Positive Emotions in Moral Judgment 1 Running head: POSITIVE EMOTIONS IN MORAL JUDGMENT Divergent Effects of Different Positive Emotions on Moral Judgment

Positive emotions are often treated as relatively similar in their cognitive-behavioral effects, and as having unambiguously beneficial consequences. For example, Valdesolo and DeSteno (2006) reported that a humorous video made people more prone to choose a utilitarian solution to a moral dilemma. They attributed this finding to increased positive affect. To determine whether such results actually stem in general from positive affect or from other more specific properties of humor, we conducted an experiment with moral dilemmas presented during an interleaved emotion-induction procedure involving mirth and another positive emotion, elevation. Mirth increased permissiveness for deontological violations, whereas elevation had the opposite effect. Furthermore, affective valence had no apparent independent influence on these judgments. Our results suggest that mirth and elevation have distinct cognitive consequences whose properties reflect their respective social functions, not their shared positive valence. Positive Emotions in Moral Judgment 3 Divergent Effects of Different Positive Emotions on Moral Judgment Although recent research has advanced our understanding of positive emotions, they are still less well understood than negative emotions. One striking difference in this understanding is the extent to which current accounts hypothesize a set of positive emotions with distinct social functions, neural bases, and physiological correlates. Whereas a rich body of theory distinguishes various negative emotions along all of these dimensions, in most accounts of positive emotion, valence still plays the principal role. A prominent example of such an account is the broaden-and-build theory (Fredrickson, 2004), which posits that positive emotions serve to expand thought-behavior repertoires, whereas negative emotions narrow this scope. However, this theory does not distinguish among the effects of different positive emotions, and previous empirical tests have focused on how different positive emotional states lead to comparable behaviors (Fredrickson & Branigan, 2005). Such views have encouraged experimenters to use a diverse set of stimuli for inducing positive affect, in effect treating diverse positive emotional states as equivalent (Isen et al., 1985; Hamann et al., 2002). The valence-based approach to positive emotion has permeated many areas of research, including moral psychology. In a study that has gained widespread attention, Valdesolo and DeSteno (2006) used a single humorous video to induce positive mood before having participants resolve a moral dilemma, the footbridge problem (Thomson, 1976), which concerns whether it is acceptable to push a stranger off a bridge in order to prevent a trolley from killing several other people underneath. The humorous video increased participants‘ tendency to favor the unconventional utilitarian solution, which entails a deontological violation (pushing the stranger off the bridge). Explaining their results, Valdesolo and DeSteno claimed that positive Positive Emotions in Moral Judgment 4 affect attenuates the negative affect that would normally accompany the deontological violation. We will call this account the negativity-diminishment hypothesis, which resembles Fredrickson et al.‘s (2000) undoing hypothesis (an outgrowth of the broaden-and-build theory whereby positive emotions ̳undo‘ the cognitive and physiological effects of negative emotions). Meanwhile, in recent years, an alternative view of emotions has gained traction. The appraisal-tendency framework (ATF; Lerner & Keltner, 2000) posits that the influence of an emotion stems from multiple dimensions. Emotions that have the same valence nevertheless differ in terms of properties associated with their unique cognitive, motivational, and social functions. ATF has been successfully tested for negative emotions (Lerner et al., 2004), although it remains largely untested for positive emotions. Under the appraisal-tendency hypothesis, Valdesolo and DeSteno‘s (2006) results may not be related to positive affect, but rather to the distinct properties of mirth, the positive emotion associated with humor (Martin, 2007). Although an appraisal tendency for mirth has not previously been articulated, we will suggest one here. A central property of humor is that topics—particularly those usually afforded deep respect—are taken lightly. Therefore, mirth‘s appraisal tendency may be to increase irreverence and remove the gravitas of otherwise serious ideas. In the context of moral judgment, this suggests that mirth would increase permissiveness for moral violations, including deontological violations. If this explanation is correct, then other positive emotions might not have the same effect on moral judgments. The alternative positive emotion that we consider here is elevation, which arises from witnessing acts of moral beauty, and establishes a mindset whereby people want to act in a more noble, saint-like way (Haidt, 2003; Silvers & Haidt, 2008). While both mirth and elevation are positively valenced, they oppose each other in some important respects: elevation is Positive Emotions in Moral Judgment 5 associated with reverence, whereas mirth is associated with irreverence. When it comes to footbridge-type moral dilemmas, elevation might make people less likely to endorse the violence or direct harm entailed by utilitarianism, instead encouraging decisions that seem ―decent‖ rather than cold and calculating. Given these considerations, our research has two related objectives. First, we investigate the extent to which positive affect necessarily increases utilitarian moral judgment. Second, we investigate whether particular positive emotions can have consequentially distinct influences on judgment and decision-making. The results of our study have implications for both moral cognition and the study of emotion. Empirical Approach To achieve our objectives, we have devised a new interleaved emotion-induction procedure, which includes multiple trials for each participant. On each trial, an emotioninduction stimulus was presented, a participant rated his or her degree of emotional response to it, and then rated the permissibility of a deontological violation in a moral dilemma presented immediately afterwards. Across trials, multiple induction stimuli and moral dilemmas occurred, enabling us to measure the degree of induced emotion and how much it affected permissibility ratings. To avoid complex interactions between qualitatively different emotional states, we had each participant serve under only one emotion-induction condition. Our procedure has several advantages. It provides repeated measures of the effects caused by induced emotions, testing to what extent they occur reliably across individual induction stimuli and moral dilemmas. This approach has considerably more power and generality than single-trial methods, which involve many fewer (and possibly idiosyncratic) Positive Emotions in Moral Judgment 6 stimuli. Such power and generality may be especially important when assessing subtle effects of emotions like mirth and elevation. Method Preliminary Norming Study Before our main experiment, we conducted a preliminary stimulus-norming study. Thirty audio clips were assembled for each of three emotion-induction conditions: mirth, elevation, and neutral control. For the mirth condition, audio clips were taken from the stand-up comedy of Jerry Seinfeld, Mitch Hedberg, Steven Wright, and Demetri Martin, whose observational humor is funny but not overtly aggressive or offensive. For the elevation condition, we extracted inspirational stories from a volume of Chicken Soup for the Soul (Canfield, 2002). Audio clips based on these extracts were recorded by four student actors (two male, two female). For the neutral (control) condition, we used lectures by four professors (two male, two female) from the Science and the City podcast. These stimuli were chosen to be engaging but not to contain contentious social or political content. Each clip lasted 3.5 to 4 minutes. 1 Each of sixty-two participants was assigned randomly to one of the three conditions. For each audio clip, a participant answered several questions about it on Likert rating scales. These questions measured general positive affect and specific degrees of induced mirth, elevation, and interest (Table 1). Not all questions were asked in all conditions, both because this would have made the study too long for each subject to rate all clips, and because it was thought that some questions might have dual meanings depending on the condition (for instance, stand-up comedy might be considered ―uplifting‖ in a qualitatively distinct way from acts of moral beauty, and we Positive Emotions in Moral Judgment 7 Table 1 Mean Ratings in Response to Emotion-Induction Stimuli in the Preliminary Norming Study. Note. “Variable‖ indicates labels for predictors used in subsequent analyses. Ratings were made on 1-7 Likert scales. Statistically reliable contrasts, controlling for valence, are indicated by **> and **<. Blank entries indicate the questions was not asked under that condition. Positive Emotions in Moral Judgment 8 did not want to falsely conflate these measures). Ratings in response to questions that assessed degrees of mirth (funny, laughing, smiling) and elevation (uplifting, inspiring, tearing, chest warmth, message) were positively correlated with rated positive affect, confirming that these are both positive emotions (Table 2). Stimuli for Main Experiment Emotion Induction To assess the effects of positive emotions on moral judgments in the main experiment, we used 12 audio clips for each of three emotion-induction conditions: mirth, elevation, and neutral control. These clips were selected from the larger sets of the preliminary norming study with the goal of maximizing relevant affective experiences and minimizing other irrelevant experiences while keeping overall valence high and nearly equal across the positive-emotion conditions. Table 3 shows the mean parameters of the audio clips selected for the main experiment. Moral Dilemmas The experiment used 24 moral dilemmas similar to the footbridge problem (Greene et al., 2001), modified to have third-person phrasing. Participants Participants (29 females, 26 males) in the main experiment were native English-speaking, American undergraduates who received either course credit or payment. Design and Procedure Participants wore headphones while sitting in front of a computer with a display monitor at eye level and a keypad nearby. Each participant was assigned randomly to one of the three emotion conditions. There were 12 successive trials, each involving a randomly paired emotionPositive Emotions in Moral Judgment 9 Table 2 Correlations between Ratings about Emotion Variables and Positive Valence in the Preliminary Norming Study. Note. Pearson product-moment (r) correlation coefficients. t-values associated with the tests are all reliably different from zero (p <= 0.0001). induction clip and moral dilemma. No clip or dilemma was presented more than once per participant. On each trial, a participant heard an audio clip and then rated it on a 1–6 scale appropriate to the assigned condition: 1 = not at all funny/uplifting/interesting, 6 = extremely funny/uplifting/interesting. After making this rating by using the keypad, the participant read a moral dilemma displayed on the computer monitor, and rated how permissible a deontological violation was on a 1–6 scale (1 = ―forbidden‖, 6 = ―completely permissible‖). Each participant gave permissibility ratings for 12 of the dilemmas. Positive Emotions in Moral Judgment 10 Table 3 Mean Values of Ratings Associated with Emotion-Induction Stimuli Used in the Main Experiment. Note. Numerical values come from 1–7 Likert scales and represent the mean ratings for audio clips selected from stimuli in the preliminary norming study. Blank entries indicate that particular questions were not asked in some conditions. Results To analyze our data we used mixed-effects linear regression models (Baayen et al, 2008; Gelman & Hill, 2006), which quantified systematic relationships between trial-by-trial permissibility ratings and predictor variables based on emotion ratings from the main experiment and preliminary norming study. One of our regression models tested whether positive valence was sufficient to account for the obtained permissibility ratings. It had the form permissibility-rating = positive-valence + Positive Emotions in Moral Judgment 11 participant-random-effect + dilemma-random-effect + trial-number. Here permissibility rating corresponded to individual moral judgments made on each trial using the 1–6 scale described previously. Positive-valence corresponded to mean ratings of ―positive feeling‖ obtained in the preliminary norming study for each induction stimulus presented on trials of the main experiment. Another augmented regression model added emotion-clip-rating to the right side of the preceding equation. It tested to what extent permissibility ratings depended on the strength of the target emotion induced on each trial. Trial did not significantly predict permissibility ratings in any of the models. Positive-valence effects. Subjective positive valence induced by audio clips did not account reliably (p > .4) across trials for permissibility ratings about moral dilemmas; t = 0.29, 0.07, and -0.85 in the mirth, elevation, and neutral conditions, respectively. Combining conditions, there was still no reliable effect of positive valence (t = -0.50, p > .6). Furthermore, when other condition-specific predictors (i.e., ―funniness‖, ―uplifted‖, and ―interest‖) were taken into account, the correlation between permissibility ratings and positive valence was essentially zero (Table 4). Such null results cast doubt on the sufficiency of Valdesolo and DeSteno‘s (2006) negativity-diminishment hypothesis. Mirth effects. For the mirth condition with positive-valence effects partialed out, trial-bytrial ―funniness‖ ratings about the humorous audio clips in the main experiment accounted well for permissibility ratings (Table 4). There was a reliable (t = 2.34, p = 0.02) positive partial correlation: the higher the rated ―funniness‖, the more the participants tended to favor deontological violations (Figure 1). This confirms our prediction based on the hypothesis that mirth has an appraisal tendency of irreverence. Positive Emotions in Moral Judgment 12 Elevation effects. For the elevation condition with positive-valence effects partialed out, trial-by-trial ―uplifted‖ ratings obtained in response to the audio clips of the main experiment did not, by themselves, account reliably (t = 0.42, p = 0.67) for permissibility ratings about the moral dilemmas. This outcome might conceivably be explained in two ways: our theorizing about emotional elevation is incorrect, or ―uplifted‖ ratings are not the best indicator for the presence of elevation. Consistent with the second possibility, Silvers and Haidt (2008) found that elevation may be better measured through ratings of subjective inspiration and self-reported physiological responses such as eye watering. Hence, we performed an additional regression analysis with another proxy measure for the induced degree of elevation in the main experiment, using sums of the ―inspiring‖, ―tearing‖, and ―chest warmth‖ ratings from our preliminary norming study. This predictor accounted reliably (t = -2.87, p = 0.004) for permissibility ratings in the elevation condition. Audio clips that had previously elicited higher ratings for these properties of elevation predicted lower permissibility ratings for the moral dilemmas (Table 4 and Figure 2). (Indeed, all of the elevation predictors have negative coefficients.) When this latter trend is compared to what happened in the mirth condition, there is a reliable cross-over interaction (t =3.68, p = 0.0003). These findings cast further doubt on the negativity-diminishment hypothesis while supporting the appraisal-tendency hypothesis. 1 “Interesting” control effects. For the neutral condition, trial-by-trial ratings of ―interest‖ induced by the audio clips did not account reliably (t = 1.23, p = 0.22) for permissibility ratings in response to the moral dilemmas. With positive valence partialed out, these ratings had correlations and regression slopes close to zero (Table 4 and Figure 1). Positive Emotions in Moral Judgment 13 Table 4 Summary of the Relationships between Ratings of Emotion-Induction Stimuli and Permissibility Ratings in the Main Experiment, Controlling for Positive Valence. Note. The estimates are coefficients for standardized predictors obtained from mixed-effects linear regression models using mean valence ratings from the initial norming study as covariates. The t-values and p-values come from statistical tests about whether these coefficients differ reliably from zero. ―*‖ indicates coefficients that are reliably different from zero at the  = 0.05 level. Positive Emotions in Moral Judgment 14 Figure 1. Permissibility ratings for deontological violations in moral dilemmas in the main experiment as a function of rated emotional experiences induced by preceding audio clips. The emotion ratings for the clips pertain to funny in the mirth condition, a sum of tearing, chest warmth, and inspiration in the elevation condition, and interesting in the neutral condition (see Table 4). The regression lines are based on coefficient estimates for predictor variables used in mixed-effects linear regression modeling (with standardized predictors), which demonstrate opposing effects of mirth and elevation on moral judgments. The plotted points represent means from approximately equal-size bins, each of which sums over about 20 trials. Mixed-effects statistical tests indicate that the slopes for the mirth and elevation conditions are reliably different from zero and from each other. The mirth slope is also different from neutral; see text for more details. Positive Emotions in Moral Judgment 15 Furthermore, when one takes the top-rated emotion clip (or clips in the case of a tie) for each participant—the funniest, most interesting, or most elevating—a reliable linear trend emerges in the permissibility ratings across the three emotion conditions, with mirth leading to the highest mean permissiveness and elevation the lowest (Figure 2), t(51) = 2.44, p=0.018. 2 Figure 2. Mean permissibility ratings obtained in the three emotion conditions of the main experiment for trials involving clips that had the most powerful experiences of the emotions. Confidence intervals indicate standard errors around group means. Positive Emotions in Moral Judgment 16 Discussion Contrary to the negativity-diminishment hypothesis, our results cannot be explained simply in terms of positive affect. Instead, the present study supports the view that positive emotions are functionally distinct, and influence decision-making according to emotion-specific appraisals. The opposing effects of mirth and elevation on permissiveness towards deontological violations suggest that specific properties of positive emotions can have substantially different consequences for human cognition and action. While we have made progress toward understanding the nature of positive emotions, a number of explanations for our observed effects remain. Earlier in this paper we suggested that mirth‘s appraisal tendency might be increased irreverence, leading to higher permissiveness for deontological violations. This suggests that mirth would not be discriminating in the types of moral behavior it would condone; indeed, if our moral dilemmas were rephrased to probe the acceptability of utilitarian violations, we may well find that mirth also increased permissiveness for these actions as well. It is nonetheless possible that mirth specifically favors deontological violations, for example because they represent an especially strong form of irreverence. At the very least, the findings of Valdesolo and DeSteno (2006) were not due to an idiosyncratic relationship between the original footbridge problem (which involves a fat stranger) and a comedy starring Chris Farley (an elephantine man who would make conspicuously good ballast). 4 There is less ambiguity in the finding that elevation lowers permissibility ratings. Our clear prediction was that elevation would lower permissiveness because elevation is associated with moral reverence. It may not be a coincidence that elevation is associated with moral precursors like affiliation, empathy, and helping, which are the same tendencies proposed to be Positive Emotions in Moral Judgment 17 driving decisions about deontological solutions to footbridge-type moral dilemmas (Mendez et al., 2005; Koenigs et al., 2007). It is critical to note that elevation does not influence all aspects of morality equally; for instance, increased elevation did not make subjects more attuned to saving the most lives. Instead, elevation appears to activate and amplify moral responses unrelated to simple utility maximization. In the future, methods like those used here may have useful applications for a variety of other investigations regarding the functional effects of emotion. Interleaved emotion-induction differs from standard mood-induction procedures in that it allows the experimenter to measure emotion induction on a trial-by-trial basis. The increased power of this approach stems from both its repeated-measures design and its sensitivity to individual self-reports about the strength of emotion felt at the start of each moral-judgment trial. These reports were augmented with information gathered about specific items in an independent norming study involving the same population. Such methods may be especially useful when inducing emotions that require stimuli for which there is a high degree of individual variation in personal reactivity, as is the case with mirth and elevation. The present research also bears on other recent explanations of moral decision making, which emphasize its grounding in both cognitive and affective systems (Greene & Haidt, 2002). We have demonstrated that such decisions may be modulated indirectly—in different ways—by emotions induced with stimuli separate from the targets of the decisions themselves. Such distal effects are not necessarily implied by an assumption that moral decisions involve an immediate proximal affective process. Our results thus provide significant new evidence to constrain theories about the overall relationship of moral and affective processing. Positive Emotions in Moral Judgment 18 Furthermore, these constraints add nuance to the popular view that positive emotions uniformly promote good or beneficial behavior. Given that mirth and elevation have opposing effects on moral judgment, it makes little sense to describe such positive emotions as bringing about a ̳desirable‘ outcome in this setting. Rather, these effects reflect the emotion‘s function and the context in which it is placed. Positive emotions may be functional, but this is distinct from claiming their effects are unambiguously positive (Norem & Chang, 2002; Campos, 2003). Moving away from normative labels will let us better appreciate the nature of positive emotions: their properties, functions, and relation to one another. Positive Emotions in Moral Judgment 19 Acknowledgments Nina Strohminger was supported in part by a graduate fellowship from the William Orr Dingwall Foundation. We thank Robert Mankoff, Joshua Carp, and Jonathan Haidt for helpful comments. Positive Emotions in Moral Judgment 20 ReferencesBaayen, R. H., Davidson, D. H., & Bates, D. M. (2008). Mixed-effects modeling with crossedrandom effects for subjects and items. Journal of Memory and Language, 59, 390-412.Campos, B. (2003). When the negative becomes positive and the reverse: Comments onLazarus‘s critique of positive psychology. Psychological Inquiry, 14, 110-172.Canfield, J. (2002). Chicken Soup for the Volunteer’s Soul. Deerfield Beach, FL: HCI.Fessler, D. M. T. & Haley, K. J. (2003). The strategy of affect: Emotions in human cooperation. In P. Hammerstein (Ed.), Genetic and Cultural Evolution of Cooperation (pp. 7-36).Cambridge, MA: MIT Press.Fredrickson, B. L. (2004). The broaden-and-build theory of positive emotions. PhilosophicalTransactions of the Royal Society of London B, 359, 1367-1377.Fredrickson, B. L., Mancuso, R.A., Branigan, C., & Tugade, M. M. (2000). The undoing effectof positive emotions. Motivation and Emotion, 24, 237-258.Fredrickson, B.L. & Branigan C. (2005). Positive emotions broaden the scope of attention andthought-action repertoires. Cognition and Emotion, 19, 313-332.Gelman, A. & Hill, J. (2006). Data Analysis Using Regression and Multilevel/HierarchicalModels. New York: Cambridge University Press.Greene, J. D., Sommerville, R. B., Nystrom, L. E., Darley, J. M., & Cohen, J. D. (2001). AnfMRI investigation of emotional engagement in moral judgment. Science, 293, 2105-2108.Greene, J. D. & Haidt, J. D. (2002). How (and where) does moral judgment work? Trends inCognitive Science, 6, 517-523. Positive Emotions in Moral Judgment 21 Haidt, J. D. (2003). Elevation and the positive psychology of morality. In C. L. M. Keyes & J.D. Haidt (Eds.), Flourishing: Positive psychology and the life well lived (pp. 275-289).Washington DC: American Psychological Association.Hamann, S. B., Ely, T. D., Hoffman, J. M., & Kilts, C. D. (2002). Ecstasy and agony: Activationof the human amygdala in positive and negative emotion. Psychological Science, 13, 135-141.Isen, A. M., Johnson, M. M. S., Mertz, E., & Robinson, G. F. (1985). The influence of positive affect on the unusualness of word associations. Journal of Personality and SocialPsychology, 48, 1413-1426.Kirchsteiger, G., Rigotti, L., & Rustichini, A. (2006). Your morals might be your moods.Journal of Economic Behavior and Organization, 59, 155-172.Koenigs, M., Young, L., Adolphs, R. Tranel, D., Cushman, F., Hauser, M., & Damasio, A.(2007). Damage to the prefrontal cortex increases utilitarian moral judgments. Nature, 446,908-911.Lerner, J.S., Small, D.A., & Loewenstein G. (2004). Heart strings and purse strings: Carryovereffects of emotions on economic decisions. Psychological Science, 15, 337-341.Lerner, J. S. & Keltner, D. (2000). Beyond valence: Toward a model of emotion-specificinfluences on judgment and choice. Cognition and Emotion, 14, 473-493.Martin, R. A. (2007). The Psychology of Humor. Burlington, MA: Elsevier Academic Press. Mendez, M. F., Anderson, E. & Shapira, J. S. (2005). An investigation of moral judgment infrontotemporal dementia. Cognitive Behavioral Neurology, 18, 193–197.Norem, J. K. & Chang, E. C. (2002). The positive psychology of negative thinking. Journal ofClinical Psychology, 58, 993–1001. Positive Emotions in Moral Judgment 22 Silvers, J. & Haidt, J. D. (2008). Moral elevation can induce nursing. Emotion, 8, 291-295.Thomson, J. J. (1976). Killing, letting die, and the trolley problem. The Monist, 59, 204-217.Tversky, A. & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.Science, 185, 1124-1130.Valdesolo, P. & DeSteno, D. (2006). Manipulations of emotional context shape moral judgment.Psychological Science, 17, 476-477. Positive Emotions in Moral Judgment 23