Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Contents lists available at ScienceDirect

Organizational Behavior and Human Decision Processes

journal homepage: www.elsevier.com/locate/obhdp

A dynamic perspective on moral choice: Revisiting moral hypocrisy

Stephanie C. Lin a, *, Dale T. Miller b
a
INSEAD, Singapore
b
Stanford University, United States

A R T I C L E I N F O A B S T R A C T

Keywords: We propose a dynamic model of moral decision making whereby people revise their morally relevant preferences
Moral hypocrisy as their options evolve. We employ this model to reinterpret prior findings on moral hypocrisy. In particular, we
Moral decision making revisit the finding that, when tasked to assign themselves and another person to tasks that differ in pleasantness,
Justifications
participants who “flip a coin” to determine the task allocation assign themselves to the preferable task more than
Prosocial behavior
Self-concept
fifty percent of the time. This result was originally thought to reveal that people will take moral credit for flipping
a coin while simultaneously harboring the intention of disregarding its outcome if it is negative. We suggest
instead that people flip the coin not with the intention of disregarding the outcome, but with the hope of
maximizing their self-interest without self-reproach (Studies 1 and 2); only when this outcome proves
unachievable do they resort to rationalizing their self-interested assignment (Studies 3 and 4). These findings
offer a novel perspective on the flexibility of moral decisions.

1. Introduction Thompson, & Chen, 2002; Batson, Thompson, Seuferling, Whitney, &
Strongman, 1999) to examine how one possible preference structure
Policy makers, organizations, and individuals often engage in prac­ functions in a two-stage process: In the first stage, people are primarily
tices that benefit both themselves and others. Environmentally sus­ motivated to achieve a self-interested outcome while simultaneously
tainable behavior benefits society but also provides economic benefits to being moral; only when that option becomes unavachievable do they
the organizations that implement them. Donating to a charitable orga­ shift to the second stage in which they recruit justification strategies to
nization benefits its cause but also elevates the reputation of the donor. move onto their next choice—self-interest. In doing so, we provide a
Organizations and individuals often engage in these behaviors touting dynamic perspective on moral choice that offers insight into people’s
their prosocial motivations. What happens when these behaviors can no hierarchical preference structure and moral flexibility in prosocial de­
longer benefit the actors—for instance, if environmental protection cision making.
becomes expensive, or donations are anonymized? If they ultimately
renege on their environmental practices or charitable donations, does 2. Stage 1: Hoping for self-interest without self-reproach in self-
this signify that they never had prosocial intentions from the start? other tradeoffs
Although abandoning the original prosocial behaviors could be taken as
evidence of self-interested motivations, and even worse, hypocrisy, we Self-other tradeoffs leave many people feeling trapped between two
argue that such a judgment may be premature. While actors’ original unpleasant outcomes: engaging in prosocial behavior at a cost to their
intentions may not qualify as purely prosocial, they may not qualify as resources (time, money, effort), and engaging in a self-interested
purely self-interested either. behavior at a cost to their self-regard (Berman & Small, 2012; Lin,
Research on self-other tradeoffs typically portrays a binary choice Schaumberg, & Reich, 2016; Liu & Lin, 2018). Trade-offs of this form are
between acting prosocially (for the benefit of others) or out of self- so wrought with ambivalence that people often avoid making the deci­
interest (for the benefit of oneself). In the present paper we claim that sion altogether (Andreoni, Rao, & Trachtman, 2017; Cain, Dana, &
prosocial decisions are dynamic and that people’s preference structure Newman, 2014; Dana, Cain, & Dawes, 2006), even in private and despite
and decision strategies evolve as the available options change, so as to a cost to themselves (Lin et al., 2016).
maximally serve themselves. We employ a familiar paradigm (Batson, An influential and now-classic experiment on self-other tradeoffs

* Corresponding author at: 1 Ayer Rajah Ave., Singapore 138676, Singapore.

E-mail address: [email protected] (S.C. Lin).

https://doi.org/10.1016/j.obhdp.2021.02.005
Received 13 May 2019; Received in revised form 15 February 2021; Accepted 15 February 2021
Available online 14 April 2021
0749-5978/© 2021 Elsevier Inc. All rights reserved.
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

claims that people bypass this conflict by appearing moral to themselves justify—rather than fail to internalize—their disregard of the coin flip’s
and others without actually intending to be so (Batson & Collins, 2011; outcome.
Batson, Kobrynowicz, & Dinnerstein, 1997; Batson, Thompson, & Chen, The original moral hypocrisy paradigm makes salient one particular
2002; Batson, Thompson, Seuferling, Whitney, & Strongman, 1999). In justification: assigning themselves the positive task is still in accordance
this paradigm, participants are instructed to allocate two tasks to with the initial instructions. As the coin flip is optional, participants
themselves and another participant—a pleasant task with positive have the prerogative to assign tasks however they wish. Consistent with
monetary consequences, and a neutral task with no positive conse­ other research that indicates that salient counterfactual outcomes (e.g.,
quences. They can assign the tasks according to their own preference or being randomly assigned a relatively negative outcome before witness­
flip a coin to decide (with the toss’s outcome known only to themselves). ing more positive outcomes that one “almost” achieved) can justify
Many participants (between 50-67%) choose to flip the coin, but sub­ dishonesty (Shalvi et al., 2011), participants can recruit the counter­
sequently assign themselves to the positive task more often than chance factual that they “could have” assigned themselves to the positive
would predict. This occurs even when the coin is explicitly labeled “SELF outcome from the start to justify their revoking of the coin flip’s
to POS” and “OTHER to POS” (Batson et al., 1999). Batson and his outcome. Different from this past research, though, this specific coun­
colleagues describe this behavior as an example of “moral hypoc­ terfactual is their own choice at an earlier decision point (“I could have
risy”—the coin flip acts as a pretense for having moral standards (to chosen to assign myself to the positive task from the start”) rather than a
others and oneself) that the participants do not actually have (Batson & chance outcome (e.g., “the coin could have flipped in my favor”). We
Collins, 2011). People have no actual intention of being moral—only argue that participants backtrack to this earlier stage, reasoning that
appearing so (Batson et al., 1999; Batson & Thompson, 2001). They give assigning the desirable outcome to themselves is prescriptively norma­
themselves moral credit for flipping the coin without internalizing the tive (as it is within the parameters of the instructions) whether this is
fact that they lied about the outcome (i.e., they internalize the coin flip done before or after the coin toss.
but do not compare their disregard for its outcomes against their moral Moreover, this specific justification could lead participants to over­
standards, Batson & Collins, 2011). estimate how often others choose to assign tasks without even using the
These interpretations imply that people settle on a decision strategy randomizer. This would further justify their self-interested action as
before flipping the coin. Consciously or not, they decide to follow a self- people feel more comfortable engaging in self-interested behavior when
interested path; flipping the coin first is a mere “shortcut” to appear to be it is descriptively normative (Bandura, Barbaranelli, Caprara, & Pas­
moral, both to others and to themselves (Tsang, 2002). We suggest an torelli, 1996; Cialdini, Reno, & Kallgren, 1990; Gino, Ayal, & Ariely,
additional interpretation: Participants’ choice to flip the coin is made in 2009; Keizer, Lindenberg, & Steg, 2008). In other words, given that they
the hope (a 50% chance) that they will receive the favorable out­ could have assigned themselves to the positive task from the start,
come—thus being rewarded with the positive task without incurring participants may generate the expectation that most others will have
self-reproach. This claim is supported by the finding that people are done so, thereby justifying their switch from relying on the randomizer
happier being assigned a self-interested option than they are choosing it (the coin) to retroactively electing not to use the randomizer and simply
for themselves (Berman & Small, 2012). Furthermore, instead of having assigning themselves to the positive task.
the premeditated intention of dishonestly claiming the favorable
outcome if the toss goes against them (per the moral hypocrisy account), 4. Overview of studies
we posit that participants decide to flip the coin without considering
what they might do if the toss does not favor them. At this stage, they We posit that the dishonest behavior observed in the original moral
may even believe they will obey the coin regardless of its outcome. This hypocrisy paradigm (Batson, Kobrynowicz, & Dinnerstein, 1997; Bat­
follows prior research which shows that people will choose to be son, Thompson, Seuferling, Whitney, & Strongman, 1999) and other
randomly assigned to a prosocial (vs. self-interested) option when that situations is not always (consciously or unconsciously) premeditated,
choice is binding (Lin & Reich, 2018). Thus, we argue that in the classic but can instead evolve in stages as the situation changes. In the first
moral hypocrisy paradigm, many participants choose to flip a coin stage, people hope to "have their cake and eat it too"—achieve self-
because it is the only option that makes it possible to avoid both a ma­ interest without self-reproach. When that proves impossible, they
terial cost (i.e., doing a boring task with no chance of winning any move on to the second stage, resorting to justification strategies that
money) and a cost to self-regard (i.e., guilt about selfish behavior). allow them to impose their self-interested choice while minimizing self-
reproach. We thus propose a dynamic model of moral decision making
3. Stage 2: Adjustment of decision strategies via justification whereby people prioritize resolving conflicting goals to be moral and
self-interested in a way that achieves both goals before shifting to self-
When their initial goal—achieving a self-interested outcome without interested motivations.
the accompanying self-reproach—is no longer possible, we suggest that We report four studies that reexamine the original paradigm and
people confront the second stage of the decision process: justifying their make the case that our dynamic perspective on moral choice offers an
self-interest. Specifically, when the coin flip does not favor them and additional explanation for the behavior of “moral hypocrites.” Studies 1
they are given an opportunity to revoke it, we contend that people adjust and 2 test the first stage of the dynamic process: that many people’s
their moral standards to allow themselves to disregard the coin flip. Such primary motivation is to simultaneously be moral and achieve the self-
rationalization strategies are common (Bandura, Barbaranelli, Caprara, interested outcome. In Study 1, we provide evidence that people view
& Pastorelli, 1996; Paharia, Vohs, & Deshpandé, 2013; Pittarello, Leib, the self-interested outcome via coin flip as the procedural outcome that
Gordon-Hecker, & Shalvi, 2015; Shalvi, Eldar, & Bereby-Meyer, 2012; leads to the most happiness. This supports our argument that people
Shalvi, Gino, Barkan, & Ayal, 2015), and allow people to minimize the choose the randomized outcome in hope of achieving the self-interested
self-threat associated with unethical or self-interested behavior. In order outcome without any corresponding self-reproach.
to engage in self-interested or unethical actions without harming their In Study 2, we replicate the original paradigm examining real
moral self-view people will, for example: take advantage of ambiguous behavior in an online setting (Batson et al., 1999, 2002) and also show
interpretations of their behavior (Pittarello et al., 2015; Shalvi, Dana, that a significant proportion of people choose to use the randomizer
Handgraaf, & De Dreu, 2011), reframe their behavior so it seems benign even if they anticipate being forced to adhere to the outcome. The fact
or even positive in its consequences for others (Bandura et al., 1996; that people choose the randomizer even when bound to its outcome
Wiltermuth, 2011), or change their standards of behavior so as to make suggests that their shift in strategy when they receive the undesirable
acceptable their desired behavior (Paharia et al., 2013). Thus, when the outcome is not premeditated as originally implied. Instead, it suggests
coin assigns them to the undesirable task, we suggest that some people that people value being randomly assigned the positive outcome so

204
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

highly that they are willing to risk their own self-interested outcome for which participants compared the four possible outcomes from the
it. original paradigm to one another.
Studies 3 and 4 provide evidence for the second stage of the dynamic
process: when they can no longer achieve their primary goal (i.e., to 5.1.2. Procedure
simultaneously be moral and achieve the self-interested outcome), Participants imagined “J,” a participant who was taking part in a
people recruit justifications to shift their decision strategy. We posit that psychological study. They read that J was given the exact instructions
when the randomizer assigns the unfavorable (prosocial) outcome, from the original paper (Batson et al., 1999). To summarize, J was told
people find justifications for assigning the pleasant task to themselves that there were two tasks: a positive consequences and neutral conse­
rather than adhering to the outcome of the coin toss. Study 3 tests for quences task. The positive consequences task would reward correct re­
one such strategy that the moral hypocrisy paradigm leaves availa­ sponses with raffle tickets to ultimately win a $30.00 gift certificate to
ble—that disregarding the random outcome is within the parameters of the store of J’s choosing. The neutral consequences task was relatively
the instructions. We find that ignoring the outcome of the toss and dull, and would involve performing a task and being given feedback
assigning themselves to the self-interested task is viewed as more justi­ with no monetary consequences. Participants further read that the
fiable when the coin flip is optional (i.e., they are permitted to assign experimenter explained to J that they would be assigning themselves
tasks without it) than when it is mandatory. and another participant to the two tasks. Importantly, this included the
Finally, in Study 4, we examine whether rationalization strategies description of J’s opportunity to flip a coin to make their decision
are linked to the decision to switch from using the randomizer to directly (Batson et al., 1999, p. 528):
assigning the favorable task to themselves. When we compare partici­
Most participants feel that giving both people an equal chance—by,
pants who are denied the positive outcome with those who are assigned
for example, flipping a coin—is the fairest way to assign themselves
the positive outcome, the former estimate that fewer people used the
and the other participant to the tasks (we have provided a coin for
randomizer from the start. This in turn leads them to surreptitiously
you to flip if you wish, designating one side “SELF to POS” for pos­
revoke their original decision to use the randomizer and to simply assign
itive consequences for yourself and the other side “OTHER to POS”
themselves the positive task. This suggests that only when they are de­
for positive consequences for the other participant). But the decision
nied their preference to achieve the self-interested outcome fairly (i.e.,
is entirely up to you. You can assign yourself and the other partici­
via coin toss) do people use the fact that they could have assigned
pant however you choose. The other participant does not and will not
themselves to the self-interested outcome from the start to justify
know that you are assigning tasks; he or she will think that the task
switching retroactively.
assignment was purely by chance. Because of this and because the
Across all studies, we ensured ample power by seeking at least 100
two of you will never meet, your anonymity is assured.
participants per condition (Simmons, Nelson, & Simonsohn, 2013). We
report all measures and manipulations used, and any a priori exclusion Participants then read about the possible outcomes (we refer to each
criteria and subsequent exclusions. Data and analysis scripts are avail­ specific outcome by the corresponding bracketed term):
able at https://osf.io/6x8kr/?view_only=a3e43590171a493b8cfc6ae
0087ba06d. All studies received Institutional Review Board approval. Assume that, if J flips the coin, J must do what the coin says.
Therefore, the possible outcomes for J are the following: J chooses to
5. Study 1 flip the coin and gets SELF to POS [flip-self], J chooses to flip the coin
and gets OTHER to POS [flip-other], J chooses NOT to flip the coin
Whereas the moral hypocrisy account suggests that people seek to and chooses to assign themselves1 to POS [assign-self], J chooses
appear moral without having to be moral, we suggest that they are NOT to flip the coin and chooses to assign the other participant to
primarily motivated to be moral while simultaneously achieving a self- POS [assign-other].
interested outcome. To examine this reasoning, Study 1 tested whether Participants then indicated how mixed J’s feelings would be toward
people showed a preference for the option that would allow them to each outcome (i.e., subjective ambivalence, Snyder & Tormala, 2017) on
achieve a self-interested outcome through a moral route. Participants a 7-point scale (1 = not at all mixed, 7 = extremely mixed). They then
read a description of the original paradigm (Batson et al., 1999), and indicated how positive J would feel, ignoring all negative feelings about
estimated how participants completing this study would feel under the the outcome (1 = no positive thoughts and feelings, 7 = maximum positive
different possible outcomes: flipping the coin and being assigned the thoughts and feelings), and vice-versa, a measure of objective ambivalence
positive task, flipping the coin and assigning the other participant to the (i.e., the coexistence of positive and negative evaluation of the same
positive task, assigning themselves to the positive task without flipping attitude object). Objective ambivalence was calculated from these items
the coin, and assigning the other participant to the positive task without using the Griffin method, which takes into account both the intensity and
flipping the coin. We propose that flipping the coin and being fairly similarity of positive and negative reactions (i.e., (Pos + Neg)/2 – |Pos­
assigned to the positive task allows participants to achieve both the goal –Neg|, Thompson, Zanna, & Griffin, 1995). They then indicated how
to be moral and to be assigned the positive task. We thus expected guilty J would feel (only shown to 72 participants due to study pro­
participants to anticipate that flipping the coin and being randomly gramming error; 1 = not at all guilty, 7 = extremely guilty), and how happy
assigned the positive task would be the most desirable outcome. This J would feel (1 = extremely unhappy, 7 = extremely happy) with each
would support our claim that participants who chose to flip the coin in outcome. Participants then ranked J’s preference for the outcomes, “with
the original paradigm did so in the hope that they would achieve the 1 being MOST preferred and 4 being LEAST preferred,” using a dragging
self-interested outcome without compromising their self-regard, but not rank-order question. Lastly, they indicated their age and gender.
with the intention that they would disregard the toss outcome if it went
against them.
5.2. Results
5.1. Method
Because participants were asked to comparatively evaluate all four
5.1.1. Participants and design possible outcomes (flip-self, flip-other, assign-self, assign-other), we
We aimed to collect 100 participants from Amazon Mechanical Turk.
We stopped collecting data when the request was complete, resulting in
99 participants (62.63% male, 37.37% female, Mage = 34.86). We did 1
Colloquial grammar used for participants and throughout results and dis­
not exclude any participants. The study was a within-subject design in cussion to increase fluency.

205
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Fig. 1. All dependent measures predicted by outcome. Note: Error bars are 95% confidence intervals.

Table 1
Means and standard deviations across dependent measures in Study 1.
Outcome

Dependent measure Flip-self Flip-other Assign-self Assign-other

Happiness 5.91 (1.26) 3.58 (1.49) 5.24 (1.28) 4.04 (1.48)

Objective ambivalence 0.79 (2.68) 2.05a (2.06) 1.82a (2.36) 2.17a (2.20)
Subjective ambivalence 3.04 (1.91) 4.49a (1.73) 4.41a (1.97) 4.79a (1.65)
Guilt 2.72a1 (1.97) 2.68a2 (1.85) 4.50 (1.70) 3.1712 (1.99)
Rank 1.48 (0.83) 2.76 (0.78) 2.32 (1.02) 3.44 (0.82)

Notes. Standard deviations are parenthetical following means. Means that share the same subscript letter within rows are not significantly different from one another.
Means that share the same subscript number within rows are marginally different from one another.

treated the outcomes as falling under one factor. To make these com­ each outcome, being randomly assigned the positive outcome was the
parisons we used linear mixed-models, controlling for by-participant least ambivalence-inducing outcome, significantly differing from all
intercepts. We used dummy codes to code different outcomes as the other outcomes: versus assign-self, t(295) = 4.72, p < .001, d = .382;
reference group. versus flip-other, t(294) = 5.76, p < .001, d = 0.53; versus assign-other, t
(294) = 6.32, p < .001, d = 0.51. None of the other outcomes differed
5.2.1. Happiness from each other (see Web Appendix A). Patterns were the same for
Each of the outcomes was significantly different from each other in subjective ambivalence (see Table 1, Fig. 1, and Web Appendix A).
happiness ratings in descending order from most to least happiness: flip-
self, assign-self, assign-other, flip-other. Participants believed that flip­ 5.2.3. Guilt
ping “SELF to POS” would induce more happiness than assigning oneself Unsurprisingly, participants expected that J would feel the guiltiest if
to the positive outcome without flipping the coin (see Fig. 1 and Table 1 they assigned themselves to the positive outcome without the coin:
for descriptive statistics of all dependent measures), t(294) = 3.54, p < versus assign-other, t(213) = 5.34, p < .001, d = 0.72; versus flip-self, t
.001, Cohen’s d = 0.53, which induced more happiness than assigning (213) = 7.19, p < .001, d = 0.90; versus flip-other, t(213) = 7.36, p <
the other participant to the positive outcome without flipping the coin, t .001, d = 1.07. Participants expected that J would feel equally guilty
(294) = 6.39, p < .001, d = 0.87, which induced more happiness than when choosing to flip the coin regardless of whether they flipped “SELF”
flipping “OTHER to POS,” t(294) = 2.47, p = .014, d = 0.31. See Web or “OTHER” to “POS,” t(213) = − 0.17, p = .87. Participants also ex­
Appendix A for detailed analyses for all measures. pected that assigning the participant to the positive outcome would
cause marginally more guilt than either flipping “OTHER to POS,” t
5.2.2. Objective ambivalence (213) = 1.80, p = .074, d = 0.24, or flipping “SELF to POS,” t(213) =
When considering the simultaneous positive and negative aspects of 1.97, p = .051, d = 0.22. This was not anticipated but is not discussed

2
Throughout the paper, we report effect sizes only for marginal or significant
effects (p < .10).

206
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

further as the effects were marginal and this preference was not actions of at least some of the participants classified as moral hypocrites
consistently reflected across the other measures. in prior literature. We created an online version of the original moral
hypocrisy paradigm, allowing participants to use a digital coin (i.e., click
5.2.4. Rank on a randomizer) if they wished. We manipulated whether the ran­
Each of the outcomes was ranked significantly differently from each domizer’s outcome was automatic or revocable: Participants were told
other in the following order from most to least preferred: flip-self, that if they used the randomizer they would either automatically be
assign-self, flip-other, assign-other (see Table 2 for frequency of rank­ assigned to the random outcome received (automatic randomizer condi­
ings, Web Appendix A for most common rank orders). Participants tion), or that they would indicate their chosen assignment after viewing
thought J would prefer flipping “SELF to POS” (ranked first 68.75% of the random outcome (revocable randomizer condition, similar to the
the time) over assigning themselves to the positive outcome without the original paradigm).
coin, t(380) = 6.74, p < .001, d = 1.02, which J would prefer over We anticipated replicating the original results in the revocable
flipping “OTHER to POS,” t(380) = 3.49, p < .001, d = 0.43, which J randomizer condition: more than half of those who used the randomizer
would prefer over assigning the other participant to the positive would ultimately assign the positive outcome to themselves. Addition­
ally, and more importantly, we predicted that a substantial number of
Table 2 participants would show a preference for the randomizer even if it
Frequency of rank order of each outcome. would automatically (and irrevocably) assign them to the random
outcome. The moral hypocrisy account implies that participants have no
Rank order
intention to be moral; instead, they have a premeditated plan to take the
Outcome Rank 1 Rank 2 Rank 3 Rank 4
positive outcome for themselves and simply flip the coin first in order to
Flip-self 68.75% 4.17% 22.92% 4.17% “appear moral without being so.” As an automatic randomizer renders
Flip-other 19.79% 32.29% 39.58% 8.33% this impossible, the moral hypocrisy account would predict that par­
Assign-self 6.25% 46.88% 19.79% 27.08%
ticipants would opt not to use the randomizer and directly assign
Assign-other 5.21% 16.67% 17.71% 60.42%
themselves to the positive task instead. In contrast, our dynamic moral
shifting account suggests that many people who choose the randomizer
outcome without the coin, t(380) = 5.41, p < .001, d = 0.83. The lowest are primarily motivated to be moral and achieve the self-interested
ranked option was assigning the other participant to the positive outcome simultaneously, and are prepared to commit to its outcome
outcome without the coin. in the hope it will result in a favorable outcome (i.e., assign the positive
task to them). We posit that the decision to revoke the randomizer is not
5.3. Discussion foreseen, but occurs only after the randomizer reveals an unfavorable
outcome (i.e., assigning the other participant the positive task). Thus,
The results of Study 1 supported our prediction that participants in our model predicts that a significant number of participants (i.e., a
the original moral hypocrisy paradigm view being randomly assigned proportion greater than 0, as determined by a 95% confidence interval)
the positive task as the most desirable outcome. Ranked as the most would still choose to use the randomizer, even if it would automatically
preferred outcome, it was expected to make participants both the assign them to the outcome.
happiest and the least ambivalent. Assigning themselves to the positive Although we hypothesized that a significant number of people would
outcome was viewed as producing the second highest level of happiness. still choose the randomizer in the automatic randomizer condition, we
This pattern supports the hypothesis that people’s decision strategy were open to the possibility that fewer people would choose the
evolves over time from their most preferred outcome (i.e., self-interest randomizer in the automatic than the revocable randomizer condition.
without selfishness) to, if they lose the toss, their second preferred This difference would represent the presence of “moral hypocrites” who
outcome (i.e., choosing the self-interested outcome). were willing to choose the randomizer only when they could revoke its
Participants viewed all outcomes except the flip-self outcome with outcome if unfavorable. Thus, another aim of this study was to estimate
relatively high ambivalence. This is consistent with our contention that what proportion of those who revoke the random outcome are “moral
self-other tradeoffs force people into a lose-lose choice. In flipping the hypocrites” (i.e., internalizing the use of the randomizer while simul­
coin and having the other participant being assigned the positive task, J taneously intending a priori to ignore its outcome) versus “dynamic
(the target) may feel that they acted fairly, but would be disappointed by moral shifters,” (i.e., who revoke the coin flip only after they receive the
the outcome. Similarly, in assigning someone else to the positive task, J unfavorable outcome), and to examine whether we could distinguish
may feel they have acted altruistically, but again would be unhappy with these two groups psychologically. To this end, we measured a series of
the task they must consequently complete. Finally, although it was the individual differences separately from the main study. This also allowed
second most preferred choice, by assigning themselves to the positive us to test whether those who revoke the unfavorable outcome have
task without using the coin, J was expected to feel guilty about how they higher moral rationalization tendencies, which would support our
came to be completing it (participants indeed anticipated that J would contention that those who ultimately revoke the unfavorable outcome
feel less guilty about being randomly assigned to the positive task by the use rationalization to do so.
coin than assigning themselves to the positive task). Interestingly, par­ Finally, we tested the extent to which participants perceived their
ticipants perceived these three outcomes to be equally mixed (i.e., own decision-making process to be moral. Our dynamic moral shifting
ambivalence-inducing). Thus, the flip-self outcome is the only one that account suggests that participants rationalize that they could have just
allows participants to sidestep the lose-lose situation; it allows them to assigned themselves the positive task from the start (without using the
retain moral self-regard while also achieving the self-interested randomizer). Thus we expected those who revoke the randomizer’s
outcome.3 outcome to view their decision strategy as equally moral to those who
assigned themselves the positive outcome directly (without using the
6. Study 2 randomizer), rather than as more moral (as found originally, Batson
et al., 2002).
Study 2 tested the claim that dynamic moral shifting accounts for the
6.1. Method

3
See Web Appendix A for additional discussion regarding differences in 6.1.1. Participants and design
ranking and happiness measures. After conducting a similar pilot study (see Web Appendix B), we pre-

207
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

registered this study (http://aspredicted.org/blind.php?x=rg3te9) with moved on to another page where they indicated their final assignment.
an aim to recruit at least 1000 participants to complete a two-part study; Those in the automatic assignment condition who opted for the
the first part was a battery of scale items (opened to 2000 HITs on randomizer were told “you will complete the [positive / neutral] task”
Amazon Mechanical Turk) and the second part was the main study depending on what the randomizer revealed on the previous page.
(opened to those who legitimately completed the first part of the study). After making their choice, participants indicated the extent to which
This led to a total of 1620 viable responses (50.03% male, 49.14% fe­ they felt each of four self-conscious emotions about the way they
male, 0.56% non-binary, Mage = 38.06). Any worker who appeared assigned tasks—shame, guilt, embarrassment, and pride (1 = not at all, 7
more than once (according to Worker ID) in the dataset was removed = extremely). Participants also responded to the moral self-image (MSI)
from the sample entirely4. The main study had two rounds of conditions: scale (Jordan, Leliveld, & Tenbrunsel, 2015). MSI and self-conscious
first, participants were assigned to a randomizer type (automatic or emotions were counterbalanced. Finally, participants indicated the
revocable; i.e., the randomizer’s outcome would either result in auto­ perceived morality of their task assignment method (“Do you think the
matic assignment or the participants would be responsible for assigning way you made the task assignment was morally right?” 1 = not at all, 9
themselves to tasks regardless of whether they used the randomizer or = yes, totally), as per the original research (Batson et al., 2002). Given its
not). Then, participants who chose to use the randomizer were assigned centrality to the original argumentation and its direct relationship to the
to outcome condition: either self or other to the positive outcome. Thus, constructs of interest, we focus on the last measure in our analyses and
all participants were assigned to a randomizer type condition (automatic report all other analyses in Web Appendix B.
vs. revocable), but only those who chose to use the randomizer were Participants then completed a filler task. Depending on whether they
assigned to an outcome condition (self or other to positive outcome). ultimately completed the positive or neutral task, they were told they
would be entered into the lottery or that the other participant would be.
6.1.2. Procedure They finally indicated their age and gender and exited the study. A
In the first part of the study, participants responded to the following lottery payment was made at the end of the study to a random partici­
scales in random order: the self-importance of moral identity scale pant who assigned him or herself to the positive consequence task.
(Aquino & Reed, 2002, looked at separately as both internalized and
symbolic); the eight-item moral disengagement scale (Moore, Detert, 6.2. Results
Klebe Treviño, Baker, & Mayer, 2012); the moral clarity scale (Wilter­
muth & Flynn, 2013); the 13-item short form social desirability scale 6.2.1. Replication of original findings
(Ballard, 1992), and the justice perspective and relationship-care single- First, we tested whether the present results replicated the original
item measures (Batson, Kobrynowicz, & Dinnerstein, 1997). paradigm within the revocable randomizer condition (Batson et al.,
All of those who completed the first study were invited to participate 1999) (see Fig. 2 for decision tree of results). Of the 276 who did not use
in the main part of the study via the MTurk platform a few days later the randomizer, 231 (83.70%) assigned themselves to the positive
(without being informed why they qualified). In this part, participants outcome. Of the 540 who used the randomizer, 394 (72.96%) assigned
read instructions from Study 1, which were directed at them as actors themselves to the positive outcome, which is a proportion significantly
rather than directed at an imagined third party, and adapted to the greater than chance, p < .001. Those who flipped5 “OTHER to POSITIVE
MTurk context. Instead of lottery entries for the $30.00 gift card, par­ TASK” were more likely to assign the other participant to the positive
ticipants were told that the positive task would be linked to lottery en­ task (47.35%) than those who flipped “SELF to POSITIVE TASK”
tries for a $10.00 MTurk bonus, which is a high amount on MTurk (e.g., (7.61%), B = 1.64, SE = 0.19, p < .001, Odds ratio (OR) = 5.16. How­
the main study paid $0.72 to complete). For the randomizer option, ever, conceptualizing this finding another way (and as pre-registered),
participants read: those who flipped “OTHER to POSITIVE TASK” were less likely to
Most participants feel that giving both people an equal chance—by, obey the randomizer’s outcome than those who flipped “SELF to POSI­
for example, flipping a coin—is the fairest way to assign themselves TIVE TASK” (47.33% vs. 92.39%), B = 2.60, SE = 0.26, p < .001, OR =
and the other participant to the tasks. Thus, if you wish, you can click 13.50.
the button below, which will either display “SELF to POSITIVE
TASK” for positive consequences for yourself or “OTHER to POSI­ 6.2.2. Hypothesis testing: effect of randomizer type
TIVE TASK” for positive consequences for the other participant. Next, and crucial to the purpose of the study, we compared the
automatic randomizer condition to the revocable randomizer condition.
In the automatic randomizer condition, participants read “If you view Participants in the automatic randomizer condition were marginally less
the random outcome, you will automatically be assigned to the random likely to use the randomizer (61.57%) than those in the revocable
outcome you view when you move on to the next page.” In the revocable randomizer condition (66.18%), B = 0.20, SE = 0.10, p = .054, OR =
randomizer condition, participants instead read, “You can then report 1.22 (see Fig. 2 for full results). Importantly, even though they were
your assignment directly on the next page.” As instructions read in Study committed to its outcome, the number of participants using the
1, participants were reminded that the decision was up to them, they randomizer was high and significantly above 0 (61.57%, 95% CI =
could assign the other participant however they chose, and that the [58.1%, 65.9%]).
other participant would not be aware that they were assigning tasks. If
participants clicked on the randomizer, they randomly viewed either 6.2.3. Individual differences
“SELF to POSITIVE TASK” or “OTHER to POSITIVE TASK” beneath the We tested whether we could identify differences between “moral
randomizer button. hypocrites” and “dynamic moral shifters.” We compared individual
Both those in the automatic randomizer condition who chose not to differences between those who used the randomizer in the revocable
use the randomizer and all those in the revocable randomizer condition condition (in which some were classic “moral hypocrites” who planned
to revoke the randomizer) and those who used the randomizer in the
automatic assignment condition (in which they were forced to commit
4
This deviates slightly from the preregistration that stated that only those
who failed the English speaking check but re-entered the study would be
5
removed (n =10). An unanticipated 24 participants also simply completed the Although there was no physical coin involved, we hereby refer to receiving
study more than once. Although we did not pre-register their removal, we an outcome via randomizer as having “flipped” that outcome (e.g., those who
reasoned that their data should not be included; however, including them in the clicked the randomizer and randomly viewed “OTHER to POSITIVE TASK” are
sample does not change the results. said to have “flipped OTHER to POSITIVE TASK”).

208
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Fig. 2. Frequency of choice outcomes from Study 2, organized as a tree that begins with condition assignment, then branches from left to right in temporal order
of choice.

Table 3
Means and standard deviations across individual difference in Study 2.
Behavioral Pattern Statistical comparisons

Measure (1) Revokers: Flip “OTHER to POSITIVE (2) No randomizer, assign (3) Flip “OTHER to POSITIVE Group 1 vs. 2 Group 1 vs. 3
TASK,” assign self to positive task self to positive task TASK,” assign other to positive task

Moral disengagement 2.50 (1.52) 2.05 (0.88) 2.14 (0.95) t = 3.60, p < .001, t = 2.51, p = .012,
d = 0.29 d = 0.28
Social desirability 4.69 (0.25) 4.68 (0.27) 4.61 (0.26) t = 0.54, p = .59 t = 2.76, p = .006,
d = 0.34
Symbolic moral identity 3.19 (0.99) 2.87 (0.93) 3.12 (0.89) t = 3.22, p = .001 t = 0.66, p = .51
d = 0.32
Internalized moral identity 4.18 (0.73) 4.37 (0.61) 4.34 (0.64) t = − 2.60, p = t = − 1.91, p =
.010, d = 0.26 .057, d = 0.23
Social justice 3.70 (0.58) 3.78 (0.50) 3.74 (0.52) t = − 1.43, p = .15 t = − 0.60, p = .55
Moral clarity 4.18 (1.00) 4.15 (0.90) 4.17 (0.82) t = 0.34, p = .74 t = 0.11, p = .91
Relationship care 6.50 (1.93) 6.42 (1.95) 6.92 (1.59) t = 0.42, p = .68 t = − 1.91, p =
.057, d = 0.24
Justice 6.92 (1.40) 6.83 (1.53) 6.95 (1.53) t = 0.57, p = .57 t = − 0.17, p = .86

Notes. Standard deviations are parenthetical following means; t-values have 810 degrees of freedom.

to the randomizer’s outcome). We did not find any differences on these participants into six a priori determined behavioral patterns reflecting
variables (see Web Appendix B for statistics; discussed further below). the six branches in the revocable randomizer condition (see Fig. 2). We
Our dynamic model predicts that many of those who revoke the were most interested in “revokers”—the group who used the random­
random outcome will morally rationalize their behavior. To test this, we izer, flipped “OTHER to POSITIVE TASK,” and subsequently assigned
analyzed the revocable randomizer condition only and categorized themselves to the positive task. As pre-registered, we compared these

209
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Fig. 3. Moral judgment of decision making process for those who engaged in the positive task, as a function of how they arrived at the positive task. Note: (1)
"Revoke randomizer for self" and (2) "Assign self" both differ from (3) "Forced (automatic assignment)" and (4) "Obey randomizer", ps < 0.002; (1) and (2) do not
differ, p = .38; (3) and (4) marginally differ, p = .099.

participants to those who flipped “OTHER to POSITIVE TASK” and 6.3. Discussion
honestly assigned the other participant to the positive task, and to those
who simply assigned themselves the positive task without using the The results of Study 2 replicated much of the original paradigm in an
randomizer. To summarize, revokers were significantly higher in social online setting in the revocable randomizer condition. This is reassuring
desirability and moral disengagement, and marginally lower in inter­ because of the change in setting and also because the original studies
nalized moral identity, and in relational care than those who flipped often had very few participants by today’s standards (e.g., Batson et al.,
“OTHER to POSITIVE TASK” and obeyed the randomizer. Revokers were 1999, 2002).
significantly higher in moral disengagement and symbolic moral iden­ Most relevant to the main purpose of the study, we found that a
tity, and lower in internalized moral identity than those who assigned substantially large proportion (62%) of participants opted to use a
themselves to the positive task without the use of the randomizer (see randomizer even when they knew they would be forced to follow its
Table 3). outcome. Being automatically and irrevocably assigned to a task was not
sufficient to deter people from using the randomizer. They were willing
6.2.4. Perceived morality of the assignment to click the randomizer with a 50% chance of being assigned to the
We examined participants’ perceptions of the morality of task negative task. Anticipating the ability to revoke the randomizer’s
assignment. Following our analysis plan, we examined the subset of outcome led only 5% more participants to use the randomizer. The fact
those who completed the positive task only. We categorized participants that 62% of the randomizer choosers in the revocable randomizer con­
using the four means by which they could have ended up with the dition would have done so knowing they could not renege supports our
positive task, which we refer to by the bracketed terms: (1) those who proposition that people adjust their decision-making preferences as their
flipped “OTHER to POSITIVE TASK” in the revocable randomizer con­ options evolve. They first seek to achieve the self-interested outcome
dition [revoked randomizer], (2) those who assigned themselves to the without self-reproach by using the randomizer: They hope that they will
positive task with no randomizer (in either randomizer type condition) be assigned the positive outcome, but are prepared to accept the unfa­
[assigned self], (3) those who flipped “SELF to POSITIVE TASK” in the vorable outcome7. When they are not assigned the positive task—and
automatic assignment condition [forced by randomizer], and (4) those when they have the chance to renege—they adjust their decision making
who flipped “SELF to POSITIVE TASK” in the revocable randomizer strategy.
condition [obeyed randomizer]. Entering this coded variable into a Individual difference measures did not identify unique characteris­
regression predicting morality of task assignment revealed a significant tics of Batson’s “moral hypocrites” (i.e., those who choose the
overall effect of route to the positive task, F(3, 1139) = 20.99, p < .001. randomizer only when they know they can revoke the outcome). This
Those who revoked the randomizer viewed their decision to be as moral may be because “moral hypocrites” were a small minority (i.e., 5% of
(M = 6.46, SD = 1.92) as those who assigned themselves to the positive 66%) of those who chose the randomizer in the revocable randomizer
task (M = 6.29, SD = 2.17), t(1139) = − 0.88, p = .38, but to be less condition. Thus, detecting differences in individual characteristics of
moral than both those who were forced to follow the randomizer (M = those who chose the randomizer in the revocable versus automatic
7.39, SD = 1.68), t(1139) = 4.42, p < .001, d = 0.49, and those who randomizer may require more statistical power.
willingly obeyed the randomizer (M = 7.10, SD = 1.84), t(1139) = 3.06, However, exploratory analyses did shed light on people who revoke
p = .002, d = 0.33 (see Fig. 3).6 the randomizer. First, those who flipped the unfavorable outcome and
revoked it were higher in moral disengagement tendencies than both
their counterparts who flipped the unfavorable outcome and obeyed it,
6
and those who assigned themselves the positive task without using the
Those who assigned themselves to the positive task viewed their decision as
randomizer. As moral disengagement reflects the tendency to rationalize
less moral than both those who were forced to follow the randomizer, t(1139) =
7.08, p < .001, d = 0.51, and who willingly followed the randomizer, t(1139) =
5.31, p < .001, d = 0.37. Those who were forced to follow the randomizer
7
viewed themselves as marginally more moral than those who willingly followed Or, at least not considering what they would do when they viewed the
the randomizer, t(1139) = –1.65, p = .099, d = 0.17. unfavorable outcome.

210
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

one’s immoral decisions, this finding supports our dynamic model’s 7.1.2. Procedure
assumption that revoking the randomizer involves rationalization. Also, Participants read the instructions and description of the study as in
revokers were higher in symbolic moral identity but lower in internal­ Study 1, again imagining a participant J participating in the original
ized moral identity than those who directly assigned themselves the experimental paradigm. However, after reading that the participant
positive task without using the randomizer. This is reminiscent of the would be assigning themselves and another participant to tasks and that
moral hypocrisy account—those who choose to ignore the unfavorable they were provided a coin to flip, participants in the mandatory coin flip
outcome care about seeming moral but not about being moral. However, condition read that J was explicitly asked to use the coin to assign tasks:
another interpretation lends itself to our dynamic account: Those who “We ask that you please use this coin to assign yourself and the
flip the coin hope to achieve the self-interested positive outcome (i.e., other participant to tasks. The other participant does not and will not
reflecting low internalized moral identity) but to do so via a moral path know that you are assigning tasks via the coin flip; he or she will think
(i.e., reflecting high symbolic moral identity). that the task assignment was done beforehand.” Participants in the
Finally, we found that those who revoked the randomizer’s outcome optional coin flip condition read that J was given the option of using the
to assign themselves the positive task judged their behavior as less moral coin flip or not, as in the original instructions: “But the decision is
than those who used the randomizer and were assigned the positive entirely up to you. You can assign yourself and the other partici­
outcome, and as equally moral to those who assigned themselves to the pant however you choose. The other participant does not and will not
positive outcome from the start. This finding is inconsistent with the know that you are assigning tasks; he or she will think that the task
original interpretation that people take moral credit for using the assignment was purely by chance.”8
randomizer even when they do not adhere to its outcome. We contend On the next page, participants in the optional coin flip condition read:
that people use rationalization processes to revoke the randomizer and “Recall that J can either flip the coin or make an assignment as he or she
assign themselves to the positive task, for instance: “I could have wishes.” All participants then read, “J flips the coin and sees that the
assigned myself without using the randomizer from the start.” This coin comes up OTHER to POS. After a moment, J fills out the assignment
rationalization strategy should lead participants to view their behavior sheet. J assigns him/herself to the positive task and the other participant
as morally equivalent to directly assigning themselves to the positive to the neutral task.” Participants indicated how justifiable this behavior
task without using the randomizer, which is what we found. In the next was (1 = extremely unjustifiable, 7 = extremely justifiable), how good or
studies, we examine this posited rationalization process. bad the behavior was (1 = extremely bad, 7 = extremely good), and how
unethical or ethical it was (1 = extremely unethical, 7 = extremely ethical).
7. Study 3 We combined these items to form our justifiability composite (ɑ = 0.90).
Participants then indicated in free response form what they thought was
The moral hypocrisy account suggests that participants give them­ going through J’s mind while making the assignment. Finally, they
selves moral credit for flipping the coin while not internalizing the fact indicated their age and gender.
that they ignored the coin flip’s unfavorable outcome. Our dynamic
account conceptualizes the moral self-concept maintenance process as 7.2. Results
having two stages. First, participants flip the coin to try to achieve the
self-interested outcome without accompanying self-reproach. We have As predicted, participants judged the fact that J ignored the coin’s
thus far provided evidence that people indeed find this outcome most outcome to be more justifiable when J was given a choice whether to flip
appealing (Study 1) and that it is so appealing that they will choose to be the coin or not (M = 3.91, SD = 1.32) than when J was mandated to flip
randomly assigned even when such random assignment cannot be the coin (M = 2.79, SD = 1.41), t(197) = 5.77, p < .001, d = 0.82.
revoked (Study 2). We then suggest that if they flip the unfavorable
outcome, they move on to the second stage: justifying their decision to
7.3. Discussion
ignore the coin’s outcome. Study 3 tests a particular justification strat­
egy—that they were told they could choose to assign tasks however they
Study 3 supports the claim that people use the fact that they were
wanted as the use of the coin was not mandatory. Participants imagined
given a choice to flip or not flip the coin as a justification for ignoring its
a participant, J, who flipped “OTHER to POS” but revoked the outcome,
outcome. As assigning the positive outcome to themselves was within
assigning him or herself to the positive outcome. For half of them the
the parameters of the rules, retroactively deciding not to rely on the coin
coin flip was optional for J (as in the original paradigm), and for the
was their prerogative. Thus, whereas the moral hypocrisy account sug­
other half, J was required to flip the coin and report the outcome. We
gests that participants use the coin flip as a moral "cover story" for their
expected people to judge J’s decision to ignore the outcome of the toss as
premeditated intention to assign themselves the positive task, we sug­
more unethical and less justifiable when J was directed to make the
gest that participants develop a justification strategy to ignore the coin’s
assignment via a coin flip than when J had been given the option of not
outcome when unfavorable and present evidence for a specific justifi­
tossing the coin and simply making the assignment up front. This result
cation for imposing their choice. In Study 4, we directly test whether
would suggest that rather than crediting themselves for the use of the
people will engage in this rationalization strategy only when they can no
coin yet not internalizing their disregard for the outcome, participants
longer achieve the self-interested outcome without self-reproach (i.e.,
acknowledge their disregard of the outcome of the toss but justify it by
the outcome of the toss is unfavorable).
claiming the prerogative to assign tasks as they deemed appropriate.

8. Study 4
7.1. Method

Study 4 sought more direct evidence of our proposed two-stage

7.1.1. Participants and design
We aimed to collect 200 responses on Amazon Mechanical Turk and
stopped data collection when the request was filled, resulting in a total 8
We removed language about what most participants thought to be a fair
of 199 participants (44.73% male, 54.77% female, 0.50% non-binary,
way to assign tasks (see Study 1 Method for wording) because we reasoned that
Mage = 37.22). We did not exclude any participants. This study was a
it would not make sense in the mandatory coin flip condition. However, this may
two-condition (mandatory coin flip vs. optional coin flip) between-subjects have interacted with our manipulation such that the optional coin flip condition
design. was only seen as more moral when the fairness was not emphasized; therefore,
we conducted a variant of this study using the original instructions that repli­
cated the effect found in this study, reported in Web Appendix C.

211
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Fig. 4. The choice offered to participants. Top: The appearance of the random assignment upon clicking on the randomizer. Bottom: The appearance of the question
upon clicking on the skip randomizer button. Note, the randomizer’s assignment has disappeared.

process. We predicted that when participants were denied the ability to with 374 participants (52.67% male, 46.52% female, 0.80% non-binary,
achieve self-interest without any accompanying self-reproach (i.e., after Mage = 36.72), with 203 who chose to use the randomizer. This study
an unfavorable outcome), they would move onto the second stage: was a two-condition between-subjects design for those who used the
justifying their disregard for the randomizer’s outcome. We predicted randomizer (outcome: self or other to positive task).
that the fact that the use of the randomizer was optional would be more
salient to those who flipped the unfavorable outcome than to those who 8.1.2. Procedure
flipped the favorable outcome. We assessed this salience by measuring The procedure was identical to the main portion of Study 2, with one
how common participants believed the use of the randomizer was major exception—in addition to the “View a random outcome below”
among other participants. We also modified the online paradigm from button (which they could click or skip in the Study 2), participants also
Study 2 so that participants who initially chose to use the randomizer had a “Skip randomizer and indicate assignment on next page” option
could “undo” that choice after viewing its outcome. Specifically, they (see Fig. 4).9 In addition to their final choice, we surreptitiously recor­
could explicitly switch decision strategies to (retroactively) “skip” the ded whether they ever clicked each of the options. Importantly, when
randomizer and directly assign themselves to an outcome. We expected they clicked “View a random outcome below,” either the outcome “SELF
those who flipped the unfavorable outcome to estimate that others were to POS” or “OTHER to POS” would appear (via random assignment)
less likely to (even) use the randomizer, justifying their decision to below (see Fig. 4, top); however, if they changed their selection to the
retroactively switch strategies and assign themselves to the positive “skip randomizer” option, the random outcome disappeared (see Fig. 4,
outcome. bottom). If participants clicked back to the random outcome option, the
text would reappear, always repeating the prior randomly assigned
outcome. We a priori decided that all participants who ever clicked the
8.1. Method
randomizer would be classified as having used the randomizer (even if
they ultimately clicked on the “skip randomizer” option). Those who
8.1.1. Participants and design
clicked on the randomizer but ended their selection on the “skip
We aimed to recruit 400 participants from Amazon Mechanical Turk
randomizer” option were classified as (1) having used the randomizer
so that we would have about 200 participants who chose to flip the coin
and (2) having switched from using the randomizer to retroactively
in our final sample. Participants first responded to a few comprehension
skipping the randomizer.
questions to ensure they were truly English speakers, due to issues with
the MTurk pool (Dreyfuss, 2018). Those who responded to any questions
incorrectly were asked to return the HIT. We eliminated any participants
9
whose Worker ID appeared more than once in the dataset. This left us Also, the randomizer outcomes and accompanying instructions read “SELF
to POS” and “OTHER to POS” rather than “SELF to POSITIVE TASK” and
“OTHER to POSITIVE TASK.”

212
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Fig. 5. Frequency of choice outcomes from Study 4, organized as a decision tree that branches from left to right in temporal order.

Once participants had selected the randomizer or not, they moved on than those who flipped “SELF to POS” (87.63%), B = − 1.69, SE = 0.37,
to another page where they made their final assignment. They then p < .001, they were more likely to disregard the coin’s outcome (56.60%
responded to our justification measure: “When assigning people to tasks, vs. 12.37%), B = 2.22, SE = 0.37, p < .001 and to indicate a self-
people could assign tasks however they wanted or use the randomizer. assignment that mismatched the coin’s assignment.
Which of the following do you think is most likely to be true about how We next examined participants’ switching behavior—the tendency
people make assignments?” (1 = Almost everyone chooses without using for participants to retroactively switch from using the randomizer to
the randomizer, 2 = Most people choose without using the randomizer, 3 = skipping the randomizer. Those who flipped “OTHER to POS” were over
About half use the randomizer and half choose without it, 4 = Most people three times as likely to switch to the skip randomizer option (17 out of
use the randomizer, 5 = Almost everyone uses the randomizer). This 106, 16.04%—16 of whom proceeded to assign themselves to the pos­
captured the salience of the available justification—using the random­ itive task) than those who flipped “SELF to POS” (5 out of 97, 5.15%—4
izer in itself was a choice, and participants were given the prerogative of whom proceeded to assign themselves to the positive task), B =
not to use it at all—without directly asking about how justified their − 1.26, SE = 0.53, p = .018. Thus, of those who revoked the “OTHER to
action was. POS” outcome to ultimately assign themselves the positive task, 26.67%
Next, participants completed a filler task (actually a different, un­ did so by explicitly switching decision strategies (again, refer to Fig. 5).
related scenario study). Depending on their self-assignment, they were
told they would be entered into the lottery or that the other participant 8.2.2. Justification to switch strategies via salience of the option to self-
would be. They finally indicated their age and gender. A lottery payment assign outcomes without the randomizer
was made at the end of the study to a random participant who assigned Next, we examined participants’ estimates of what others would do
him or herself to the positive consequence task. in their situation. Of participants who used the randomizer, those who
flipped “OTHER to POS” estimated that fewer participants used the
8.2. Results randomizer (M = 2.85, SD = 0.95) than those who flipped “SELF to POS”
(M = 3.12, SD = 0.94), t(201) = 2.07, p = .040. This suggests that
8.2.1. Replication of original findings participants who were tempted by a self-interested option—ignoring the
The results replicated the effects from Study 2′ s revocable random­ random outcome and assigning themselves to the positive out­
izer condition and the original paradigm (Batson et al., 1999). First, of come—tried to convince themselves that others did not even use the
the 367 participants, 197 (53.68%) chose to click the randomizer button randomizer. Believing that a minority of participants used the
at least once. Of those who did not use the randomizer, 151 (88.30%) randomizer may allow participants to think their initial use of the
assigned themselves to the positive task and 20 (11.70%) assigned the randomizer was a commitment to fairness that most others would not
other participant to the positive task (see Fig. 5 for full results). have made. This may, in turn, allow them to retroactively renege on
Of the 203 participants who ever used the randomizer, 143 (70.44%) their commitment to the coin option.
assigned themselves to the positive task, which is significantly different We added participants’ estimates of others’ use of the randomizer
from chance, p < .001. Although participants who flipped “OTHER to into the logistic regression predicting likelihood of retroactively
POS” were less likely to ultimately complete the positive task (56.60%) switching to the skip randomizer option and found that it negatively

213
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

predicted switching, B = − 0.67, SE = 0.26, p = .009; the effect of the (respectively 62% and 5%).
random outcome on the likelihood of switching was mitigated, B = The results from Study 2 further suggest that when (but not before)
− 1.11, SE = 0.54, p = .039. A bootstrapping mediation analysis (5000 their preferred outcome is denied (i.e., when they choose the random­
simulations) suggests that those who flipped “OTHER to POS” believed izer and it assigns the other participant the positive task), participants
that others were less likely to use the randomizer than those who flipped switch decision strategies to take the positive outcome for themselves.
“SELF to POS,” leading them to switch their decision strategy from using We argue that this is the result of moral rationalization; indeed, in­
the randomizer to nominally assign themselves to tasks without using dividuals who are more skilled in moral rationalization are more likely
the randomizer, 95% CI = [0.0003, 0.0400]. to revoke the randomizer’s unfavorable outcome. Study 3 identifies one
justification readily available to them in the original paradigm—the fact
that it was explicitly their prerogative to assign the tasks without flip­
8.3. Discussion
ping the coin. Study 4 provides evidence that seizing on this justification
facilitates the decision to retroactively switch decision strategies to
Our findings were largely consistent with Study 2′ s revocable
directly assign to themselves the positive task when their earlier decision
randomizer condition. We also found that participants estimated that
to use the randomizer assigned them to the negative task.
fewer others used the randomizer when they flipped the unfavorable
outcome, a belief we assume supported their desire to revoke the use of
the randomizer. We argue that this reflects their reliance on the justi­ 9.1. Theoretical contributions
fication that it was possible and normatively acceptable to assign
themselves to the self-interested outcome without even using the Previous research suggests that people engage in many self-serving
randomizer. Furthermore, this lower estimate of randomizer usage led cognitions and behaviors, including generating overly optimistic pre­
those who flipped the unfavorable outcome to be more likely to “undo” dictions about their abilities and moral behavior (Buehler, Griffin, &
their use of the randomizer—that is, to switch from using the random­ Ross, 1994; Epley & Dunning, 2016; Khan & Dhar, 2007; Klein & Epley,
izer to retroactively skipping the randomizer and assigning themselves 2016), denying their agency in self-other tradeoffs (Andreoni et al.,
an outcome directly. These results support our dynamic analysis that 2017; Berman & Small, 2012; Dana, Weber, & Kuang, 2007; Lin et al.,
proposes that people shift their moral reasoning once the randomizer 2016; Lin & Reich, 2018), and justifying self-interested or immoral ac­
yields an unfavorable outcome and that this shift allows them to justify tions (Bandura et al., 1996; Pittarello et al., 2015; Shalvi et al., 2015;
ignoring the outcome of the randomizer. We conceptually replicated this Wiltermuth, 2011). We too document an instance of motivated
effect in another study (see Web Appendix D). reasoning in the service of self-interest. The justification we highlight
here is the evocation of the counterfactual or prescriptive norm that one
9. General discussion “could have” assigned oneself to the positive task from the start. But
providing another example of how counterfactual thinking aids and
We advocate for a dynamic model of choice whereby people priori­ abets motivated reasoning (Shalvi et al., 2011) is not the main contri­
tize securing self-interested outcomes that spare them moral self- bution of the present research, albeit important. Our main contribution
reproach and only embrace purely self-interested outcomes when their is to identify and empirically support a dynamic moral decision proc­
preferred option is no longer achievable. We test this model in the ess—one where choosing to flip the coin represents one stage of the
context of the classic moral hypocrisy paradigm, and in so doing, present decision process and justification for not following the outcome of the
an additional interpretation of the morally inconsistent behavior it toss a second stage. Furthermore, the present research illuminates the
produced (Batson et al., 1999). In this paradigm, we contend that many integration of these two processes and shows their functional interde­
people choose to assign their fate to a random chance because they pendence. That is, rather than focusing on whether certain justification
aspire to achieve a positive outcome without the guilt of deliberately strategies are used, we shed light on when justification strategies are
assigning themselves that outcome; only when their desired eventuality recruited. More nuanced than earlier conclusions that justified self-
does not materialize (and when the opportunity presents itself) do they interest is the dominant option (Gino, Schweitzer, Mead, & Ariely,
resort to justification strategies to change their earlier decision to fairly 2011; Shalvi et al., 2012), we find that people opt to achieve self-
assign outcomes and impose the outcome unilaterally. Justification thus interested outcomes without needing to engage in justification tactics
serves as a back-up to their first-choice outcome: to achieve the self- (even with a 50% chance of losing the positive outcome). Ration­
interested outcome morally. alization is only resorted to when self-interest without selfishness is
Study 1 showed that choosing to flip a coin and then being randomly impossible—and only when the situation affords them the flexibility to
assigned to the positive outcome was people’s preferred outcome in the recruit such justifications.
paradigm. Because this outcome allows people to satisfy their self- In some sense, the process for which we argue could be seen as a form
interest without having to feel selfish, participants predicted that it of ability overestimation (Epley & Dunning, 2016; Khan & Dhar, 2007;
would make people happier, less ambivalent, and less guilty than would Klein & Epley, 2016) or a novel forecasting error. Our findings imply
assigning themselves to the self-interested outcome and refusing to toss that people overestimate their prosocial inclinations, and misforecast
a coin. These results support our contention that when people flip the their tendency to rationalize their self-interested behavior. Customarily,
coin in the original paradigm (Batson et al., 1999, 2002) they hope to forecasting errors are evidenced by a discrepancy between what people
"have their cake and eat it too"—that is, achieve the self-interested goal predict they will do and what they ultimately do. Although we do not
without the accompanying moral self-reproach. empirically test this in our studies, our findings suggest a similar
More importantly, in Study 2, many participants still chose to use a discrepancy between what people think they will do (i.e., follow the
randomizer even when its outcome would be automatically assigned to coin’s outcome) and what they actually do (i.e., rationalize their
them. The moral hypocrisy account would predict that participants revoking of the outcome of the coin). That is, many people (62% in our
would avoid the randomizer option when it could not be revoked. Study 2) flip the coin even when forced to follow it. We infer from this
However, we found that people were prepared to forego the positive task that most people do not predict that, when given the opportunity, they
altogether for a 50% chance of getting randomly assigned the positive would rationalize their revoking of the coin when faced with a negative
task (without having to “selfishly” choose the positive task for them­ outcome.
selves). Furthermore, those adhering to our dynamic moral shifting ac­ Our findings call into question the conclusions drawn from the
count far outnumbered those adhering to a moral hypocrisy account original research, which has been cited as evidence that people seek to

214
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Fig. 6. Theoretical decision flow chart.

take moral credit for their intentions without having to see them through conflicts for both moral and self-interested reasons. In doing so, they
(Galen, 2012; Merritt, Effron, & Monin, 2010). In contrast to this claim, may strongly contend (and even believe) that their decisions are morally
we did not find that participants perceived "using the randomizer but driven (e.g., entering a war to liberate an oppressed minority). When
revoking its outcome" to be a more moral way to assign tasks than they later do not adhere to the moral components of their decisions (e.g.,
directly assigning themselves the positive outcome without using the when war efforts turn into harmful occupation) they are viewed as
randomizer. This suggests that they did not achieve a boost to their hypocritical. Our research suggests that people often make decisions
moral self-concept by choosing to flip a coin and then blatantly dis­ that serve both themselves and others with good intentions, and may
regarding the outcome (as suggested originally, e.g., “those who fiddled even have the conviction that they will follow through on those in­
the flip may have reasoned to the point, ‘I flipped the coin, which is fair,’ tentions if tested. Thus, although observers may see the ultimate aban­
and stopped without adding, ‘But I ignored the result, which is not’”; donment of moral ideals as evidence of negative intentions from the start
Batson & Collins, 2011).10 Instead, they viewed revoking the random­ (e.g., that those others consciously use the moral reasons for decisions as
izer as similar to directly assigning themselves the positive outcome a “cover” for their self-interested intentions), they may not understand
without using the coin, supporting our contention that psychologically the psychological evolution of the decision. These decision makers may
they “undo” the coin flip and assign themselves to the positive task as if not have anticipated turning their backs on their moral principles, but
they never flipped the coin in the first place. simply hope that they will not have to be tested, and only change their
Our research also has implications for moral decision making more tune when their original reasoning no longer applies.
broadly. Leaders and policy makers may try to optimize outcomes for
themselves and for others; for instance, they may enter international 9.2. Limitations and future directions

Our research thus far does not speak to why participants lied to ex­
10
perimenters (and other participants, Batson et al., 2002) about flipping a
The discrepancy between our findings and the previous one may be because
coin if they did not internalize their behavior, and instead engaged in
participants in the original paradigm were first asked to report how they made
other justifications. Some researchers have interpreted the original
their decision (leading many to claim they had used the coin) (Batson et al.,
1999), or what method of task assignment was most morally right (Batson et al., findings as reflecting people’s self-presentational concerns—that people
2002); these questions may have led participants to report that their decision contend they have used the coin simply to appear moral to others (Shalvi
process was morally right so as not to admit that they did not adhere to the et al., 2011, 2012). Although not tested here, we agree with other re­
method that they indicated using, or the one that they indicated as “morally searchers’ interpretation that participants’ claims to have used the coin
right.” in the original paradigm reflects self-presentational concerns.

215
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

The current studies demonstrate one specific choice evolution—that CRediT authorship contribution statement
people prefer a chance to achieve self-interest without selfishness but if
chance thwarts their self-interest, they will adjust their moral reasoning Stephanie C. Lin: Conceptualization, Methodology, Formal analysis,
to justify overturning chance in favor of self-interest. Indeed, self- Investigation, Writing - original draft, Visualization. Dale T. Miller:
interest without selfishness was identified as the preferred scenario for Writing - review & editing, Resources.
most participants, and self-interest was most commonly people’s second
choice. However, people arrive at their decisions following different Declaration of Competing Interest
paths (see Fig. 6). In Study 2, a small group (5%) followed the moral
hypocrisy pattern (Path G). Some (28%) behaved out of self-interest None.
from the start (Path F) and a few (6%) prosocially from the start (Path
A); some (34%) utilized the randomizer and got lucky, assigning Appendix A. Supplementary material
themselves to the positive outcome (Path C); others (15%) utilized the
randomizer and got unlucky, but obeyed its outcome (Paths B and E). Supplementary data to this article can be found online at https://doi.
These idiosyncratic choice structures are highlighted by the prefer­ org/10.1016/j.obhdp.2021.02.005.
ence patterns in Study 1; although many held self-assignment as the
second most preferred option, many also indicated a preference for the References
flip-other outcome (e.g., perhaps those who choose Path B). We do not
claim to capture the full picture of decision patterns—we simply contend Andreoni, J., Rao, J. M., & Trachtman, H. (2017). Avoiding the ask: A field experiment
that people are morally adaptive in dynamic choice settings and on altruism, empathy, and charitable giving. Journal of Political Economy, 125,
625–653.
demonstrate one instantiation of this evolution. Generalizations based Aquino, K., & Reed, A., II. (2002). The self-importance of moral identity. Journal of
on main effects in dishonesty do a disservice to the nuanced nature of the Personality and Social Psychology, 83, 1423–1440.
complexity of people’s moral decision making. Future research should Ballard, R. (1992). Short Forms of the Marlowe-Crowne Social Desirability Scale.
Psychological Reports, 71, 1155–1160.
do more to distinguish these groups from each other.11
Bandura, A., Barbaranelli, C., Caprara, G. V., & Pastorelli, C. (1996). Mechanisms of
One notable group is made of those who prefer assigning themselves moral disengagement in the exercise of moral agency. Journal of Personality and
the positive outcome to assigning the other to the positive outcome via Social Psychology, 71, 364–374.
Batson, C. D., & Collins, E. C. (2011). Moral hypocrisy: A self-enhancement/self-
the coin flip, but also prefer obeying the coin (even at a cost to them­
protection motive in the moral domain. In M. D. Alicke, & C. Sedikides (Eds.),
selves) over defying the coin’s outcome (Path E). If this group knew Handbook of self-enhancement and self-protection (pp. 92–111). The Guilford Press.
ahead of time that they would lose the toss, they may indeed have Batson, C. D., & Thompson, E. R. (2001). Why don’t moral people act morally?
chosen to assign themselves the positive outcome without using it. Motivational considerations. Current Directions in Psychological Science, 10, 54–57.
Batson, C. D., Kobrynowicz, D., & Dinnerstein, J. L. (1997). In a very different voice:
However, having flipped the coin, they feel that this behavior would be Unmasking moral hypocrisy. Journal of Personality and Social Psychology, 72,
even worse than had they simply assigned themselves in the first place. 1335–1348.
This is in line with prior research that suggests that people commit to a Batson, C. D., Thompson, E. R., & Chen, H. (2002). Moral hypocrisy: Addressing some
alternatives. Journal of Personality and Social Psychology, 83, 330–339.
moral path once they start down it to avoid feeling hypocritical (Lin, Batson, C. D., Thompson, E. R., Seuferling, G., Whitney, H., & Strongman, J. A. (1999).
Zlatev, & Miller, 2017). Although those who lost the coin flip were less Moral hypocrisy: Appearing moral to oneself without being so. Journal of Personality
likely to obey the coin than those who won the flip, they were also more and Social Psychology, 77, 525–537.
Berman, J. Z., & Small, D. A. (2012). Self-interest without selfishness: The hedonic
likely to assign others to the prosocial task than those who chose to benefit of imposed self-interest. Psychological Science, 23, 1193–1199.
assign tasks without using the coin. Perhaps some felt trapped by the Buehler, R., Griffin, D., & Ross, M. (1994). Exploring the “planning fallacy”: Why people
coin’s outcome (and were more likely to deny themselves to the positive underestimate their task completion times. Journal of Personality and Social
Psychology, 67, 366–381.
task than if they had simply assigned themselves to tasks), whereas
Cain, D. M., Dana, J., & Newman, G. E. (2014). Giving versus giving in. The Academy of
others were willing to take the hit to their moral self-concept for Management Annals, 8, 505–533.
reneging, grasping the justifications that were available to them. Future Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative
conduct: Recycling the concept of norms to reduce littering in public places. Journal
research examining dynamic moral choice might examine when and
of Personality and Social Psychology, 58, 1015–1026.
how people justify straying from a moral path they have started down, as Dana, J., Cain, D. M., & Dawes, R. M. (2006). What you don’t know won’t hurt me: Costly
opposed to when they feel committed to that path. (but quiet) exit in dictator games. Organizational Behavior and Human Decision
Processes, 100, 193–201.
Dana, J., Weber, R. A., & Kuang, J. X. (2007). Exploiting moral wiggle room: Experiments
10. Conclusion demonstrating an illusory preference for fairness. Economic Theory, 33, 67–80.
Dreyfuss, E. (2018, December 8). A bot panic hits amazon mechanical turk. Retrieved
Most people are not inherently hypocritical—they do not wish to December 24, 2018, from https://www.wired.com/story/amazon-mechanical-turk-b
ot-panic/.
appear fair without any intention of being so. When possible, many try Epley, N., & Dunning, D. (2016). The mixed blessings of self-knowledge in behavioral
to reconcile moral and self-interested motivations—to achieve self- prediction: Enhanced discrimination but exacerbated bias. Personality and Social
interested outcomes through fair means. A majority of people are Psychology Bulletin, 32, 641–655.
Galen, L. W. (2012). Does religious belief promote prosociality? A critical examination.
willing, even, to put themselves at risk of being bound to prosocial ac­ Psychological Bulletin, 138, 876–906.
tion for a chance to achieve this outcome. Only when self-interest Gino, F., Ayal, S., & Ariely, D. (2009). Contagion and differentiation in unethical
without selfishness is no longer achievable—and when they have the behavior: The effect of one bad apple on the barrel. Psychological Science, 20,
393–398.
opportunity to change their minds—do they yield to their self-interested
Gino, F., Schweitzer, M. E., Mead, N. L., & Ariely, D. (2011). Unable to resist temptation:
motives. In short, they update their preferences in moral choice to best How self-control depletion promotes unethical behavior. Organizational Behavior and
serve their own self-image at each point in the decision process. Human Decision Processes, 115, 191–203.
Jordan, J., Leliveld, M. C., & Tenbrunsel, A. E. (2015). The moral self-image scale:
Measuring and understanding the malleability of the moral self. Frontiers in
Psychology, 6, 16–33.
Keizer, K., Lindenberg, S., & Steg, L. (2008). The spreading of disorder. Science, 322,
1681–1695.
Khan, U., & Dhar, R. (2007). Where there is a way, is there a will? The effect of future
11
One branch not represented in Fig. 6 is the group (2.6% in Study 2) who choices on self-control. Journal of Experimental Psychology: General, 136, 277–288.
used the randomizer, were assigned the positive outcome, and then assigned the Klein, N., & Epley, N. (2016). Maybe holier, but definitely less evil, than you: Bounded
self-righteousness in social judgment. Journal of Personality and Social Psychology,
other participant to the positive task. Perhaps they are concerned that a self- 110, 660–674.
interested assignment would imply dishonesty, even if they were honestly Lin, S. C., & Reich, T. (2018). To give or not to give? Choosing chance under moral
assigned the positive task. conflict. Journal of Consumer Psychology, 28, 211–233.

216
S.C. Lin and D.T. Miller Organizational Behavior and Human Decision Processes 164 (2021) 203–217

Lin, S. C., Schaumberg, R. L., & Reich, T. (2016). Sidestepping the rock and the hard Shalvi, S., Eldar, O., & Bereby-Meyer, Y. (2012). Honesty requires time (and lack of
place: The private avoidance of prosocial requests. Journal of Experimental Social justifications). Psychological Science, 23, 1264–1270.
Psychology, 64, 35–40. Shalvi, S., Gino, F., Barkan, R., & Ayal, S. (2015). Self-serving justifications: Doing wrong
Lin, S. C., Zlatev, J. J., & Miller, D. T. (2017). Moral traps: When self-serving attributions and feeling moral. Current Directions in Psychological Science, 24, 125–130.
backfire in prosocial behavior. Journal of Experimental Social Psychology, 70, Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2013). Life after p-hacking. Advances in
198–203. Consumer Research, 41, 775.
Liu, P. J., & Lin, S. C. (2018). Projecting lower competence to maintain moral warmth in Snyder, A. I., & Tormala, Z. L. (2017). Valence asymmetries in attitude ambivalence.
the avoidance of prosocial requests. Journal of Consumer Psychology, 28, 23–39. Journal of Personality and Social Psychology, 112, 555–576.
Merritt, A. C., Effron, D. A., & Monin, B. (2010). Moral self-licensing: When being good Thompson, M. M., Zanna, M. P., & Griffin, D. W. (1995). Let’s not be indifferent about
frees us to be bad. Social and Personality Psychology Compass, 4, 344–357. (attitudinal) ambivalence. In R. E. Petty, & J. A. Krosnick (Eds.), Attitude strength:
Moore, C., Detert, J. R., Klebe Treviño, L., Baker, V. L., & Mayer, D. M. (2012). Why Antecedents and consequences: 4. Ohio State University series on attitudes and persuasion
employees do bad things: Moral disengagement and unethical organizational (pp. 361–386). Lawrence Erlbaum Associates, Inc.
behavior. Personnel Psychology, 65, 1–48. Tsang, J.-A. (2002). Moral rationalization and the integration of situational factors and
Paharia, N., Vohs, K. D., & Deshpandé, R. (2013). Sweatshop labor is wrong unless the psychological processes in immoral behavior. Review of General Psychology, 6, 25–50.
shoes are cute: Cognition can both help and hurt moral motivated reasoning. Wiltermuth, S. S. (2011). Cheating more when the spoils are split. Organizational Behavior
Organizational Behavior and Human Decision Processes, 121, 81–88. and Human Decision Processes, 115, 157–168.
Pittarello, A., Leib, M., Gordon-Hecker, T., & Shalvi, S. (2015). Justifications shape Wiltermuth, S. S., & Flynn, F. J. (2013). Power, moral clarity, and punishment in the
ethical blind spots. Psychological Science, 26, 794–804. workplace. Academy of Management Journal, 56, 1002–1023.
Shalvi, S., Dana, J., Handgraaf, M. J. J., & De Dreu, C. K. W. (2011). Justified ethicality:
Observing desired counterfactuals modifies ethical perceptions and behavior.
Organizational Behavior and Human Decision Processes, 115, 181–190.

217