Individual Differences in Working Memory Capacity, Executive Control, and Moral Judgment

Recent findings suggest that exerting executive control influences responses to moral dilemmas. In our study, subjects judged how morally appropriate it would be for them to kill one person to save others. They made these judgments in 24 dilemmas that systematically varied physical directness of killing, personal risk to the subject, inevitability of the death, and intentionality of the action. All four of these variables demonstrated main effects. Executive control was indexed by scores on working-memory-capacity (WMC) tasks. People with higher WMC found certain types of killing more appropriate than did those with lower WMC and were more consistent in their judgments. We also report interactions between manipulated variables that implicate complex emotion-cognition integration processes not captured by current dual-process views of moral judgment. When Pilate saw that he could prevail nothing, but that rather a tumult was made, he took water, and washed his hands before the multitude, saying, I am innocent of the blood of this just person. . . . (Matthew 27:24, King James Version) The synoptic gospels are notoriously ambiguous in assigning earthly culpability for Jesus’ death. Rather than actively sentencing Jesus to crucifixion, Pilate passively allows it by letting the multitude decide whether to release Jesus or Barabbas. When the crowd chooses Barabbas, Pilate claims his hands are clean regarding Jesus’ fate. We suggest that the gospel authors captured something important about humankind’s moral sense, particularly regarding the morality of being the direct agent of killing. Philosophers have long considered such issues in developing prescriptive ethics, but scientists are only now studying moral judgments in order to understand the generative processes that influence them. A lively theoretical debate concerns the role of emotion versus that of normative-rule-based cognition in driving moral judgments (e.g., Blair, 1995; Haidt, 2001; Lakoff, 2002; Mikhail, 2007; Nichols, 2002; Pizarro & Bloom, 2003; Prinz, 2006). Much of the excitement surrounding this debate derives from the neuroimaging work of Greene and his colleagues (Greene, Nystrom, Engell, Darley, & Cohen, 2004; Greene, Sommerville, Nystrom, Darley, & Cohen, 2001; see also Borg, Hynes, Van Horn, Grafton, & Sinnott-Armstrong, 2006). Using functional magnetic resonance imaging (fMRI), Greene et al. investigated the neural substrates of judgments about hypothetical personal versus impersonal dilemmas. In personal dilemmas, subjects contemplated (a) causing serious bodily harm or death to (b) another person or persons, (c) in a way that did not simply deflect preexisting harm onto another person; personal dilemmas thus involved ‘‘up close and personal’’ harmful acts the subjects directly initiated. Impersonal dilemmas did not meet these three criteria. These authors illustrated personal dilemmas with the footbridge problem, in which a runaway trolley hurtles toward five unaware workmen; the only way to save them is to push a heavy man (standing nearby on a footbridge) onto the track, where he will die in stopping the trolley. In an impersonal version of this dilemma, the trolley problem, one may save the workmen by throwing a switch that simply diverts the trolley onto another track, killing one worker. Most people find the action in the impersonal dilemma to be morally more acceptable than the action in the personal dilemma, despite their identical consequences (e.g., Hauser, Cushman, Young, Jin, & Mikhail, 2007). Why? Emotion prevails in personal dilemmas. Greene et al. (2001, 2004) demonstrated that personal dilemmas engage brain regions involved in emotion (e.g., amygdala, posterior cingulate/ precuneus), whereas impersonal dilemmas activate areas involved in deliberative reasoning and working memory (e.g., middle frontal gyrus, bilateral parietal lobe). Moreover, when subjects judge resolutions to personal dilemmas to be morally Address correspondence to Adam B. Moore, Department of Psychology, Green Hall, Princeton University, Princeton, NJ 08550, e-mail: [email protected]. PSYCHOLOGICAL SCIENCE Volume 19—Number 6 549 Copyright r 2008 Association for Psychological Science appropriate, they respond slowly and engage networks associated with executive control (e.g., anterior dorsolateral prefrontal cortex, anterior cingulate). Greene et al. thus proposed a dualprocess model: Personal moral dilemmas evoke an automatic, ‘‘hot’’ emotional response that biases against responding in a way that causes harm, but this response may be overridden by ‘‘cold’’ cognitive control. In our study, we tested this dual-process theory by asking whether individual differences in control, indexed by scores on working-memory-capacity (WMC) tasks, would predict cold, consequentialist responses to personal dilemmas. Research with nonemotive tasks indicates that (a) WMC variation predicts executive ability to override prepotent responses (e.g., Kane, Bleckley, Conway, & Engle, 2001; Kane & Engle, 2003), (b) WMC variation predicts abstract-reasoning performance (e.g., Kane, Hambrick, & Conway, 2005; Oberauer, Schulze, Wilhelm, & Sü , 2005), and (c) experimentally increasing working memory load impairs reasoning about future consequences (e.g., Hinson, Jameson, & Whitney, 2003). Therefore, we hypothesized that by controlling emotion and engaging deliberative processing, people with greater WMC would more rationally evaluate consequences within personal moral dilemmas. This is a strong prediction because executive theories of WMC (e.g., Engle & Kane, 2004; Hasher, Lustig, & Zacks, 2007) are supported almost exclusively by studies of cold cognition (but see Braver, Gray, & Burgess, 2007), and individual differences influencing such moral judgments are rare (Hauser et al., 2007; O’Neill & Petrinovich, 1998). Our second goal was to test whether the personal-impersonal distinction would survive a retooling of the dilemmas presented to subjects. To do this, we redesigned the materials used by Greene et al. (2001, 2004) to address several significant problems with them: (a) More personal than impersonal dilemmas involved death or injury, and more impersonal than personal dilemmas involved lying or stealing; (b) unlike personal dilemmas, many impersonal dilemmas presented abstract, probabilistic-reasoning problems; (c) several scenarios posed nondilemmas (e.g., should a child murder his grandmother for not buying him a gift?); (d) the descriptions of personal dilemmas were longer than the descriptions of impersonal dilemmas (Ms 5 124 and 102 words, respectively); (e) whether or not subjects’ lives were endangered was varied unsystematically across the dilemmas; and (f) subjects saw multiple versions of some scenarios, so carryover effects were possible (a potential problem for other, similar studies as well: Hauser et al., 2007; Mikhail, 2007; Petrinovich, O’Neill, & Jorgensen, 1993; Valdesolo & DeSteno, 2006). Some of these shortcomings reflect vagueness in how Greene et al. (2001, 2004) defined impersonal dilemmas. Personal dilemmas met three criteria (bodily harm that was caused to persons and did not simply deflect existing harm), and any dilemmas meeting none, one, or two of these criteria were considered impersonal. Consequently, very different dilemmas were contrasted, so the materials failed to systematically capture the vexing psychological distinction between the footbridge and trolley problems. For example, dilemmas involving whether to make charitable donations, steal or damage property, or enact environmental-hazard policies were included in the impersonal dilemmas along with those involving deflecting impending threats of death (e.g., a runaway trolley, poisonous fumes) from an innocent group toward an innocent individual. To better equate the severity of outcomes of personal and impersonal dilemmas, we redefined the personal-impersonal distinction to reflect more versus less direct killing (see also Royzman & Baron, 2002; Spranca, Minsk, & Baron, 1991). All critical dilemmas involved killing one individual to save more people, but the killing contemplated in personal dilemmas was more physically direct than the killing contemplated in impersonal dilemmas; that is, the killing involved in personal dilemmas was less mediated through mechanical-technological means or through other people’s actions. Our third goal was to test whether other philosophically relevant variables influence moral judgment (perhaps moderating any effects of the personal-impersonal variable). First, because everyday moral dilemmas often involve costs and benefits to the self, and trolley-dilemma judgments change with subjects’ relationships to the hypothetical victims (Petrinovich et al., 1993), we systematically varied self-risk. We predicted that killing to save oneself and other people would be more acceptable than killing to save only other people; we view this self-other distinction as evolutionarily ingrained and thus automatically and intuitively processed (cf. Petrinovich et al.). Second, we manipulated whether the intended object of harm would die regardless of subjects’ actions, and we predicted that subjects would find killing more acceptable if it merely hastened an inevitable death than if the death were avoidable; we view this inevitable-avoidable distinction as relatively subtle and thus more rationally and deliberatively processed than the self-other distinction. Third, as have other researchers (Borg et al., 2006; Cushman, Young, & Hauser, 2006; Hauser et al., 2007; Mikhail, 2007), we examined Thomas Aquinas’ (trans. 1988, pp. 226– 227) doctrine of double effect (DDE), according to which harm is more permissible if it is a foreseen but unintended consequence than if it is an intended means to an end. Normative sensitivity to the DDE may partially explain the effects of the personal-impersonal distinction: In footbridge/trolley-type dilemmas, one either intentionally uses the bystander’s body (e.g., by throwing him off the footbridge) or unintentionally sacrifices another workman (e.g., by throwing the switch). People may therefore blanch at resolutions to personal dilemmas because they have internalized the DDE. Thus, half of our impersonal dilemmas required killing as a means, and half required killing as an unintended but foreseeable outcome. Many moral-judgment studies rely exclusively on subtle variations of the trolley problem (e.g., Hauser et al., 2007; Petrinovich, O’Neill, & Jorgensen, 1993; Valdesolo & DeSteno, 2006). 550 Volume 19—Number 6 Working Memory and Moral Judgment