some piece of information attracts no emotional response, then people will have trouble assessing its importance and relevance. Among others, this could explain people’s often peculiar attitudes toward risk and probability as well as their surprising assessments of potential outcomes when deciding on future policies (Finucane et al., 2000; Slovic et al., 2002). Incidentally, it has been observed that time-pressure increases the inclination to employ emotion-based assessments instead of reflective evaluations (Finucane et al., 2000; Slovic et al., 2002) – just as the heuristics model predicts.

24. Compare and contrast (Deonna & Teroni, 2012).

25. See footnote 13 above.

26. I wish to thank Paul Russell, Alessandra Tanesini, Jonathan Webber as well as the anonymous referees of this journal for helpful comments and discussions on various versions of this paper. Further, I gratefully acknowledge the financial support of the Lund Gothenburg Responsibility Project (LGRP) funded by the Swedish Research Council.