Pushing Moral Buttons: The Interaction Between Personal Force and Intention in Moral Judgment

In some cases people judge it morally acceptable to sacrifice one person’s life in order to save several other lives, while in other similar cases they make the opposite judgment. Researchers have identified two general factors that may explain this phenomenon at the stimulus level: (1) the agent’s intention (i.e. whether the harmful event is intended as a means or merely foreseen as a side-effect) and (2) whether the agent harms the victim in a manner that is relatively “direct” or “personal.” Here we integrate these two classes of findings. Two experiments examine a novel personalness/directness factor that we call personal force , present when the force that directly impacts the victim is generated by the agent’s muscles (e.g., in pushing). Experiments 1a-b demonstrate the influence of personal force on moral judgment, distinguishing it from physical contact and spatial proximity. Experiments 2a-b demonstrate an interaction between personal force and intention, whereby the effect of personal force depends entirely on intention. These studies also introduce a method for controlling for people’s real-world expectations in decisions involving potentially unrealistic hypothetical dilemmas.

What explains this pattern of judgment?Neuroimaging (Greene, et al., 2001(Greene, et al., , 2004)), lesion (Ciaramelli et al., 2007;Koenigs et al., 2007;Mendez et al., 2005), and behavioral (Bartels, 2008;J. Greene et al., 2008;Valdesolo & DeSteno, 2006) studies indicate that people respond differently to these two cases because the action in the footbridge dilemma elicits a stronger negative emotional response.But what features of this action elicit this response?Recent studies implicate two general factors.First, following Aquinas (unknown/2006), many appeal to intention and, more specifically, the distinction between harm intended as a means to a greater good (as in the footbridge dilemma) and harm that is a foreseen but "unintended" side-effect of achieving a greater good (as in the switch dilemma) (Cushman et al., 2006;Hauser et al., 2007;Mikhail, 2000;Schaich Borg et al., 2006).Second, many studies appeal to varying forms of "directness" or "personalness," including physical contact between agent and victim (Cushman et al., 2006), the locus of intervention (victim vs. threat) in the action's underlying causal model (Waldmann & Dieterich, 2007), whether the action involves deflecting an existing threat (Greene et al., 2001), and whether the harmful action is mechanically mediated (Moore et al., 2008;Royzman & Baron, 2002).The aim of this paper is to integrate these two lines of research.
We present two experiments examining a directness/personalness factor that we call personal force.An agent applies personal force to another when the force that directly impacts the other is generated by the agent's muscles, as when one pushes another with one's hands or with a rigid object.Thus, applications of personal force, so defined, cannot be mediated by mechanisms that respond to the agent's muscular force by releasing or generating a different kind of force and applying it to the other person.Although all voluntary actions that affect others involve muscular contractions, they do not necessarily involve the application of personal force to another person.For example, firing a gun at someone or dropping a weight onto someone by releasing a lever do not involve the application of personal force because the victims in such cases are directly impacted by a force that is distinct from the agent's muscular force, i.e. by the force of an explosion or gravity.The cases of direct harm examined by Royzman and Baron (2002) are not so direct as to involve the application of personal force.The direct/indirect distinction described by Moore and colleagues (2008) is similar to the distinction drawn here between personal and impersonal force, but Moore and colleagues do not systematically distinguish between physical contact and personal force.
Experiments 1a-b aim to document the influence of personal force, contrasting its effect with those of physical contact (1a-b) and spatial proximity (1a) between agent and victim.Experiment 1a also introduces a method for controlling for effects of unconscious realism, i.e. a tendency to unconsciously replace a moral dilemma's unrealistic assumptions with more realistic ones.("Trying to stop a trolley with a person is unlikely to work.")Experiments 2a-b examine the interaction between personal force and intention.More specifically, we ask whether the effect of personal force depends on intention and vice versa.

Experiment 1a
We compared four versions of the footbridge dilemma to isolate the effects of spatial proximity, physical contact, and personal force on moral judgments concerning harmful actions.We also tested the unconscious realism hypothesis by controlling for subjects' real-world expectations.
The mean age was 31.Subjects were recruited anonymously in public venues in New York City and Boston.Subjects were paid $3.

Design, materials, and procedure
Subjects responded to one of four versions of the footbridge dilemma in a between-subject design, indicating the extent to which the proposed action is "morally acceptable."In the standard footbridge dilemma (n = 154, Figure 1a), the agent (named Joe) may save the five by pushing the victim off the footbridge using his hands.This action involves spatial proximity, physical contact, and personal force.In the remote footbridge dilemma (n = 82, Figure 1d), Joe may drop the victim onto the tracks using a trap door and a remote switch.This action involves none of the three aforementioned factors.The footbridge pole dilemma (n = 72, Figure 1b) is identical to the standard footbridge dilemma except that Joe uses a pole rather than his hands to push the victim.This dilemma involves spatial proximity and personal force without physical contact.The footbridge switch dilemma (n = 160, Figure 1c) is identical to the remote footbridge dilemma except that Joe and the switch are adjacent to the victim.This dilemma involves spatial proximity without physical contact or personal force.Comparing remote footbridge to footbridge switch isolates the effect of spatial proximity.Comparing standard footbridge to footbridge pole isolates the effect of physical contact.Comparing footbridge switch to footbridge pole isolates the effect of personal force.The text of the standard footbridge dilemma is as follows: An empty runaway trolley is speeding down a set of tracks toward five railway workmen.There is a footbridge above the tracks in between the runaway trolley and the five workmen.On this footbridge is a railway workman wearing a large, heavy backpack.If nothing is done, the trolley will proceed down the main tracks and cause the deaths of the five workmen.(See diagram below) It is possible to avoid these five deaths.Joe is a bystander who understands what is going on and who happens to be standing right behind the workman on the footbridge.Joe sees that he can avoid the deaths of the five workmen by pushing the workman with the heavy backpack off of the footbridge and onto the tracks below.The trolley will collide with the workman, and the combined weight of the workman and the backpack will be enough to stop the trolley, avoiding the deaths of the five workmen.But the collision will cause the death of the workman with the backpack.
Note: Joe cannot avoid the deaths of the five workmen by jumping himself because he is not heavy enough to stop the trolley.There is also not enough time to remove the backpack from the workman.
Is it morally acceptable for Joe to push the workman off of the footbridge in order to avoid the deaths of the five workmen, causing the death of the single workman instead?Subjects answered (YES/NO) and rated the moral acceptability of the action on a nine-point scale.The above text was accompanied by a diagram (Figure 1a).Similar text and diagrams (Figures 1c-d and 3) were used for other dilemmas, with changes reflecting the experimental manipulations.Complete materials are available at [url].
The instructions acknowledged that the dilemmas were not necessarily realistic and requested that subjects "suspend disbelief."Data from 31 (of 664) subjects who reported being unable/unwilling to suspend disbelief ("conscious realists") were excluded form analysis, as were data from 10 subjects reporting confusion.
To control for unconscious realism, we asked subjects (after they responded to the dilemma) to report on their real-world expectations concerning the likely consequences of Joe's actions.Subjects estimated the likelihood (0-100%) that the consequences of Joe's action would be (a) as described in the dilemma (five lives saved at the cost of one), (b) worse than this, or (c) better than this.These estimates (respectively labeled PLAN, WORSE, and BETTER) were modeled as covariates.The predictive value of these variables indicates the extent to which subjects' judgments may reflect unconscious realism.
Data were analyzed using a general linear model.Here and in Experiment 2a, the three "realism covariates" and gender were included as first-order covariates and allowed to interact with the dilemma variable.
In Experiment 2a these factors were allowed to interact with both main effects and the interaction of interest.Because the realism covariates are likely correlated, this analysis is adequate to control for the their collective effects but inadequate to resolve their respective contributions.
(See Figure 2) There was a significant main effect of WORSE (F(1, 417) 12 = 5.80, p = .02)with actions expected to be less successful eliciting lower moral acceptability ratings, consistent with unconscious realism.There were no significant effects of PLAN, BETTER, gender, or higher order covariates (p > .05).These results indicate that harmful actions involving personal force are judged to be less morally acceptable.Moreover, they suggest that spatial proximity and physical contact between agent and victim have no effect and that a previously reported effect of physical contact (Cushman et al., 2006) is in fact an effect of personal force.In all four of the dilemmas examined in this study, the harmful event is intended as a means to achieving the agent's goal, raising the possibility that the effect of personal force is limited to cases in which the harm is intended as a means.Experiments 2a-b examine the interaction between personal force and intention.

Experiment 1b
To ensure that the results concerning personal force and physical contact observed in Experiment 1a generalize to other contexts, we conducted an additional experiment using a different set of moral dilemmas, as well as a different rating scale.

Experiment 2a
This experiment examined the independent effects of personal force and intention and, most critically, their interaction, by comparing four dilemmas using a 2 (personal force absent vs. present) x 2 (means vs. side-effect) design.

Method
Methods follow Experiment 1a unless otherwise noted.

Design, materials, and procedure
Each subject responded to one of four dilemmas.In the loop dilemma (Hauser et al., 2007;Mikhail, 2000;Thomson, 1985;Waldmann & Dieterich, 2007), Joe may save the five by turning the trolley onto a looped side-track that reconnects with the main track at a point before the five people (n = 152, Figure 3a).There is a single person on the sidetrack who will be killed if the trolley is turned, but who will prevent the trolley from looping back and killing the five.Here the victim is harmed as a means (i.e.intentionally), but without the application of personal force.
The loop weight dilemma (Hauser et al., 2007;Mikhail, 2000) is identical to the loop dilemma except that a heavy weight positioned behind the victim on the side-track, rather than the victim, stops the trolley (n = 74, Figure 3b).Here the victim is killed as a side-effect (i.e.without intention) and, again, without the application of personal force.In the obstacle collide dilemma, the victim is positioned on a high and narrow footbridge in between Joe and a switch that must be hit in order to turn the trolley and save the five (n = 70, Figure 3c).To reach the switch in time, Joe must run across the footbridge, which will, as a side-effect, involve his colliding with the victim, knocking him off the footbridge and to his death.Thus, this dilemma involves personal force, but not intention.The obstacle push dilemma (n = 70) is identical to the obstacle collide dilemma except that Joe must push the victim out of the way in order to get to the switch.
Although the victim is not used to stop the trolley, Joe performs a distinct body movement (pushing) that is both harmful and necessary for the achievement of the goal.Thus, this dilemma involves the application of personal force that is intentional.
There was no significant effect of BETTER or other higher order covariates (p > .05).

Discussion
In two sets of experiments, harmful actions were judged to be less morally acceptable when the agent applied personal force to the victim.In Experiments 1a-b the effect of personal force was documented and distinguished from effects of physical contact (Cushman et al., 2006) and spatial proximity (1a only), which were not significant.Experiments 2a-b revealed that personal force interacts with intention, such that the personal force factor only affects moral judgments of intended harms, while the intention factor is enhanced in cases involving personal force.Put simply, something special happens when intention and personal force co-occur.
(We note that all key results held using categorical (YES/NO) judgments when they were collected.) In Experiments 2a-b, personal force exhibited no effect in the absence of intention, a striking result in light of Experiments 1a-b and previous work.In Experiment 2a, the action in the obstacle collide dilemma was judged to be as acceptable as those in the loop, and loop weight dilemmas despite the fact that obstacle collide, unlike the other two dilemmas, involves direct harm (Moore et al., 2008;Royzman & Baron, 2002), physical contact (Cushman et al., 2006), harm not caused by the deflection of an existing threat (Greene et al., 2001), and an alteration of the victim's causal path (Waldmann & Dieterich, 2007).(One may interpret Waldmann & Dieterich as assuming that victim interventions are necessarily intended, in which case this result is consistent with their theory.)Experiment 2b showed that this finding generalizes to several additional dilemma contexts, strongly suggesting that the effect of personal force is limited to cases involving harm as a means.
Experiments 2a and 2b also demonstrate that the effect of the intention factor on moral judgment is enhanced in cases involving personal force, and Experiment 2a found no effect of intention in the absence of personal force, suggesting that intention operates only in conjunction with other factors such as, but not necessarily limited to, personal force.Our finding of equivalence between the loop (intentional harm) and loop weight (harmful side-effect) dilemmas directly contradicts some earlier findings (Hauser et al., 2007;Mikhail, 2000), 2 but is consistent with other earlier findings (Waldmann & Dieterich, 2007).
Following Waldmann & Dieterich, we attribute the effects observed by Hauser et al. (2007) and Mikhail (2000) to a confound whereby the loop dilemma, but not the loop weight dilemma, refers to the victim as a "heavy object."("There is a heavy object on the side track… The heavy object is 1 man…" vs. "There is a heavy object on the side track… There is 1 man standing on the side track in front of the heavy object…").
The statistical significance of the "unconscious realism" covariates included in Experiments 1a and 2a provides limited support for the unconscious realism hypothesis.This support is limited for at least two reasons.First, subjects' assessments of the likely real-world effects of the actions in question may be post-hoc rationalizations (Haidt, 2001).
Second, a correlation between real-world expectations and moral judgments is not sufficient to establish a causal relationship.
Nevertheless, these results indicate that effects of unconscious realism may be real and that researchers who use hypothetical cases to study decision-making should consider controlling for such effects as done here.
One might wonder why the actions judged to be more acceptable in Experiment 1a (footbridge switch and remote footbridge) received comparable ratings (~5) to the action judged to be less acceptable in Experiment 2a (obstacle push).First, in considering why the footbridge switch and remote footbridge dilemmas received relatively low ratings, we speculate that this may be due to the fact that the actions in these dilemmas involve dropping the victim onto the tracks, constituting an additional intentional harm (Mikhail, 2007).Second, in considering why the ratings for the obstacle push dilemma are relatively high, we suggest that this may be due to the fact that the action in the obstacle push dilemma, while involving a distinct body movement that is harmful and necessary for the achievement of the goal, does not involve using the victim, as in the four footbridge dilemmas.Each of these hypotheses will be explored in future work.
The latter hypothesis highlights more general open questions concerning the scope of agents' intentions (Bennett, 1995).In the obstacle push dilemma, the pushing is necessary, but the consequent harm, strictly speaking, is not.This observation raises parallel questions about more paradigmatic cases of intentional harm.For example, one might claim that even in the standard footbridge dilemma the harm is unintentional because the agent merely intends to use the victim's body to stop the trolley, harming him only as a foreseen side-effect of doing this.
These observations highlight the need for a theory of intentional event segmentation (Zacks & Tversky, 2001).
Other open questions concern the proper characterization of personal force: Must it be continuous (as in pushing), or may it be ballistic (as in throwing)?Is pulling equivalent to pushing?We acknowledge, more broadly, that the effects documented here under the rubric of "personal force" may ultimately be refined and reinterpreted.For example, alternative interpretations may focus on the potential for dynamic interaction between agent and victim.
Finally, we consider the significance of our finding that personal force and intention interact: Why is it that the combined presence of personal force and intention pushes our moral buttons?The codependence of these factors suggests a system of moral judgment that operates over an integrated representation of goals and personal forcerepresentations such as "goal-within-the-reach-of-muscle-force."In a general sense, this suggests a mechanism of moral judgment that is a species of embodied cognition (Gallese, Keysers, & Rizzolatti, 2004;Lakoff & Johnson, 1999;Prinz, 2002;Wilson, 2002).One natural source of such embodied goal representations is system of action planning that coordinates the application of personal force to objects to achieve goal-states for those specific objects.A putative sub-system of moral judgment, monitoring such action plans, might operate by rejecting any plan that entails harm as a goal-state (Mikhail, 2000(Mikhail, , 2007) ) to be achieved through the direct application of personal force.We propose this "actionplanning" account of the present results as an important area for further research.
At a more general level, the present study strongly suggests that our sense of an action's moral wrongness is tethered to its more basic motor properties, and specifically that the intention factor is intimately bound up with our sensitivity to personal force.This perspective contrasts with at least some versions of the "universal moral grammar" perspective (Hauser, 2006;Mikhail, 2000Mikhail, , 2007)), according to which the present moral judgments depend on goal representations of the kind one might find in a legal system, leaving little room for an 'embodied' representation involving personal force.It also presents a challenge to philosophical theories that endorse the doctrine of double effect (i.e. the intention factor) on the basis of its intuitive plausibility (Aquinas, unknown/2006;Fischer & Ravizza, 1992).Will they bless its shotgun marriage to a normatively ugly bride: the doctrine of personal force?

Figure 1 .
Figure 1.Diagrams for the (a) standard footbridge dilemma (physical

Figure 2 .
Figure 2. Results of Experiment 1: Moral acceptability ratings for four

Figure 3 .
Figure 3. Diagrams for the (a) loop dilemma (means, no personal force),

Figure 4 .
Figure 4. Results of Experiment 2: Moral acceptability ratings for four