Abstract
We report four experiments that investigate explicit reasoning and moral judgements. In each experiment, some subjects responded to the “footbridge” version of the trolley problem (which elicits stronger moral intuitions), whereas others responded to the “switch” version (which elicits weaker moral intuitions). Experiments 1–2 crossed the type of trolley problem with four reasoning conditions: control, counter-attitudinal, pro-attitudinal, and mixed reasoning (both types of reasoning). Experiments 3–4 examine whether moral judgements vary based on (a) when reasoners engage in counter-attitudinal reasoning, (b) when they make the moral judgement, and (c) by the type of moral dilemma. These two experiments comprised five conditions: control (judgement only), delay-only (2-minute wait then judgement), reasoning-only (reasoning then judgement), reasoning-delay (reasoning, then 2-minute delay, then judgement), and delayed-reasoning (2-minute delay, then reasoning, then judgement). These conditions were crossed with the type of trolley problem. We find that engaging in some form of counter-attitudinal reasoning led to less typical judgements (regardless of when it occurs), but this effect was mostly restricted to the switch version of the dilemma (and was strongest in the reasoning-delay conditions). Furthermore, neither pro-attitudinal reasoning nor delayed judgements on their own impacted subjects’ judgements. Reasoners therefore seem open to modifying their moral judgements when they consider opposing perspectives but might be less likely to do so for dilemmas that elicit relatively strong moral intuitions.
Keywords
Philosophers have long debated whether moral judgements are driven by emotional intuitions (i.e., the emotivist or intuitionist view; Hume, 1777/1960) or by explicit logic and reasoning (i.e., the rationalist view; Kant, 1785/1959). Emotional intuitions are posited to occur automatically, without much need for explicit processing (Smith & DeCoster, 2000), whereas reasoning is believed to be a more controlled and deliberative process that can be readily verbalised (Evans, 2003, 2008; Rips, 2001; Sloman, 1996). Many prominent philosophers have been rationalists (e.g., Socrates, Plato, Descartes, and Kant) and have thus held that people are rational agents who can objectively analyse the most relevant factors, weigh their importance, and subsequently make a calculated choice (Dworkin, 1986; Haidt, 2001; Hauser et al., 2007). However, other renowned scholars (e.g., Hume) have challenged this idea and argued that moral judgements are predominantly driven by emotional impulses.
Over the past 60 years, a large body of empirical research in the cognitive and social sciences has helped to inform this debate. Contrary to the once popularly held view, this work has shown that humans often fail to behave rationally (Kahneman & Tversky, 1979; Tversky & Kahneman, 1974, 1981). Moreover, many of our choices are often affected by factors of which reasoners are not explicitly aware (Nisbett & Schachter, 1966; Nisbett & Wilson, 1977; Zajonc, 1980). For example, decision outcomes can be strongly influenced by the manner in which a scenario is framed (Kahneman & Tversky, 1984; Tversky & Kahneman, 1974), the size of the group the reasoner is a part of (Darley & Latane, 1968), and even the clothes the reasoner is wearing (Slepian et al., 2015). Furthermore, many of our choices seem to occur automatically, without explicit deliberation (Nisbett & Wilson, 1977). Recent work has also shown that even philosophers, who undergo many years of rigorous training in reasoning, show many of these same biases (Horvath & Wiegmann, 2022; Schwitzgebel & Cushman, 2012, 2015; Wiegmann et al., 2020). In addition, related work has shown that explicitly encouraging subjects to engage in objective reasoning often fails to produce judgements that differ from subjects’ typical choices (Kneer et al., 2022; Kneer & Machery, 2019). Taken together, these findings challenge the rationalist perspective and call into question the extent to which explicit reasoning affects our daily choices.
In line with this idea, theoreticians have posited that moral judgements are primarily driven by moral intuitions, which are emotional intuitions about moral judgements (Greene, 2009; Greene & Haidt, 2002; Haidt, 2001, 2007). Under this view, moral dilemmas give rise to moral intuitions, and reasoners subsequently generate explicit reasons to justify those intuitions. Thus, the reasoning process is posited to occur after a judgement has been implicitly made (Haidt, 2001, 2007).
Some evidence for this hypothesis can be found by the occurrence of moral dumbfounding, wherein reasoners express strong conviction that their positions are driven by objective reasoning but are unable to explicitly justify the reasons behind their moral judgements (Haidt et al., 1993, 2000; Haidt & Hersh, 2001; McHugh et al., 2017). Furthermore, although reasoners are typically unaware of many of the factors that impact their judgements, to explain their behaviour, they often construct post hoc narratives, which are based on erroneous attributions and illusory correlations (Hauser et al., 2007; Nisbett & Wilson, 1977; Schachter & Singer, 1962; Wheatley & Haidt, 2005). Critically, because reasoning is posited to primarily be an explicit process (Rips, 2001; Sloman, 1996), if moral judgements were mainly driven by reasoning, people should be able to accurately verbalise the reasons behind their choices, and thus many of the phenomenon described here (e.g., moral dumbfounding) should rarely occur.
Conceptually related work has also shown that when dealing with issues that are important to the reasoner’s world view, they tend to primarily engage in reasoning that justifies their initial positions, and in doing so, mostly focus on supporting evidence while disregarding contradictory information (formally known as the confirmation bias; Kuhn, 1989, 1991; Kunda, 1990; Nickerson, 1998; Wason, 1960). Accordingly, this type of reasoning rarely results in reasoners changing their initial position and often leads to polarised views (Haidt, 2001, 2007; Lord et al., 1979).
On the contrary, reflecting and reasoning about various other choices that appear to be mainly driven by automatic, intuitive processes (e.g., taste preferences) can sometimes result in reasoners making less typical judgements (Wilson et al., 1984; Wilson & Schooler, 1991). Thus, given that reasoning can help people to override some automatic choices, it is possible that it can also be used to override an emotional or moral intuition. Although the research on moral decision making is quite extensive, much of this work has not directly examined how moral judgements are directly affected by explicit reasoning.
For example, various recent studies have examined the relationship between various factors that are related to reasoning and moral judgements (e.g., Bago & De Neys, 2019; Białek & De Neys, 2017; Gamez-Djokic & Molden, 2016; Gürçay & Baron, 2017; Lee & Holyoak, 2020; Parker & Finkbeiner, 2020; Patil et al., 2021). However, these studies have not manipulated (or even directly investigated) how explicitly reasoning about a given moral dilemma affects reasoners’ subsequent judgements.
Moral intuitions and reasoning
When considering this relationship, it is important to note that some dilemmas might elicit stronger moral intuitions than others. For example, consider the footbridge and switch versions of the trolley problem (Foot, 1967; Thomson, 1985): In each version, there is a runaway trolley that is headed towards five people who will be killed if it continues on its course, and one person must be sacrificed to save the others. These dilemmas differ by the type of hypothetical action that the subject must take to save the group of people. In the footbridge version of the dilemma, a large man is standing over a footbridge and the subject must physically push him onto the trolley’s path; in the switch version, the subject must pull a lever, which will divert the trolley to a different set of tracks and onto the path of another person. Although the outcomes in these dilemmas are identical, subjects often endorse actions that lead to different outcomes in each dilemma. For the switch version of the trolley problem, subjects typically favour the utilitarian outcome (i.e., sacrificing one person to save five), whereas for the footbridge version, they tend to endorse the non-utilitarian outcome (for a review of this work see Railton, 2014).
One explanation for these different choices is that the footbridge version of the trolley problem elicits a stronger moral intuition than the switch version (Greene, 2007; Horne & Powell, 2016). In line with this idea, extensive work has found suggestive evidence that judgements for the footbridge version of the trolley problem are more strongly driven by automatic, moral intuitions, whereas judgements for the switch version are more strongly impacted by deliberative, explicit processes (Cushman et al., 2006; Greene, 2003, 2007, 2014, 2017; Greene et al., 2001, 2004, 2008). 1
Given these findings, one possibility is that the impact of reasoning on moral judgements varies based on the strength of the corresponding moral intuition. In line with this idea, one of the few extant studies on how reflection about a moral dilemma impacts judgements found that subjects who explicitly reflected on different versions of the trolley problem made slightly more typical judgements for the footbridge version (i.e., showed less support for the utilitarian outcome) than those who did not reflect on these dilemmas (Schwitzgebel & Cushman, 2015). However, reflection had no impact on subjects’ judgements for the switch version of the trolley problem.
One possibility is that reasoners have greater motivation to justify stronger moral intuitions than weaker ones. Hence, when a moral intuition is relatively strong, reasoning might primarily function to justify that intuition. On the contrary, when a moral intuition is relatively weak, reasoning might be less motivated and more objective, and may therefore help reasoners evaluate the dilemma and make a choice that is based on explicit reasoning (as proposed by rationalists).
However, it is important to note that Schwitzgebel and Cushman (2015) did not directly manipulate reasoning in their study. It is thus possible that a different set of results might have been observed under a more direct or explicit reasoning manipulation. In one such study, Paxton et al. (2012, Experiment 2) presented subjects a dilemma pertaining to incest between two siblings (see Haidt et al., 2000), along with an argument in favour of this outcome; because subjects typically oppose the siblings’ actions (Haidt, 2001, 2007; Haidt et al., 2000); these arguments can be thought of as counter-attitudinal. For some subjects, this argument was designed to be inductively weak, and for others, it was designed to be inductively strong. In addition, some subjects judged the moral acceptability of the siblings’ actions immediately after reading the presented argument, whereas others were asked to reflect on this argument for 2 minutes before making this judgement.
Paxton et al. (2012) found that subjects who were presented the strong argument judged the siblings’ actions as more morally acceptable than subjects who were presented the weak argument, but this effect was only observed when subjects were encouraged to reflect on the argument for 2 minutes before making a judgement. Thus, under certain conditions, counter-attitudinal reasoning seems to lead to less typical moral judgements. However, an important open question is whether this phenomenon depends on the strength of the moral intuition that the corresponding dilemma elicits. That is, does reflection and reasoning affect all moral judgements in the same manner, or does this effect vary based on the strength of the moral intuition that the dilemma elicits?
Delaying reasoning and moral judgements
It is also important to note that that when an explicit moral judgement is made can vary; reasoners could make an explicit judgement immediately after encountering a moral dilemma or they could wait to make this judgement (as in the delayed judgement conditions in Experiment 2 of Paxton et al., 2012). Similarly, reasoners can reason about a moral dilemma immediately after encountering the dilemma or they can wait to reason about it.
This point is important because in many domains that involve moral judgements (e.g., military, medicine, politics), there can be delays between when people reflect and reason about a given dilemma and when they make an explicit decision about it. For example, consider a scenario in which a medical doctor encounters a moral dilemma about a given medical procedure. This doctor might hold off on making a decision about the dilemma for several weeks. During this period, they may also wait to explicitly think or reason about the dilemma for several days. Furthermore, once the doctor has reasoned about the dilemma, they may still wait several more days before making an explicit decision about the dilemma.
During each of these time periods, various factors might influence reasoners’ judgements. For instance, the strength of a moral intuition might decrease when a judgement is delayed, which might lead to a less typical decision than when reasoners make an explicit judgement immediately after encountering the dilemma. In line with this idea, one study found that subjects made less typical judgements about the footbridge version of the trolley problem (i.e., were more likely to endorse the utilitarian outcome) when they engaged in a distractor task during a 3-minute delay (compared to subjects who made their decision right away; Ham & van den Bos, 2010). However, because this study did not include a direct reasoning manipulation (as in Paxton et al., 2012, Experiment 2), it does not address the question of whether delaying when people reason about a moral dilemma affects their judgements nor does it address how reasoning might be related to making a delayed moral judgement.
Counter-attitudinal reasoning and moral intuitions
One possibility is that reasoners attempt to approach moral dilemmas objectively but are implicitly motivated to make their moral intuitions logically coherent (particularly when these intuitions are relatively strong). Indeed, reasoners typically believe that they are objective in their analyses of a given situation and that their corresponding views are a byproduct of careful reasoning and reflection (Haidt & Hersh, 2001; Robinson et al., 1995). Hence, when a moral intuition occurs, it may lead to “tunnel logic,” wherein reasoners unknowingly fail to explicitly consider opposing perspectives, and instead focus on finding a logical justification for their intuitions, which results in an increasingly narrow perspective. Reasoners may therefore become fixated on the reasons that support their intuitions, which might make it difficult to consider opposing perspectives.
Given enough time to explicitly reason about a moral dilemma, reasoners might thus thoroughly justify their intuitions, which may result in polarised judgements. The further along a reasoner gets into the justification process, the more difficult it might be to override a moral intuition through reason and reflection. Once this justification process has run its course, if reasoners are unable to find a reasonable counterpoint to their moral intuitions, they might come to believe that their moral judgements are objectively correct (given that they have made an honest effort to objectively reason about the dilemma). In such cases, reasoners might be reluctant to consider opposing views and may not be willing to change their positions.
It therefore seems to follow that there might be a critical period in which explicit reasoning can impact moral judgements. Once this critical period has passed and the justification process has run its course (e.g., after a delay), presenting the reasoner with counter-attitudinal arguments for private reflection may have little effect on their moral judgements.
However, if reasoners earnestly attempt to approach moral dilemmas objectively and simply fail to thoroughly consider counterpoints to their moral intuitions, introducing such arguments before this critical period has passed might lead to moral judgements that are less in line with reasoners’ moral intuitions (i.e., less typical judgements). In this latter case, reasoners might demonstrate more objective decision making that accords with the rationalist perspective, wherein moral judgements follow from reasoning (as opposed to reasoning being driven by moral intuitions).
Furthermore, it is possible that these predictions vary based on the strength of the moral intuition that the corresponding dilemma elicits in reasoners. For example, delaying a moral judgement might have a greater impact on a decision outcome for a dilemma that produces a stronger moral intuition than one that produces a weaker intuition.
This discussion thus leads to two primary hypotheses. First, counter-attitudinal reasoning can generally lead to less typical moral judgements. 2 Thus, reasoners who engage in counter-attitudinal reasoning should generally make less typical judgements than reasoners who do not engage in counter-attitudinal reasoning. Second, this prediction might vary based on the strength of the moral intuition that the dilemma produces. Thus, how counter-attitudinal reasoning affects the judgements that reasoners make for a dilemma that elicits weaker moral intuitions should be different than how it affects the judgements that reasoners make for dilemmas that elicit stronger moral intuitions.
With respect to the effects of delay, we have outlined three primary possibilities. First, if reasoners have enough time to explicitly justify their moral intuitions, asking them to engage in counter-attitudinal reasoning will not impact their moral judgements. For instance, consider a case in which reasoners encounter a moral dilemma and have a few minutes to think about it, during which they can justify their moral intuitions. If these reasoners were to then consider perspectives that run counter to their moral intuitions, this reasoning task should not affect their moral judgements. These reasoners should therefore make similar judgements to reasoners who do not engage in counter-attitudinal reasoning at all.
On the other hand, a second possibility is that if reasoners engage in counter-attitudinal reasoning before they have explicitly justified their moral intuitions (i.e., prior to the passage of the “critical period”), they will make less typical moral judgements. Hence, if reasoners encounter a moral dilemma and immediately engage in counter-attitudinal reasoning, they should make less typical moral judgements than reasoners who wait a short period to engage in counter-attitudinal reasoning, because these latter reasoners can justify their moral intuitions during this delay.
Alternatively, a third possibility is that if there is not a critical period to moral reasoning, yet reasoners attempt to make objective moral judgements and unintentionally neglect opposing perspectives, counter-attitudinal reasoning should lead to less typical moral judgements regardless of when it occurs. Under this hypothesis, reasoners who engage in counter-attitudinal reasoning at any point in the decision-making process should make similar moral judgements to one another, but less typical judgements than reasoners who do not engage in counter-attitudinal reasoning at all.
To test these hypotheses, we report four experiments that investigate the relationship between explicit reasoning and moral judgements across various decision-making conditions. Specifically, we examine whether different types of explicit reasoning affect moral judgements on different types of dilemmas (e.g., dilemmas that elicit weaker vs. stronger moral intuitions). In Experiments 1–2, we manipulated argument type, wherein some subjects reviewed counter-attitudinal arguments, others reviewed pro-attitudinal arguments, and some reviewed both types of arguments. In addition, in Experiments 3–4, we examine whether subjects’ moral judgements are impacted by when they reason about a moral dilemma (e.g., immediately vs. after a short delay), as well when they make an explicit judgement about it (e.g., immediately after reasoning vs. after a short delay).
Experiment 1
First, we examined whether the type of explicit reasoning that subjects engage in affects their moral judgements. Thus, we manipulated the type of reasoning task that subjects were asked to complete. Specifically, some subjects engaged in counter-attitudinal reasoning, others engaged in pro-attitudinal reasoning, and a third group (the mixed reasoning condition) engaged in both kinds of reasoning. This latter condition was included to examine whether considering both counter and pro-attitudinal arguments impacts subjects’ moral judgements. We also included a control condition, in which subjects were not given an explicit reasoning task and only read the dilemma that they were presented and made a moral judgement about it. The control condition provides a baseline measure of how these different types of reasoning affect subjects’ moral judgements.
To examine whether subjects’ judgements in these conditions vary based on the strength of the moral intuition that a dilemma elicits, these conditions were crossed with two versions of the trolley problem (footbridge vs. switch; Foot, 1967; Thomson, 1985); as a reminder, the footbridge version is posited to produce a stronger moral intuition (and emotional response) than the switch version (Greene, 2007; Greene et al., 2001, 2004). These dilemmas were used as a testbed for the aforementioned hypotheses because they have been extensively studied and reasoners’ choices for each dilemma are well documented, as are the neurological and emotional correlates of each dilemma (see Greene et al., 2001; also see Horne & Powell, 2016). How these dilemmas impact reasoners and how reasoners respond to them are thus well understood. These dilemmas are therefore ideal stimuli for investigating how explicit reasoning alters reasoners’ typical moral judgements and how this effect might vary as a function of the strength of the moral intuition that is produced by the moral dilemma.
Control subjects were asked to read the dilemma that they were presented and then rate the moral permissibility of the action that leads to the utilitarian outcome (i.e., sacrificing one individual to save five people). For subjects in the three reasoning conditions, after reading the presented dilemma, they were shown a list of arguments and were asked to indicate which ones they were considering.
For the footbridge version of the trolley problem, subjects generally endorse the non-utilitarian outcome, whereas for the switch version, they tend to endorse the utilitarian outcome (for a review, see Railton, 2014). Thus, the counter-attitudinal arguments for the footbridge version of the dilemma support the utilitarian outcome, whereas in the switch version, they support the non-utilitarian outcome (see Figure 1). In contrast, the pro-attitudinal arguments support the non-utilitarian outcome in the footbridge version of the dilemma and the utilitarian outcome in the switch version.
Arguments that were used in Experiments 1–2, which were against (a) and in favour of (b) the utilitarian outcome in the trolley problems.
If reasoners are indeed open to considering opposing perspectives but simply fail to consider them, then engaging in counter-attitudinal reasoning should lead to less typical moral judgements. Hence, subjects in the counter-attitudinal and mixed reasoning conditions should make less typical moral judgements than subjects in the control and pro-attitudinal reasoning conditions.
However, because people tend to seek out information that accords with their own views (Kuhn, 1989, 1991; Kunda, 1990; Nickerson, 1998; Wason, 1960), they might be implicitly motivated to consider positions that justify their moral intuitions. In such cases, reasoners might primarily engage in pro-attitudinal reasoning. Given that this strategy might be more typical than counter-attitudinal reasoning, it is possible that encouraging subjects to engage in pro-attitudinal reasoning might further convince them that their moral intuitions are correct, which might thus result in more polarised moral judgements than subjects in the other conditions.
On the contrary, it is possible that subjects are already engaging in this type of reasoning on their own, and thus an explicit pro-attitudinal reasoning task might be completely redundant and have no effect on moral judgements. Finally, we note that if any of these predictions depend on the strength of the moral intuition that is produced by a given dilemma, an interaction should be observed between the decision-making conditions and the type of dilemma that subjects are presented.
Method
Subjects
A total of 802 subjects were recruited from Prolific (www.prolific.co) and were paid .55¢ for their participation, which amounts to an average hourly wage of $9.48. This and all other experiments reported in this article were approved by the institutional review board at Syracuse University. Informed consent was obtained from all subjects who participated in this and all other experiments reported in this article.
Design
Subjects were randomly assigned to one of eight conditions; footbridge conditions: control (n = 101), counter-attitudinal reasoning (n = 92), pro-attitudinal reasoning (n = 97), and mixed attitudinal reasoning (n = 123); switch conditions: control (n = 101), counter-attitudinal reasoning (n = 94), pro-attitudinal reasoning (n = 94), and mixed attitudinal reasoning (n = 100).
Materials and procedure
This experiment was completed online. All stimuli were presented on a computer screen and responses were entered with the computer’s keyboard and mouse or touchpad. In each phase of the experiment, subjects could move on by clicking on a “continue” button.
In the first phase of the experiment, all subjects read the moral dilemma that they were assigned (i.e., either the footbridge or switch version of the trolley problem; Foot, 1967; Thomson, 1985). When subjects were ready to proceed, the dilemma was removed from the screen and instructions for the next phase of the experiment were presented.
Reasoning task
Subjects in the reasoning conditions were then presented a list of eight arguments. Subjects in the counter-attitudinal reasoning conditions were presented counter-attitudinal arguments, subjects in the pro-attitudinal reasoning conditions were presented pro-attitudinal arguments, and subjects in the mixed reasoning conditions were presented both types of arguments. For subjects in the mixed reasoning conditions, the arguments that they were presented was randomised, subject to the constraint that four were pro-attitudinal arguments and four were counter-attitudinal arguments.
Subjects in the reasoning conditions were instructed to read the presented arguments and indicate any that they were considering in their upcoming decision; to select an argument, subjects clicked on a box that was next to each argument. The order that these arguments were presented was randomised for each subject in all of the reasoning conditions.
Moral judgements
In the last phase of the experiment, all subjects were asked to rate the moral permissibility of the action that leads to the utilitarian outcome in the presented dilemma (i.e., sacrificing one individual to save five others). This judgement was made using a 7-point Likert-type scale (1 = completely morally impermissible, 7 = completely morally permissible). Next, the screen was cleared, and subjects were asked to indicate how confident they were in their choice. This decision was made on a 5-point Likert-type scale, with higher numbers indicating greater confidence. Subjects in the control conditions made these two choices after they indicated that they had read the presented dilemma (i.e., after the first phase of the experiment).
Results and discussion
Moral judgements
Figure 2a shows subjects’ mean moral permissibility ratings partitioned by decision-making condition and type of dilemma. A two-way analysis of variance (ANOVA), with decision-making condition and type of dilemma as between-subject factors, revealed a main effect of type of dilemma on subjects’ moral judgements, F(1, 794) = 120.093, p < .001, MSE = 2.90,
Subjects’ mean moral permissibility (a) and confidence judgements (b) on each version of the trolley problem partitioned by condition in Experiment 1.
To further investigate this interaction, two sets of least significant difference (LSD) post hoc comparisons were conducted, wherein we examined whether subjects’ moral permissibility ratings differed among the decision-making conditions for each version of the trolley problem. For the footbridge version of the trolley problem, no statistically reliable differences were observed in subjects’ moral judgements among the decision-making conditions (all ps > .110, all ds < .148). However, for the switch version of the trolley problem, subjects in counter and mixed attitudinal reasoning conditions made less typical moral judgements than subjects in the control and pro-attitudinal reasoning conditions (all ps < .019, all ds > .330). However, no differences in moral judgements on the switch version of the trolley problem were observed between subjects in the counter and mixed attitudinal reasoning conditions (p = .578).
Confidence judgements
Figure 2b shows subjects’ mean confidence judgements partitioned by decision-making condition and type of dilemma. A two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors, found a main effect of decision-making condition on subjects’ confidence judgements, F(3, 794) = 8.829, p < .001, MSE = 1.527,
To examine the main effect of decision-making condition further, we conducted an LSD post hoc comparison, which revealed that subjects in the mixed attitudinal reasoning conditions were more confident in their judgements than subjects in the other decision-making conditions (all ps < .004, all ds > .281). No other differences in subjects’ confidence judgements were observed among the decision-making conditions (all ps > .446, all ds < .077).
Number of arguments
In addition, we examined whether subjects in the reasoning conditions differed in the number of arguments that they selected during the reasoning task. A two-way ANOVA, with type of reasoning (counter-attitudinal vs. pro-attitudinal vs. mixed attitudinal reasoning) and type of dilemma as between-subject factors, found no main effect of type of dilemma and no interaction (both ps > .601). However, a main effect of type of reasoning was found, F(2, 594) = 7.610, p = .006, MSE = 4.047,
To remind the reader, subjects in the pro-attitudinal reasoning conditions reviewed arguments that were in line with the typical judgement that subjects typically make in the dilemma that was presented and subjects in the mixed attitudinal reasoning conditions reviewed half of these arguments. Thus, subjects who were presented at least some pro-attitudinal arguments indicated that they considered a greater number of arguments than subjects who only reviewed counter-attitudinal arguments.
Finally, to examine whether there was a difference between the pro and counter-attitudinal arguments that subjects in the mixed attitudinal reasoning conditions selected, we conducted a paired-samples t-test. This analysis revealed that subjects in the mixed attitudinal reasoning conditions selected a greater number of pro-attitudinal arguments (M = 2.120, SE = .093) than counter-attitudinal arguments (M = 1.625, SE = .101), t(199) = 3.449, p < .001, SE = .143, d = .244. Thus, subjects in the mixed attitudinal reasoning condition seemed to consider more pro-attitudinal arguments than counter-attitudinal arguments.
Summary
The present findings show that whether reasoning impacts moral judgements depends on the type of reasoning in which reasoners engage. Specifically, the results demonstrate that engaging in some form of counter-attitudinal reasoning (counter and mixed attitudinal reasoning conditions) leads to less typical moral judgements in general (i.e., compared to the control condition) and can also lead to less typical moral judgements than engaging in pro-attitudinal reasoning. Critically, however, these effects were only observed on the switch version of the trolley problem, as the type of reasoning that subjects engaged in did not impact their moral judgements on the footbridge version of the dilemma.
Furthermore, pro-attitudinal reasoning did not impact subjects’ moral judgements on either version of the trolley problem. Thus, we did not find evidence that engaging in pro-attitudinal reasoning impacts the reasoners’ moral judgements or their confidence in those judgements. One possibility is that reasoners already engage in pro-attitudinal reasoning on their own, such that they primarily search for reasons that support their moral intuitions. For this reason, having subjects engage in pro-attitudinal reasoning might be redundant, and as a result, this type of task may not produce a notable impact on their moral judgements.
Experiment 2
To bolster our confidence in the findings from Experiment 1, we conducted Experiment 2 as a direct replication. Thus, the design, materials, and procedures in the present experiment were identical to Experiment 1.
Method
Subjects
A total of 803 subjects were recruited from Prolific (www.prolific.co) and were paid .55¢ for their participation.
Design
Subjects were randomly assigned to one of eight conditions; footbridge conditions: control (n = 102), counter-attitudinal reasoning (n = 91), pro-attitudinal reasoning (n = 116), and mixed attitudinal reasoning (n = 99); switch conditions: control (n = 98), counter-attitudinal reasoning (n = 96), pro-attitudinal reasoning (n = 101), and mixed attitudinal reasoning (n = 100).
Results and discussion
Moral judgements
Figure 3a shows subjects’ mean moral permissibility ratings partitioned by decision-making condition and type of dilemma. A two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors, revealed a main effect of type of dilemma on subjects’ moral judgements, F(1, 795) = 186.72, p < .001, MSE = 2.70,
Subjects’ mean moral permissibility (a) and confidence judgements (b) on each version of the trolley problem partitioned by condition in Experiment 2.
To investigate this interaction further, two sets of LSD post hoc comparisons were conducted, in which we examined whether subjects’ moral permissibility ratings differed among the decision-making conditions for each version of the trolley problem. For the footbridge version of the trolley problem, subjects in the counter-attitudinal reasoning conditions made less typical moral judgements than subjects in all of the other decision-making conditions (all ps < .014, all ds > .335); no other statistically reliable differences were observed in subjects’ moral judgements on the footbridge version of the trolley problem among the other decision-making conditions (all ps > .698, all ds < .052).
For the switch version of the trolley problem, subjects in counter-attitudinal and mixed reasoning conditions made less typical moral judgements than subjects in the control and pro-attitudinal reasoning conditions (all ps < .015, all ds > .342). However, no differences in moral judgements on the switch version of the trolley problem were observed among any of the other decision-making conditions (all ps > .752, all ds < .046).
Confidence judgements
Figure 3b shows subjects’ mean confidence judgements partitioned by decision-making condition and type of dilemma. A two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors, found a marginal effect of decision-making condition on subjects’ confidence judgements, F(1, 795) = 3.82, p = .091, MSE = 1.56,
To examine the marginal effect of decision-making condition further, we conducted exploratory post hoc comparisons with Bonferroni corrections. Although subjects in the counter-attitudinal reasoning conditions reported numerically lower confidence in their judgements than subjects in the other decision-making conditions, this difference was not statistically reliable (all ps > .139). No other differences were observed in subjects’ confidence judgements among the decision-making conditions (all ps > .949).
Number of arguments
We also examined whether subjects in the reasoning conditions differed in the number of arguments that they selected during the reasoning task. A two-way ANOVA, with type of reasoning (counter-attitudinal vs. pro-attitudinal vs. mixed reasoning) and type of dilemma as between-subject factors, found no main effect of type of dilemma and no interaction (both ps > .136). However, a main effect of type of reasoning was found, F(2, 597) = 3.977, p = .019, MSE = 4.121,
We also note that a paired-samples t-test once again revealed that subjects in the mixed attitudinal reasoning conditions selected a greater number of pro-attitudinal arguments (M = 2.000, SE = .096) than counter-attitudinal arguments (M = 1.568, SE = .104), t(198) = 2.938, p = .004, SE = .145, d = .208. These results thus suggest that subjects in the mixed attitudinal reasoning condition considered a greater number of pro-attitudinal arguments than counter-attitudinal arguments.
Summary
The primary results from Experiment 2 mostly replicate those from Experiment 1, as subjects who viewed counter-attitudinal arguments (counter and mixed reasoning conditions) made less typical moral judgements than subjects in the control and pro-attitudinal reasoning conditions on the switch version of the trolley problem. Furthermore, as in Experiment 1, this effect depended on the type of dilemma that subjects were presented, as the mixed reasoning condition did not impact subjects’ moral judgements on the footbridge version of the trolley problem. However, it is important to note that unlike in Experiment 1, subjects in the counter-attitudinal reasoning condition made less typical moral judgements on the footbridge version of the trolley problem than subjects in the other decision-making conditions (i.e., showed more support for the action that led to the utilitarian outcome); this same trend was observed in Experiment 1, although that effect was smaller than in the present experiment. Counter-attitudinal reasoning therefore led to less typical moral judgements on both types of dilemmas, whereas this effect was restricted to the switch version of the dilemma for the mixed reasoning condition. In contrast, pro-attitudinal reasoning had no impact on subjects’ moral judgements on either version of the trolley problem (as in Experiment 1).
Experiment 3
The findings from Experiments 1–2 show that engaging in counter-attitudinal reasoning can lead to less typical moral judgements for dilemmas that elicit relatively weak moral intuitions. However, two outstanding questions are (1) whether when reasoners engage in counter-attitudinal reasoning and (2) when they make a subsequent decision affects their moral judgements. To address these questions, Experiment 3 consisted of five decision-making conditions: (1) control, (2) delay-only, (3) reasoning-only, (4) reasoning-delay, and (5) delayed-reasoning. Because our primary question centres on counter-attitudinal reasoning and moral judgements, the mixed and pro-attitudinal reasoning conditions were not included in the present experiment. 3 As in Experiments 1–2, the decision-making conditions were crossed with the footbridge and switch versions of the trolley problem (Foot, 1967; Thomson, 1985).
The control conditions in this experiment were identical to the control conditions in Experiments 1–2. In addition, the counter-attitudinal reasoning conditions were similar to the counter-attitudinal reasoning conditions in Experiments 1–2. One difference was that in the present experiment, subjects in the reasoning conditions were asked to review nine arguments instead of eight (see Figure 4); we added one argument in favour of the utilitarian outcome for the footbridge version of the dilemma and one in favour of the non-utilitarian outcome for the switch version. This change was made in an attempt to bolster the potency of the counter-attitudinal reasoning task.
Arguments that were used in Experiments 3 and 4, which were against (a) and in favour of (b) the utilitarian outcome in the trolley problems.
Subjects in the delay-only condition waited 2 minutes after reading the dilemma before making a choice about it. For subjects in the three reasoning conditions, after reading the dilemma, they were shown a list of counter-attitudinal arguments and were asked to indicate which ones they were considering.
For the reasoning-only condition, this list of arguments was presented immediately after subjects read the dilemma (as in the counter-attitudinal reasoning conditions in Experiments 1–2), which was followed by them rating the moral permissibility of the action that leads to the utilitarian outcome in the dilemma. For the reasoning-delay condition, after reviewing these arguments, subjects waited 2 minutes before making this choice. For the delayed-reasoning condition, subjects waited 2 minutes after reading the dilemma before they were presented these arguments, after which they rated the moral permissibility of the action that leads to the utilitarian outcome in the dilemma.
These conditions were carefully selected to test how explicit reasoning, along with when reasoning occurs and when the corresponding moral judgement is made affect subjects’ moral choices. This design thus allows us to isolate these factors and examine their individual and cumulative effects on moral judgements. We specifically focus on addressing four questions: (1) whether counter-attitudinal reasoning impacts subjects’ moral judgements (i.e., whether the results from Experiments 1–2 replicate), (2) whether delaying the explicit choice that subjects make impacts their moral judgements, (3) whether varying when subjects review counter-attitudinal arguments (immediately vs. after a delay) leads to different moral judgements, and (4) whether any of these outcomes vary based on the strength of the moral intuition that a given dilemma produces.
If delaying moral judgements and explicit reasoning each lead to less typical judgements, then subjects in the delay- and reasoning-only conditions should make judgements about the dilemmas that are somewhat incongruent with those of control subjects (since the control condition provides a baseline of subjects’ judgements about the dilemmas).
Furthermore, if reasoners are more objective about ideas that contradict their intuitions early on, before those intuitions have been justified, then subjects who review counter-attitudinal arguments right away (reasoning-only and reasoning-delay conditions) should make less typical judgements than subjects who wait to review these arguments (delayed-reasoning condition). Finally, we note that if any of these predictions depend on the strength of the moral intuition that is produced by a given dilemma, an interaction should be observed between the decision-making conditions and the type of dilemma that subjects are presented.
Method
Subjects
A total of 1,001 subjects were recruited for this study from Prolific (www.prolific.co) and were paid .55¢ for their participation.
Design, materials, and procedure
Subjects were randomly assigned to 1 of 10 conditions; footbridge conditions: control (n = 107), delay-only (n = 98), reasoning-only (n = 107), reasoning-delay (n = 96), and delayed-reasoning (n = 94); switch conditions: control (n = 105), delay-only (n = 97), reasoning-only (n = 105), reasoning-delay (n = 94), and delayed-reasoning (n = 98).
Figure 5 illustrates the order in which subjects in each decision-making condition completed the tasks in this experiment. The first and third phase of the experiment (i.e., reading the dilemma and making a decision about it) were identical to Experiments 1–2, as were the instructions for the reasoning tasks.
Flowchart of the procedure for subjects in each decision-making condition in Experiments 3 and 4. Arrows indicate the temporal order in which each task was completed.
Counter-attitudinal reasoning task
After the first phase of the experiment, subjects in the reasoning-only and reasoning-delay conditions were presented a list of nine counter-attitudinal arguments. The order that these arguments were presented was randomised for each subject in all of the reasoning conditions.
Delayed reasoning and judgements
For subjects in the delayed-reasoning conditions, before being presented these arguments (but after reading the presented dilemma), they were instructed to wait 2 minutes before proceeding and that they would be notified when it was time to continue. At the 2-minute mark, these subjects were presented a “continue” button and a prompt, which thanked them for their patience and asked them to click on the “continue” button when they were ready to proceed, after which they completed the counter-attitudinal reasoning task.
The delay and instructions were identical for all conditions in which a delay was included. For subjects in the reasoning-delay conditions, this delay occurred after they completed the counter-attitudinal reasoning task. For subjects in the delay-only condition, this delay occurred after they indicated that they had read the presented dilemma.
Finally, all subjects made a decision about the presented dilemma and rated their confidence in this judgement (as in Experiments 1–2).
Results and discussion
Moral judgements
Figure 6a shows subjects’ mean moral permissibility ratings partitioned by decision-making condition and type of dilemma. For the primary analysis, we conducted a two-way analysis of variance (ANOVA) with decision-making condition (control vs. delay-only vs. reasoning-only vs. reasoning-delay vs. delayed-reasoning) and type of dilemma (footbridge vs. switch) as between-subject factors and subjects’ moral permissibility ratings as the dependent measure.
Subjects’ mean moral permissibility (a) and confidence judgements (b) on each version of the trolley problem partitioned by condition in Experiment 3.
This analysis revealed a main effect of type of dilemma, F(1, 991) = 112.420, p < .001, MSE = 2.69,
To further investigate this interaction, we conducted two sets of LSD post hoc comparisons, wherein we examined whether subjects’ moral permissibility ratings differed among the decision-making conditions on each version of the trolley problem. These analyses revealed that for the footbridge version of the trolley problem, no statistically reliable differences were observed in subjects’ moral judgements among any of the decision-making conditions (all ps > .064, all ds < .289). However, for the switch version of the trolley problem, subjects in all of the reasoning conditions (i.e., reasoning-only, reasoning-delay, delayed-reasoning) made less-typical moral judgements, and thus rated the action that leads to the utilitarian outcome as less morally permissible than subjects in the control and delay-only conditions (all ps < .015, all ds > .343).
Furthermore, for the switch version of the trolley problem, subjects in the reasoning-delayed condition (M = 3.936, SE = .185) made less-typical judgements than subjects in the reasoning-only condition (M = 4.381, SE = .149; p = .049, d = .258); subjects in the reasoning-delayed condition also made marginally less typical judgements on the switch version of the trolley problem than subjects in delayed-reasoning condition (M = 4.378, SE = .185; p = .055, d = .257). No other differences were observed among the decision-making conditions on the switch version of the trolley problem (all ps > .625, all ds < .074).
Confidence judgements
As a secondary analysis, we examined whether subjects’ confidence in their choices varied as a function of the condition that they were assigned to. Figure 6b shows subjects’ mean confidence judgements partitioned by decision-making condition and type of dilemma. To examine this question, we used a two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors and subjects’ confidence ratings as the dependent measure.
This analysis revealed a main effect of decision-making condition, F(4, 991) = 4.803, p < .001, MSE = 1.403,
Number of arguments
Finally, to examine whether subjects in the reasoning conditions differed in the number of arguments that they selected during the counter-attitudinal reasoning task, we conducted a two-way ANOVA, with reasoning condition and type of dilemma as between-subject factors and the number of arguments that subjects selected as the dependent measure. However, this analysis revealed no main effects and no interaction (all ps > .225).
Summary
These results demonstrate that counter-attitudinal reasoning can indeed impact moral judgements and lead to less typical choices. However, this effect seems to depend on the type of dilemma that subjects are presented, as it was only observed on the switch version of the trolley problem. This latter finding suggests that counter-attitudinal reasoning might help subjects make more objective moral judgements for dilemmas that produce weaker moral intuitions but might be less effective for dilemmas that elicit stronger moral intuitions.
Furthermore, delayed judgements on their own (i.e., delay-only conditions) did not impact subjects’ moral choices on either version of the trolley problem. However, when judgements were delayed after subjects engaged in counter-attitudinal reasoning (i.e., reasoning-delay conditions), subjects made the least typical judgements on the switch version of the trolley problem. These judgements were less typical than those from subjects in the reasoning-only condition and were marginally less typical than those from subjects in the delayed-reasoning condition. This finding is in line with those from Paxton et al. (2012) and suggest that delaying a judgement after reasoners engage in counter-attitudinal reasoning, which gives them additional time to reflect on those arguments, can lead to less typical judgements. Nevertheless, it is important to note that these effects were relatively small, and thus should be interpreted cautiously.
It is also important to note, however, that we did not find evidence for a “critical period” in reasoning about moral dilemmas, as no differences in moral judgements (on either version of the trolley problem) were observed between the reasoning-only and the delayed-reasoning conditions. Moreover, for the switch version of the trolley problem, subjects in all of the reasoning conditions made less-typical judgements than subjects in the control and delay-only condition. This finding suggests that regardless of when subjects engage in counter-attitudinal reasoning, they seem to consider and factor those arguments into their judgement. In line with this idea, no differences among the reasoning conditions were found in the number of arguments that subjects indicated that they were considering during the counter-attitudinal reasoning task, and simply engaging in this task led to subjects reporting lower confidence in their moral judgements. Taken together, these results suggest that counter-attitudinal reasoning, regardless of when in the decision-making process it occurs, seems to impact reasoners, and moves them towards more objective behaviours (e.g., having less confidence in their choices, although this shift might not always correspond to reasoners’ moral judgements).
Experiment 4
Given the small effect sizes among the reasoning conditions in Experiment 3, to increase our confidence in these results, we conducted a follow-up experiment. Experiment 4 was a direct replication of Experiment 3. These two experiments thus had identical designs, materials, and procedures.
Method
Subjects and design
Subjects were recruited from Prolific (www.prolific.co) and were paid .55¢ for participating. A total of 1,029 subjects participated in this experiment. Subjects were randomly assigned to 1 of 10 conditions; footbridge conditions: control (n = 110), delay-only (n = 95), reasoning-only (n = 109), reasoning-delay (n = 98), and delayed-reasoning (n = 101); switch conditions: control (n = 109), delay-only (n = 96), reasoning-only (n = 108), reasoning-delay (n = 100), and delayed-reasoning (n = 103).
Results and discussion
Moral judgements
Figure 7a shows subjects’ mean moral permissibility ratings partitioned by decision-making condition and type of dilemma. A two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors, revealed a main effect of type of dilemma on subjects’ moral judgements, F(1, 1019) = 110.520, p < .001, MSE = 2.45,
Subjects’ mean moral permissibility (a) and confidence judgements (b) on each version of the trolley problem partitioned by condition in Experiment 4.
To further investigate this interaction, we conducted two sets of LSD post hoc comparisons, wherein we examined whether subjects’ moral permissibility ratings differed among the decision-making conditions on each version of the trolley problem. These analyses revealed that for the footbridge version of the trolley problem, no statistically reliable differences were observed in subjects’ moral judgements among any of the decision-making conditions (all ps > .099, all ds < .220). However, for the switch version of the trolley problem, subjects in all of the reasoning conditions made less-typical moral judgements than subjects in the control and delay-only conditions (all ps < .002, all ds > .350).
Furthermore, for the switch version of the trolley problem, subjects in the reasoning-delayed condition (M = 3.730, SE = .146) made less-typical judgements than subjects in the reasoning-only (M = 4.343, SE = .137) and delayed-reasoning conditions (M = 4.272, SE = .136; both ps < .009, both ds > .380). It is important to note that this difference between the reasoning-delayed and the delayed-reasoning conditions was marginal in Experiment 3, but was indeed statistically reliable in the current experiment. No other differences in the switch version of the trolley problem were observed among the decision-making conditions (all ps > .663, all ds < .057).
Confidence judgements
As in Experiment 3, we also examined whether subjects’ confidence in their choices differed based on their condition. Figure 7b shows subjects’ mean confidence judgements partitioned by decision-making condition and type of dilemma. A two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors, revealed a main effect of decision-making condition on subjects’ confidence judgements, F(4, 1019) = 4.585, p = .001, MSE = 2.041,
Number of arguments
We also examined whether subjects in the reasoning conditions differed in the number of arguments that they selected during the counter-attitudinal reasoning task. However, a two-way ANOVA, with decision-making condition and type of dilemma as between-subject factors, revealed no main effects and no interaction (all ps > .312).
Summary
This experiment was a direct replication of Experiment 3. Accordingly, the present results are in line with and replicate those from Experiment 3. These findings thus add further support to the reliability of the primary findings from Experiment 3 and help to affirm its tentative conclusions.
General discussion
In this article, we report four experiments that examine the relationship between explicit reasoning and moral judgements. In Experiments 1 and 2, we directly manipulated the type of explicit reasoning that subjects were asked to engage in. These two experiments comprised four decision-making conditions (control vs. counter vs. pro vs. mixed attitudinal reasoning), which were crossed with two versions of the trolley problem (footbridge vs. switch; Foot, 1967; Thomson, 1985). In Experiments 3 and 4, we investigated (a) whether engaging in counter-attitudinal reasoning affects subjects’ moral judgements, (b) whether this effect varies based on when this reasoning occurs, and (c) whether such effects depend on the strength of the moral intuition that a dilemma elicits. To investigate these questions, we compared five decision-making conditions (control vs. delay-only vs. reasoning-only vs. reasoning-delay vs. delayed-reasoning) and crossed them with two versions of the trolley problem that were used in Experiments 1–2.
Counter-attitudinal reasoning
Across these experiments, several common themes arose. First, reviewing at least some counter-attitudinal arguments (i.e., counter and mixed attitudinal reasoning conditions), regardless of when this occurred, led to less-typical judgements for the switch version of the trolley problem (Experiments 1–4). These results demonstrate that reasoners are open to changing or modifying some moral judgements when they are introduced to counter-attitudinal arguments.
One reason that counter-attitudinal arguments might lead to less-typical moral judgement is that reasoners may implicitly search for reasons that support their moral intuitions, and in doing so, may inadvertently overlook opposing positions. Once these countervailing ideas are explicitly considered, reasoners might come to hold a more balanced position, which may lead to a more objective judgement.
In line with this explanation, Experiments 1 and 2 demonstrate that reasoners seem to have a bias or preference for considering arguments that align with their typical moral judgements, as subjects selected a greater number of pro-attitudinal arguments than counter-attitudinal arguments. It is also important to note that presenting subjects only counter-attitudinal arguments tended to decrease their confidence in their judgements, whereas introducing them to at least some pro-attitudinal arguments increased their confidence in their moral judgements. One possibility is that when reasoners are presented arguments that support their intuitions, it helps to further affirm their belief that those intuitions are objectively correct. On the contrary, because reasoners seem to focus on arguments that support their moral intuitions, presenting them counter-attitudinal arguments might make such arguments more salient and thereby draw the reasoner’s attention to arguments that they might not be considering or factoring into their moral judgements. As a result, reasoners’ confidence in their moral judgements may decrease.
Thus, a core contribution of this article is that we demonstrate that explicit counter-attitudinal reasoning leads to less-typical moral judgements. Moreover, our results suggest that this effect emerges because when left to their own accord, reasoners do not adequately consider positions that run counter to their intuitions.
Type of dilemma
Nevertheless, the present results point to clear limitations in the impact of counter-attitudinal reasoning, as these types of arguments were mostly ineffective at modifying subjects’ judgements on the footbridge version of the dilemma (Experiments 1, 3, and 4). To remind the reader, the footbridge version of the trolley problem is posited to elicit a stronger moral intuition than the switch version (Greene, 2007), and has also been shown to produce stronger emotional responses (Horne & Powell, 2016). Thus, one possible explanation for the present findings is that for moral dilemmas that elicit relatively weak moral intuitions, reasoners are open to changing their position when they are introduced to arguments that run counter to their typical choices. However, the extent to which such arguments attenuate moral judgements seems to vary based on the type of dilemma that is encountered, wherein the impact of counter-attitudinal arguments might be somewhat weak for dilemmas that elicit relatively strong moral intuitions.
Judgements for these latter types of moral dilemmas thus seem to be more robust to being modified through explicit reasoning. This point notwithstanding, in Experiment 2, subjects in the counter-attitudinal reasoning condition made less-typical judgements on the footbridge version of the trolley problem (i.e., showed greater support for the action that led to utilitarian outcome) than subjects in the other conditions (all ps < .014). Furthermore, although this result was not observed in Experiments 1, 3, and 4, we do note that the same trend occurred in two of these studies, wherein subjects in the counter-attitudinal reasoning conditions made (numerically) less-typical judgements for the footbridge version of the dilemma than subjects in the control conditions; when comparing the reasoning-only condition to the control condition, this difference was a nonsignificant trend in Experiments 2 (p = .190) and 3 (p = .225). Taken together, these results suggest that counter-attitudinal reasoning might indeed impact reasoners’ judgements on moral dilemmas that elicit relatively strong moral intuitions, but that this impact is weaker and less reliable than for dilemmas that elicit weaker moral intuitions.
Thus, another critical takeaway from the present findings is that reasoners might be more biased when they encounter dilemmas that produce stronger moral intuitions. To further reduce such biases, it might therefore be necessary to more strongly encourage reasoners to consider opposing perspectives than the reasoning manipulations that were used in the experiments reported in this article.
In line with this idea, some studies that have failed to show that reasoning leads to more objective judgements have often used relatively passive reasoning manipulations (e.g., asking subjects to think about a given dilemma carefully; Horvath & Wiegmann, 2022; Kneer et al., 2022; Kneer & Machery, 2019; Schwitzgebel & Cushman, 2015). In contrast, studies that have used more involved manipulations that provide subjects more structured and guided reasoning tasks have shown that reasoning can indeed lead to more objective decision outcomes, wherein subjects make less-typical choices (e.g., Ervin & Corral, 2022a, 2022b; Paxton et al., 2012; cf. Rai & Holyoak, 2010).
One possibility is that passive reasoning manipulations are somewhat ineffective because reasoners seem to have a bias towards focusing on arguments that correspond with their intuitions and preferred outcomes (as demonstrated in Experiments 1 and 2). Given the importance of objective reasoning and decision-making in many domains (e.g., science, medicine, education, philosophy, and law), an important future direction will be for researchers to identify the factors that help reasoners better consider opposing perspectives and reason about them objectively. Furthermore, because reasoners seem to struggle to objectively reason about moral dilemmas that elicit strong moral intuitions, it is important for future research to identify the types of dilemmas that are more likely to produce such responses. This knowledge can help people better identify which types of dilemmas others are more likely to approach and reason about objectively and those towards which they may be more biased.
Pro-attitudinal reasoning
Another important common theme that was observed in our experiments is that regardless of which type of the dilemma was presented (or when the reasoning occurred, i.e., right after viewing the dilemma or after a short delay), pro-attitudinal reasoning does not seem to impact subjects’ moral judgements (Experiments 1 and 2; also see the supplemental experiment in the online Supplementary Material). This finding is somewhat counterintuitive given that subjects were explicitly presented arguments that are in line with the judgement that is typically made on the corresponding dilemma. One explanation for this finding is that because reasoners seem to prefer arguments that align with their intuitions, they think and reason about similar types of ideas on their own. This explanation is further supported by the finding showing that subjects had a stronger preference for pro-attitudinal arguments than counter-attitudinal arguments. Hence, pro-attitudinal reasoning might be mostly redundant with how people typically reason about moral dilemmas, which might thus lead to such reasoning tasks having little to no impact on moral judgements.
Delays in moral reasoning and judgements
One noteworthy finding from Experiments 3 and 4 is that when subjects engaged in counter-attitudinal reasoning (i.e., right after viewing the dilemma versus after a short delay) did not seem to lessen its impact. That is, subjects who engaged in counter-attitudinal reasoning, regardless of when it occurred, made less typical judgements than subjects who did not engage in counter-attitudinal reasoning (particularly on the switch version of the trolley problem, as this effect was not observed on the footbridge version of the dilemma). These results suggest that so long as subjects are introduced to perspectives that run counter to their moral intuitions, they are likely to be impacted by them, which leads to less-typical judgements, particularly for dilemmas that elicit weaker moral intuitions.
Another important finding from Experiments 3 and 4 is that subjects in the reasoning-delay conditions made less typical judgements on the switch version of the dilemma than subjects in the other decision-making conditions. To explain the implications of this finding, it is important to remind the reader of two points. First, counter-attitudinal reasoning was involved in both the reasoning-only and delayed-reasoning conditions. Second, the delay-only and reasoning-delay conditions both involved a delayed judgement. For these reasons, the aforementioned findings cannot be attributed to counter-attitudinal reasoning or delayed judgements alone or to their pairing, but rather to their specific temporal ordering, wherein counter-attitudinal reasoning occurs prior to the delayed judgement (also see Paxton et al., 2012).
One possible explanation for these results is that when subjects engage in counter-attitudinal reasoning and then their moral judgement is delayed, they use this interim to think more carefully about the opposing perspectives that they were presented, which leads to less-typical judgements. In contrast, when reasoners engage in counter-attitudinal reasoning and make a judgement about the dilemma right away, even if there was a delay before this reasoning task, reasoners have less time to carefully think about the dilemma than when this delay occurs after they engage in counter-attitudinal reasoning. Thus, spending more time thinking about counter-attitudinal arguments might lead to less-typical judgements.
Delay-only
Another important finding to note is that in Experiments 3 and 4 (also see the supplemental experiment in the online Supplementary Material), delayed judgements on their own did not seem to affect subjects’ decisions on either version of the trolley problem, as the judgements of subjects in the delay-only conditions were no different than those from subjects in the control conditions. These findings are in line with those from Ham and van den Bos (2010), as they showed that subjects who made delayed moral judgements made similar choices on the footbridge version of the trolley problem as subjects who made a decision right after viewing the dilemma.
The present results might be taken to suggest that there is not a critical period in moral reasoning and that delayed moral judgements on their own do not impact a person’s subsequent choice one way or the other. However, we caution against drawing such a conclusion, as it is important to note that in the present set of experiments that included some form of delayed judgements, the interval between when subjects were presented a moral dilemma and when they made a decision was only 2 minutes. In contrast, in the real world, the intervals between when people encounter a moral dilemma and subsequently reason and make a decision about it could be quite lengthy (e.g., several hours, days, weeks). It is thus possible that longer delays in reasoning may indeed reveal a critical period for moral reasoning, wherein considering opposing perspectives is not particularly effective at tempering reasoners’ typical judgements. Similarly, it is also possible that longer intervals between when a moral dilemma is encountered and when a decision is made might reveal that such delays impact peoples’ moral judgements. Although these are certainly important questions, they are beyond the scope of this article, and we thus leave them for future research to address.
Limitations and future directions
It is important to note that this article only included the footbridge and switch versions of the trolley problem (Foot, 1967; Thomson, 1985). To remind the reader, these dilemmas were specifically chosen because they have been extensively studied and are thus very well understood, both in terms of how they affect reasoners and how reasoners respond to them. As a result, these dilemmas allowed us to directly examine (a) how reasoners’ typical moral judgements are impacted by explicitly reasoning about a given moral dilemma and (b) how this impact might vary based on the strength of the moral intuition that is produced by the dilemma. It will thus be important for future research to examine whether the present results extend to other types of moral dilemmas and more generally to other judgement and decision-making scenarios.
Further work will also be required to better understand why explicit counter-attitudinal reasoning leads to less typical moral judgements for dilemmas that seem to produce weaker moral intuitions but is not as effective for dilemmas that elicit stronger moral intuitions. Although this result was rather consistent across our experiments, we note that we did not directly manipulate the strength of the moral intuition that was produced by each moral dilemma, and instead relied on well-established findings over the past three decades, which have provided strong support for this conclusion (e.g., Greene, 2007; Greene et al., 2001, 2004; Horne & Powell, 2016). Nevertheless, it is an open question as to why the observed interaction between counter-attitudinal reasoning and type of moral dilemma occurs.
As noted earlier, one possibility is that reasoners are more averse to considering opposing perspectives for dilemmas that elicit stronger moral intuitions, because the strong emotion produced by the dilemma increases reasoners’ motivation to search for reasons that justify their intuitions. Consequently, reasoners might be more likely to neglect opposing perspectives. Another related possibility is that stronger moral intuitions lead reasoners to more strongly believe that their intuitions are correct, which makes corresponding judgements more robust to change through explicit reasoning than judgements that stem from weaker moral intuitions. Although this hypothesis is intuitive, it is important to note that stronger moral intuitions do not necessarily lead to more strongly held judgements, as the two are logically independent and are not always strongly associated (Horne & Powell, 2016). Addressing these questions is certainly important, but this work is beyond the scope of this article, and thus we leave it for future research to address.
Conclusion
Reasoning has traditionally been thought to be central to moral judgements (Dworkin, 1986; Kant, 1785/1959). However, considerable work over the past seven decades has called this idea into question (Haidt, 2001, 2007). This work has shown that humans are susceptible to many reasoning biases (Kahneman & Tversky, 1979; Kuhn, 1989, 1991; Kunda, 1990; Lord et al., 1979; Tversky & Kahneman, 1974, 1981) and are often unaware of the factors that influence their judgements (Nisbett & Schachter, 1966; Nisbett & Wilson, 1977; Zajonc, 1980). Nevertheless, much of this work has not directly examined the relationship between explicit reasoning and moral judgements.
This article thus presents much needed research on this topic. Our primary finding is that introducing reasoners to arguments that run counter to their preferred outcome can lead to less-typical moral judgements (particularly when time passes between when counter-attitudinal reasoning occurs and the subsequent decision). However, this effect appears to be considerably weaker for dilemmas that are posited to elicit relatively strong moral intuitions. These results therefore demonstrate that reasoning can certainly play an important role in moral judgements and can lead to more objective decision-making, but that its impact can be somewhat limited for certain types of moral dilemmas. Future work will be necessary to examine whether the present results extend to other types of moral dilemmas across various domains (e.g., medicine, military, education, politics, and economics).
Supplemental Material
sj-docx-1-qjp-10.1177_17470218231179685 – Supplemental material for The effects of explicit reasoning on moral judgements
Supplemental material, sj-docx-1-qjp-10.1177_17470218231179685 for The effects of explicit reasoning on moral judgements by Daniel Corral and Abraham M Rutchick in Quarterly Journal of Experimental Psychology
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Data accessibility statement
All materials and data are available upon email request to the corresponding author.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
