The 'omission bias' refers to the tendency to judge harm resulting from direct action as worse than harm resulting from a failure to act (Baron & Ritov, 2004). Individuals generally favour omissions (such as letting someone die) over otherwise equivalent commissions (such as actively killing someone) (Ritov & Baron, 1990).
Several studies have reported this omission bias in decision making and judgment, suggesting that people typically find acts of omission less wrong than acts of commission when the outcome is the same (Baron, 1992; Cushman et al., 2006; Ritov & Baron, 1995; Royzman & Baron, 2002; Spranca et al., 1991). For example, people condemn actively poisoning a sports rival more than they condemn failing to inform the same rival that he is about to eat something poisonous (Spranca et al., 1991). People also condemn actively stealing $100 more than they condemn failing to return $100 that is accidentally obtained (Spranca et al., 1991). Even when the outcome is worse for omission, people still show a preference for omission over commission (Baron, 1992). The majority of participants indicate unwillingness to commit a single murder to save the lives of two other individuals (Baron, 1992).
The omission bias has also been used to explain why parents are reluctant to vaccinate their children (an act of commission), even when the risks of not vaccinating (an act of omission) are greater than the vaccination itself (Asch et al., 1994; Baron & Ritov, 1994; daCosta DiBonaventura & Chapman, 2008; Meszaros et al., 1996; Ritov & Baron, 1990). When presented with the hypothetical possibility of giving a child a vaccine, carrying a small risk of death, to protect against a disease, carrying a higher risk of death, most participants still opted against the vaccination (Ritov & Baron, 1990). Further study has demonstrated that the risk of death from the vaccine needs to be 40% (less than half) the risk of death from the disease before a majority of participants will vaccinate (Asch et al., 1994).
Several reasons have been proposed to account for omission bias: an exaggeration effect; loss aversion; normality bias; legality bias; naivety; actor presence; and a preference for indirect harm (Spranca et al., 1991). These are detailed in table 1.
There are three possible explanations for why this preference for indirect harm occurs (Royzman & Baron, 2002). First, direct harm may appear worse to others than indirect harm. Second, participants often believe that direct action is more likely to be harmful than indirect action, even when it has been specified that this is not the case. Third, participants see direct harm as involving more malicious intent than indirect harm, again even when it has been specified that this is not the case.
It remains uncertain which of these explanations are most important in driving the omission bias. One way to probe the reasons for the omission bias might be to examine a form of action that is not clearly omission or commission but which involves elements of both. Nearly all previous studies in this area have focused purely on comparisons between the extremes: complete omission and commission. Judgments of harm resulting from 'active omission', in which a protagonist is only indirectly involved in the causing of harm, were yet to be investigated. In this context, active omission refers to an individual manipulating a situation to elicit a harmful consequence without directly causing the resulting harm themselves. In this sense it is more direct than omission (failing to act at all), but less direct than commission (direct action).
By presenting a number of scenarios in which harm results from either omission, commission or active omission and asking participants to rate how wrong the actions are in each of these scenarios, as well as making a number of other moral judgments, it was possible to discover where active omission stands in relation to both omission and commission. With this information it would then be possible to infer which of the previous explanations for the occurrence of omission bias are supported by the addition of active omission and which cannot fully explain the phenomenon. For example, if it really is the case that omission bias is driven by loss aversion then active omission should be judged the same as commission as they would both be considered a loss rather than a missed gain. If directness is what matters then commission, the most direct, should be considered worse than active omission and active omission should be considered worse than omission, the least direct.
The first experiment consisted of six scenarios, each including three separate endings: one in which harm was caused by omission; another in which harm was caused by active omission; and another in which harm was caused by commission. In all the scenarios, the three endings' outcomes were the same and the actor's malicious intent towards a negative outcome was held constant throughout.
One of the scenarios, for example, describes a couple who have been married for 10 years. The husband (Ben) is filing for divorce and both partners are fighting for custody of their two young children. The wife (Claire) is bitter about the divorce and believes that even though Ben is a devoted father, he gave up on the marriage and so does not deserve custody of the children. She thinks that if the family judge believes he is unfit to care and provide for the children, she will be granted sole custody. In the omission ending, Claire's lawyer tells her that he has heard that Ben uses recreational drugs and drinks in excess. Claire knows Ben is responsible and would never do this, but she says nothing. Subsequently, the judge deems Ben to be incapable of providing a stable future for the two children and grants sole custody to Claire. In the commission ending, Claire lies to her lawyer, telling him that Ben regularly drinks in excess and uses recreational drugs. Claire's lawyer takes this information to the family court and subsequently, the judge deems Ben to be incapable of providing a stable future for the two children and grants sole custody to Claire. In the active omission ending, Claire invites Ben to a party which the lawyer is also attending. She pretends to make up with Ben and then encourages him to take illicit drugs while the lawyer is present and watching. Claire's lawyer takes this information to family court. Subsequently, the judge deems Ben unfit to care for the children and grants sole custody to Claire.
Participants who took part in this experiment were asked to judge the protagonist's actions for each ending. Intuition may suggest that the actions of the protagonist in the active omission endings would be rated more severely than those in the omission endings, but less severely than the actions of the commission endings. This seems logical considering active omission can be viewed as the midpoint between the two extremes of omission and commission. However, it was predicted that the actions of the protagonist in the active omission endings would in fact be deemed worse than both the omission and commission endings. Active omissions may not be as direct as commissions, but they could be considered more cruel and manipulative.
As well as replicating experiment 1 with different scenarios, the aim of experiment 2 was to compare endings with and without malicious intent. In the endings with malicious intent, the protagonist dislikes the victim and wants harm to be inflicted upon them. In the endings without malicious intent, the protagonist does not dislike the victim, but still either inflicts harm or allows harm to be inflicted for other reasons such as preoccupation with something at the time.
By comparing these two types of endings, it would be possible to determine whether being indifferent about someone but still causing/allowing harm to be inflicted upon them is deemed better or worse than disliking someone and wanting harm to be inflicted upon them.
The second experiment consisted of two scenarios, each including six separate endings: omission without malicious intent; omission with malicious intent; active omission without malicious intent; active omission with malicious intent; commission without malicious intent; and commission with malicious intent. Table 2 lists the defining features of these endings. Appendix A shows the two scenarios used in this experiment.
As with experiment 1, participants who took part in this experiment were asked to judge the protagonist's actions for each ending. It was predicted that the endings without malicious intent would be deemed worse than the equivalent endings with malicious intent. This is because in both the protagonist has the ability to stop the impending harm; however, the protagonist with malicious intent has reason to harm the victim/allow the victim to be harmed, while the protagonist without malicious intent does not. Thus, they may be considered crueler and more to blame because of their lack of motive.
Experiment 1 involved 63 students (55 female, mean age = 19.3 years, range = 18-25) and experiment 2 involved 67 students (49 female, mean age = 19.9, range = 18-32). None of the participants from experiment 1 took part in experiment 2. All participants provided informed consent and received course credit for their participation.
All participants completed the experiment online. Before the scenarios were presented, the participants were given instructions for what was required of them as well as the layout of the study. The benefit of their participation was explained to them as well as stating there were no apparent risks associated with participation. The participants were told that their responses were confidential and they could withdraw from the study at any time. It was stressed to participants that they were to "base ratings only on the information that is given and try not to make assumptions that go beyond what the scenario has stated".
Following the presentation of each scenario ending, participants judged the protagonists' actions according to the following questions: How much is the protagonist to blame for the harm that occurred? How responsible do you think the protagonist is for the harm that occurred? How much malicious intent was shown by the protagonist? How manipulative was the protagonist? How cruel was the protagonist? How direct was the protagonist in causing harm? How socially unacceptable was the behaviour of the protagonist? Answers for each question were provided using a seven point scale ranging from 0 (least severe) to 6 (most severe), as shown in Figure 1.
The order of the endings was counterbalanced to account for order effects.
The judgment ratings for both experiments were assessed using 3-way repeated measures analysis of variance (ANOVA) to assess the effects of scenario, ending and question as well as the interactions between these variables.
Figure 2 presents the mean moral judgment ratings for each of the three endings. A 6 (scenario) x 3 (ending) x 8 (question) repeated measures ANOVA demonstrated a main effect for ending (F(1.7, 103.1) = 186.4, p < .001). Commission endings were rated as worst (M = 5.39, SE = .06), followed by active omission (M = 4.88, SE = .07) and finally, omissions (M = 4.19, SE = .10).
LSD Pairwise comparisons revealed that individual differences between the endings were significant. Active omissions were rated significantly more harshly than omissions (MD = .69, SE = .06, p < .001) and commissions were rated significantly more harshly than active omissions (MD = .51, SE = .051, p < .001).
There was also a main effect of scenario (F(4.0, 250.3) = 6.0, p < .001) and a main effect of question (F(4.8, 299.1) = 20.2, p < .001). There were significant interactions of scenario with ending (F(7.4, 458.2) = 16.3, p < .001), scenario with question (F(15.6, 968.7) = 11.5, p < .001), ending with question (F(7.1, 442.3) = 46.0, p < .001) and scenario, ending and question (F(22.9, 1419.1) = 11.8, p < .001).
Figure 3 shows the judgment ratings for the ending types within each scenario. It is clear that omission was always rated as less severe than both commission and active omission. There is, however, a less clear difference between active omission and commission. LSD Pairwise comparisons revealed that commission was rated significantly more harshly than active omission in the 'marriage' scenario (MD = .66, SE = .12, p < .001), the 'domestic violence' scenario (MD = .71, SE = .09, p < .001), the 'custody' scenario (MD = .86, SE = .10, p < .001) and the 'house' scenario (MD = .91, SE = .09, p < .001). There was no significant difference between commission and active omission in the 'hospital' scenario (MD = .07, SE = .11, p = .53). In the 'ticket' scenario, however, active omission was rated significantly more harshly than commission (MD = .13, SE = .06, p = .03).
When looking at the responses to the specific questions asked about the 'ticket' scenario's endings, it appears the main reason the active omission ending was rated so harshly was due to the protagonist's manipulative nature. (see figure 4). A paired samples t-test revealed that for this scenario, the active omission ending was rated as significantly more manipulative than the commission ending (t(62) = 5.61, p < .001).
Looking at the average question ratings encompassing all the scenarios in figure 5, it can be seen that this high manipulation rating for active omission occurred in the other scenarios. LSD Pairwise comparisons of a 6 (scenario) x 3 (ending) repeated measures ANOVA found that the active omission endings were rated as more manipulative than the commission endings (MD = .23, SE = .09, p = .02). However, commissions were rated as more wrong (MD = .49, SE = .05, p < .001), malicious (MD = .14, SE = .07, p = .045), direct (MD = .99, SE = .09, p < .001), cruel (MD = .34, SE = .06, p < .001) and socially unacceptable (MD = .52, SE = .06, p < .001) than active omissions. The protagonists in the commission endings were also considered more to blame (MD = .97, SE = .07, p < .001) and more responsible (MD = .89, SE = .08, p <.001) than the protagonists in the active omissions endings.
Figure 5 shows that the greatest difference between commission and active omission was in relation to directness. The results suggest this directness was the cause of the protagonists being considered more responsible and blameworthy in the commission endings.
Figure 6 presents the mean moral judgment ratings for each of the six endings. A 2 (scenario) x 6 (ending) x 8 (question) repeated measures ANOVA showed a main effect for ending (F(3.2, 211.8) = 173.3, p < .001). Commissions with malicious intent were rated as worst (M = 4.65, SE = .12), followed by commissions without malicious intent (M = 4.34, SE = .11). Next was active omissions with malicious intent (M = 3.99, SE = .14), followed by active omissions without malicious intent (M = 3.56, SE = .12). Then omissions with malicious intent (M = 2.85, SE = .16) and finally, omissions without malicious intent (M = 1.39, SE =
There was no main effect of scenario ((F(1, 66) = 2.805, p = .099), however, there was a main effect of question (F(4.7, 310.7) = 30.8, p < .001). There were significant interactions of scenario with ending (F(3.9, 256.5) = 7.3, p < .001), scenario with question (F(3.4, 223.6) = 12.1, p < .001), ending with question (F(16.3, 1073.5) = 13.0, p < .001) and scenario, ending and question (F(17.7, 1166.2) = 3.3 p < .001).
Figure 7 shows the judgment ratings for each specific question. For nearly all the questions, there was a continued increase through the six endings, similar to that shown in figure 6. However, the 'malicious' question evoked different responses from most of the other questions. Omission with malicious intent was not rated significantly more harshly in terms of malicious intent than active omission without malicious intent (MD = .03, SE = .17,
.13). LSD Pairwise comparisons revealed that individual differences between all the endings were significant.
p = .86). Active omission with malicious intent was also not rated significantly more harshly in terms of malicious intent than commission without malicious intent (MD = .03, SE = .14, p = .83).
The findings from the first experiment reveal that active omission is rated more harshly than omission, but less harshly than commission. Participants clearly noted the manipulative nature of the active omission endings; however, in all other aspects commission was rated more harshly.
The 'ticket' scenario was the only scenario that did not follow this trend. The active omission ending was rated significantly more harshly than the commission ending. One possible explanation for this was that in the 'ticket' scenario the active omission ending was relatively farfetched (see Appendix B).The exaggeration effect explained in table 1 suggests that this could be the reason for the high active omission rating. The participants responded with greater affect because the harmful event had an abnormal cause.
The significant main effect of scenario can be explained by the lack of homogeneity between each scenario. For example, the severity of scenario varied from deceit to murder. The motives for harm within the scenarios also differed - from getting a train ticket to saving a daughter's life. However, with each of these scenarios, the results remained the same relative to each other. For both the higher and lower rated scenarios, commissions were rated most harshly and omissions were rated least harshly.
The findings from the second experiment support those from the first experiment. Again, active omissions were rated more harshly than omissions, but less harshly than commissions. In addition to this, it was found that within each of the three types of scenario endings, those with malicious intent were rated more harshly than those without malicious intent. As was previously stated, it could be argued that in the endings with malicious intent there is at least reason for what the protagonist is doing (the disliking of the victim) whereas in the endings without malicious intent there is not. This is clearly not how the participants interpreted the endings as the former was still rated more harshly. Regardless of the motives behind the actions, harm inflicted with malicious intent was considered worse than harm inflicted without malicious intent.
One of the most interesting findings was that shown in figure 7 in which omission with malicious intent was not rated significantly more harshly in terms of malicious intent than active omission without malicious intent and active omission with malicious intent was not rated significantly more harshly in terms of malicious intent than commission without malicious intent. This result gives even more weight to the findings from experiment 1 which suggest that commissions are judged more severely than active omissions; so much so that even when there are endings describing active omission with malicious intent and others describing commission without malicious intent, there is considered to be the same amount of malicious intent in both.
It is important to note than although in the endings without malicious intent there was no malicious intent per se, the actions in these endings could still be considered malicious (see Appendix A). Although the protagonists have no foreknowledge of the harm that is going to occur, their actions are still cruel. If this study were to be repeated, endings without malicious intent would ideally involve no maliciousness at all.
With the two experiments' results, it is possible to revisit the explanations of omission bias proposed in table 1 and explain why these findings show which explanations are most important in driving the omission bias. Take, for example, the exaggeration effect. It was previously stated that the active omission ending of the 'ticket' scenario was farfetched and so the exaggeration effect may account for the high judgment rating; however, within all the scenarios the active omission endings were as, if not more, abnormal than the commissions, yet the commissions were still judged more harshly. This suggests that the exaggeration effect cannot fully explain the omission bias.
Omission is considered a missed gain rather than a loss because the harm would have occurred regardless of whether the protagonist was there or not. The protagonist can only improve the situation rather than have a detrimental input. Commission, however, is considered a loss because the protagonist actually causes the harm to occur. This is also true of active omission, in which the situation is manipulated to cause the harm. However, commission was still judged more harshly than active omission. Therefore loss aversion cannot fully account for the omission bias.
Naivety cannot explain the omission bias as active omissions would rarely result from naivety. Therefore it would be expected that the active omission endings would be judged just as harshly as commissions, but this did not happen.
However, some of the possible explanations for the omission bias are supported by this study. Commissions are likely to be punished more severely than active omissions and active omission are likely to be punished more severely than omissions. Commissions were rated more harshly than active omissions, which were rated more harshly than omissions and therefore, the legality bias is supported by this study.
The normality bias is also supported by the findings of this study. Commissions were rated as more socially unacceptable than active omissions and active omissions were rated as more socially unacceptable than omissions. The normality bias accounts for this trend.
Finally, the fact that indirect harm is preferred to direct harm is supported by this study. Commission was rated most harshly overall and most direct, while omission was rated least harshly overall and least direct. Figure 5 shows that there was a great difference between the directness ratings of each of the ending types. It appears that directness made one of the largest impacts on participant's judgments that commissions are worse than active omissions.
This impact of directness has been noted in other research. One study using the famous trolley problem for example (Hauser et al., 2007) presented two different scenarios to participants. In one an individual throws a switch to divert a trolley which is about to kill five people into one person who dies. In the other, an individual throws a person in front of the trolley which is again about to kill five people. 85% of the participants stated it would be acceptable to throw the switch, while only 12% said it would be acceptable to throw the person in front of the trolley.
This preference for indirect harm over direct harm has serious implications. While the findings of Hauser et al. (2007) suggest that most people are not willing to cause harm
directly even if it is for the 'greater good', Milgram (1974) was also aware that people can commit atrocities if the harm they are inflicting is indirect. His prediction of his famous experiment stated the following: "Any force that is placed between the subject and the consequences of shocking the victim, any factor that will create distance between the subject and the victim, will lead to a reduction of strain of the participants and thus lessen disobedience." (Milgram, 1974, p. 121).
Future research should aim to set clearer boundaries between the three types of endings. Within this study, there was quite a range within each ending. For example, omissions in some scenarios were still more direct than others. In fact it could be argued that the picking up of the ticket in the omission ending of the 'ticket' scenario (Appendix B) is classified as 'theft by finding' and therefore legally commission. This, however, does not put the findings into jeopardy. It can still be concluded that the more direct the harm the worse it is considered and that this appears to be the most influential factor in causing the omission bias; so much so that it even outweighs the effect of malicious intent.