1 Introduction
Dishonest behavior seems pervasive. For example, the estimated total damage to the American clothing industry from wardrobing—the habit of returning purchased clothes after wearing, amounts to $16 billion annually (Reference Speights and HilinskiSpeights & Hilinski, 2005), and the damage to US companies from employee theft and fraud reaches an estimate of $994 billion a year (Association of Certified Fraud Examiners, 2008). On an individual level, research on lying has found that people lie in some 30% of their daily interactions (Reference DePaulo, Kashy, Kirkendol, Wyer and EpsteindePaulo, Kashy, Kirkendol, Wyer & Epstein, 1996). In stark contrast to these findings, most people, including those who engage in the above practices, maintain a positive moral self-concept (Reference Aquino and ReedAquino & Reed, 2002; Reference Bem and BerkowitzBem, 1972; Reference Baumeister, Gilbert, Fiske and LindzeyBaumeister, 1998). If being moral is so highly valued in society, why then is unethical behavior so pervasive? And what determines its extent?
In this paper, we propose that the individual’s perspective is an important factor that affects moral behavior and determines its extent. We use the term perspective to indicate the size of the window through which individuals perceive and evaluate their choices. We propose that when evaluating options from a broad perspective, considering choices as embedded in a larger context of other choices and decisions, rather than as isolated instances, people place greater weight on the aggregate consequences of their actions. In contrast, when evaluating the same options from a narrow perspective, as is the case when focusing separately on each specific choice, people place greater weight on the specific consequences of the particular action.
Basic research in decision making shows that preferences are highly affected by normatively irrelevant factors such as the framing of the problem, the method of elicitation and the context in which the decision is made (Reference Bazerman, Moore, Tenbrunsel, Wade-Benzoni and BlountBazerman et al., 1999; Reference Bereby-Meyer, Meyer and BudescuBereby-Meyer, Meyer & Budescu 2003; Reference HseeHsee, 1996; Reference PaynePayne 1982; Reference ShafirShafir, 1993; Reference Slovic and LichtensteinSlovic & Lichtenstein 1983; Reference Tversky and KahnemanTversky & Kahneman 1986; Reference Tversky, Sattath and SlovicTversky, Sattath & Slovic 1988; Reference Tversky, Slovic and KahnemanTversky, Slovic & Kahneman 1990). The idea that the individual perspective affects preferences complements these findings in arguing that perspective can change preferences regardless of the specific context in which the decision is made.
1.1 Perspective and dishonest behavior
Review of research on dishonest behavior reveals two lines of research. One line of research focuses on the moral self-evaluations that underlie unethical behavior (Reference Mazar, Amir and ArielyMazar, Amir & Ariely, 2008). The primary tenet of this research is that people cheat only to the extent that allows them to maintain a self-concept of integrity. In a typical experiment of this sort, the manipulation acts directly on an individual’s self-concept by tapping into the individual’s religious and ideal values, contrasting in-group ethics with out-group ethics, and so forth (Reference Chance, Norton, Gino and ArielyChance, Norton, Gino & Ariely, 2011; Reference Gino, Ayal and ArielyGino, Ayal & Ariely, 2009; Reference Gino, Norton and ArielyGino, Norton & Ariely, 2010; Reference Mazar and ZhongMazar & Zhong, 2010; Reference Mead, Baumeister, Gino, Schweitzer and ArielyMead, Baumeister, Gino, Schweitzer & Ariely, 2009; Reference Vohs and SchoolerVohs & Schooler, 2008; Reference Shalvi, Dana, Handgraaf and De DreuShalvi, Dana, Handgraaf & De Dreu, 2011; Reference Shalvi, Dana, Handgraaf and De DreuShalvi, Handgraaf & De Dreu, 2011;Reference Barkan, Ayal, Gino and ArielyBarkan, Ayal, Gino & Ariely, 2012; Reference Shalvi, Eldar and Bereby-MeyerShalvi, Eldar & Bereby-Meyer, 2012).
Another line of research focuses on the cognitive processes underlying unethical behavior (Reference Banaji, Bazerman and ChughBanaji, Bazerman & Chugh, 2003; Reference Chugh, Bazerman, Banaji, Moore, Cain, Loewenstein and BazermanChugh, Bazerman & Banaji, 2005). The primary tenet of this research is that the same cognitive biases that affect human judgment in general affect individuals’ ethical behavior as well. In other words, people deviate from their own professed moral standards because they fail to notice that their current behavior violates those standards. Indeed, it has been shown in numerous studies that various individual, situational, and organizational factors inhibit the ability to notice the fact that one is engaged in actions that actually violate his/her own ethical standards (Reference Cain, Loewenstein and MooreCain, Loewenstein & Moore, 2005; Reference Gino and BazermanGino & Bazerman, 2009; Reference Gino, Shu and BazermanGino, Shu & Bazerman, 2010; Reference Kern and ChughKern & Chugh, 2009; Reference Tenbrunsel and MessickTenbrunsel & Messick, 2004; Reference Bazerman, Gino, Shu and TsayBazerman, Gino, Shu & Tsay, 2011).
In this research we suggest that unethical behavior is pervasive, in part because individuals usually adopt a narrow perspective: They tend to consider each choice they make in isolation, independent of all other choices. Because of this, individuals are not aware of the aggregated implications of their recurrent dishonest acts. From this argument it also follows that individuals are less likely to behave dishonestly when they consider their choices in the aggregate—from a broad perspective. The suggestion that perspective is a major determinant of dishonest behavior is in agreement with the theories discussed above. First, in line with the theory of self-concept maintenance (Reference Mazar, Amir and ArielyMazar, Amir & Ariely, 2008) people adopting a narrow perspective, considering each decision in isolation, are more likely to continually reset their moral self-evaluation measure. As such, their evaluations of their own unethical behavior are unlikely to exceed their self-accepted norms of (dis)honest behavior. Second, in line with the theory of bounded ethicality (e.g., Reference Chugh, Bazerman, Banaji, Moore, Cain, Loewenstein and BazermanChugh, Bazerman & Banaji, 2005) a broad perspective may raise the salience of unethical behavior by drawing attention to the aggregate of one’s dishonest acts.
Our research contributes to the broader program of research on the psychology of unethical behavior by examining how the individual’s own perspective, namely his/her evaluation system affects (dis)honest behavior. We test this suggestion by manipulating participants’ perspective through different choice procedures or through priming. Recent research provides stronger support for the idea that different perspectives—induced by either choice procedure or priming—yield different choice patterns: Participants who adopted a narrow perspective made riskier and more challenging choices than did participants who adopted a broad perspective (Reference Schurr, Avrahami, Kareev and RitovSchurr, Avrahami, Kareev & Ritov, 2012; Reference Schurr, Rodensky and ErevSchurr, Rodensky & Erev, 2012). For example, Reference Schurr, Avrahami, Kareev and RitovSchurr et al. (2012) demonstrating that the tendency to exert effort and to take risks depends on the adopted perspective: Participants who adopted a narrow perspective, by making sequential repeated ongoing choices tended to take greater risks and exert more effort than participants who adopted a broad perspective, by planning their next set of choices.
In sum, we propose that perspective has a profound effect on ethical behavior, and that a narrow perspective will induce more unethical behavior than a broad perspective. We tested this hypothesis in four experiments in which perspective was manipulated, either through choice procedures or through priming.
2 Experiment 1a: The effect of choice procedure
Experiment 1a was designed to test the hypothesis that perspective, as operationalized by choice procedures has an effect on the ethicality of behavior.
2.1 Method
Design and procedure. To test the relationship between the perspective that is induced by choice procedure and dishonest behavior, we used the “Is that the answer you had in mind?” trivial pursuit game.Footnote 1 The game resembles a computerized trivial pursuit style game, but the basic task is different from that in the regular game. Participants playing the game are presented with a four-alternative multiple-choice question and instructed to silently think of the correct answer. After indicating that they are ready with an answer, participants are presented with the correct answer and asked whether it was the answer they had in mind. Thus, participants may face an ethical dilemma, if the answer they thought of was incorrect, because they can profit by falsely indicating that they knew the correct answer.
The game is played in two stages, each consisting of 20 trials composed of easy and difficult questions (for schematic illustration of the procedure see Figure 1). The first stage is a “practice stage”, when participants are presented with easy and difficult questions in an alternating order. This stage familiarizes participants with the task and the difficulty level of easy and difficult questions; and provides investigators with a base level of performance when there is no monetary incentive to cheat.Footnote 2
Figure 1: The experimental design used in Experiments 1a—1c.
Following the practice stage, the participants engaged in the “actual game”. During this stage participants were paid according to their self-reported success. Reporting a correct answer to a difficult question yielded a higher reward than reporting a correct answer to an easy one (3 New Israeli Shekels (NIS) vs. 1 NIS; 1 NIS equals approximately $0.25).
At this stage we introduced a manipulation that was designed to prompt either a narrow or a broad perspective of the participant’s actions. We did it by letting participants decide how many difficult and how many easy questions they would face in one of two different ways: Participants in the Narrow condition chose the difficulty level of the upcoming question (easy or difficult) before each trial. Participants in the Broad condition chose in advance, before the start of the second stage, how many questions from each difficulty level they would face; the questions were then presented in a random order, according to their initial plan. The participants knew the difficulty level of each question they were facing. It is important to notice that, although the request to plan ahead is expected to induce a broad perspective, predicted to enhance ethical behavior, the question itself was not at all directed at the ethical aspect of each of the decisions. Furthermore, participants in both conditions performed the same task of indicating, after each question, if the correct answer was the answer they had thought of. Hence the possibility to misreport a specific answer was equally present in the two conditions.
Throughout the experiment, we tried to decrease the social concerns involved with cheating (Reference Mazar, Amir and ArielyMazar et al., 2008) by maintaining high standards of privacy, such as leaving the participants alone during the whole experiment. The only contact between the participants and the experimenter was at the beginning of the experiment, and then again at the end for payment.
It is also important to note that the nature of the task did not allow us to determine whether a certain participant cheated on any particular problem. However, because the knowledge called for in solving trivial pursuit questions was not expected to improve between the practice stage and the actual game stage, we reasoned that an increase in the reported number of “correct” answers between the two stages would indicate cheating, most likely caused by the monetary incentive. Thus, our measure of cheating was based on a comparison of the reported success in the practice stage and the actual game stage.
Participants: Sixty students participated in the experiment (30 males and 30 females). The participants were recruited through signs inviting students to participate in an interesting experiment for a monetary reward. They were randomly assigned to one of the two experimental conditions, with the goal of assigning an equal number of males and females to each condition.
Materials: The materials used in the experiment consisted of 60 four-alternative general knowledge questions. One half were classified as easy and the other half as difficult. These questions were selected from an initial larger pool of 170 questions following a preliminary study. In the preliminary study the 170 questions were partitioned into sets of 20 questions each. Every set was answered by at least 25 students. These students received a show up fee of 5 shekels and a bonus for correctly answering more than 15 questions. On the basis of the preliminary study, we chose 30 questions that were correctly solved in 68% (s.d = 0.10) of the cases and classified them as easy questions. Thirty questions that were correctly solved in 27% (s.d = 0.07) of the cases were classified as difficult. An example of an easy question is: “The Portrait of Dorian Gray”Footnote 3 is a novel by: A. Rudyard Kipling; B. Edgar Allan Poe; C. Mark Twain; D. Oscar Wilde. An example of a difficult question is: Samuel Langhorne Clemens is better known as: A. Rudyard Kipling; B. Edgar Allan Poe; C. Mark Twain; D. Oscar Wilde.
2.2 Results
To assess the extent of cheating, we calculated the change in reported success across the two stages. Specifically, for each participant, we calculated the change in the proportion of questions reported to have been answered correctly: A change score was computed separately for easy and difficult questions, and then weighted by the proportion of questions of each type in stage two, to obtain a normalized score in the range of −1 to +1. Specifically the weighted improvement measure (WIM) was calculated using the following formula:
Where
N = Total number of easy and difficult questions in the actual game stage.
ne = Total number of easy questions in the actual game stage.
e stage1 = Ratio of correct answers to easy questions in the practice stage.
e stage2 = Ratio of correct answers to easy questions in the actual game stage.
nd = Total number of difficult questions in the actual game stage.
d stage1 = Ratio of correct answers to difficult questions in the practice stage.
d stage2 = Ratio of correct answers to difficult questions in the actual game stage.
To illustrate, a participant who claimed to have solved correctly 6 out of 10 easy questions and 3 out of the 10 difficult questions in the practice stage, chose 6 easy and 14 difficult question in the actual game, and reported having solved 5 out of the 6 easy questions and 7 out of the 14 difficult questions, would get a score of:
A positive score on this measure indicates an improvement in the actual game relative to the practice stage, whereas a negative score indicates a lower success rate in the actual game relative to the practice stage. A score of +1 would indicate that the participant reported having failed to answer any of the questions in the first stage, but reported success in all the questions in the second stage. If the monetary incentive in the actual game stage had no effect on reported success, then there would be no difference between the two stages and the average change score would be 0. Finally, if the choice procedure had no effect on reported success, then the change scores in the two conditions would not significantly differ from each other.
The main results of this experiment are summarized in the left columns of Table 1. The improvement in reported success in the Narrow condition (M = 0.084, sd = 0.13) was significantly higher than that in the Broad condition (M = −0.004, sd = 0.13; t(58) = 2.57, p = 0.013) indicating that participants in the Narrow condition misreported having correctly answered more questions than did participants in the Broad condition. To test whether the change in reported success—between practice and game stage—was significant, we examined whether the WIM scores were significantly different from zero. A one-sample t-test revealed that, the mean WIM score in the Narrow condition was significantly different form zero (t(29) = 3.49, p = 0.002), whereas the mean score in the Broad condition was not (t(29) = −0.146, p = 0.885). These results show that perspective affected the tendency to engage in unethical behavior, with a narrow perspective increasing that tendency.
Table 1: Mean Weighted Improvement Measure (WIM) scores in Experiments 1 and 2.
Note
* indicates that the Mean WIM score is significantly different from zero (chance level) at the p<0.05 level.
** indicates that the mean WIM score is significantly different from zero (chance level) on the p<0.01 level.
Finally, it is worth contrasting participants’ actual payment to the expected one. To do this we compared participants’ actual payment with their expected payment had their reported solving skills not changed. As illustrated in Figure 2, relative to the expected payment, participants in the narrow perspective condition earned 4.2 NIS more than expected (M = 25.76, sd = 8.92 vs. M= 29.96, sd = 10.36 expected and actual respectively). In contrast, participants in the broad perspective condition earned only 0.5 NIS more than expected (M = 24.93, sd = 10.37 vs. M= 24.43 sd = 10.51). Repeated measures ANOVA with participants’ expected and actual payment as a within subject variable and condition as a between subjects variable revealed a significant main effect of payment (F(1,58) = 4.275, p = 0.043, η2p = 0.07) and more importantly an interaction between the experimental conditions and payoff (F(1,58) = 6.90, p = 0.011, η2p = 0.11) indicating that the difference between the expected and actual payment was greater in the narrow condition than in the broad condition.
Figure 2: Mean expected and actual payment.
Figure 2 shows the difference between participants’ actual payment and the expected payment had their reported solving skills in the first stage not changed.
3 Experiment 1b: Controlling for hindsight bias as an alternative explanation
An alternative explanation for the results of Experiment 1a is that participants who adopted a narrow perspective did not deliberately cheat, but rather after seeing the correct answer experienced the hindsight bias (Reference FischhoffFischhoff, 1975)—a “I knew it all along” kind of feeling. Experiment 1b was designed to test this alternative explanation. The method of Experiment 1b was identical to that of Experiment 1a, except that participants were also requested to write down the answer they had thought of before seeing the correct answer. This ensured that a “correct answer” could not have been evoked by a hindsight bias, because writing down an answer reflects an explicit self commitment to that answer, even if the slip of paper would never be seen by anyone else. We assume that a discrepancy between the written answer and the one reported must be recognized as such.
Participants: Forty students participated in the experiment (20 males and 20 females). The participants were recruited through signs inviting students to participate in an interesting experiment for a monetary reward. Students were randomly assigned to one of two experimental conditions with the goal of assigning an equal number of males and females to each condition. None of the participants took part in any of our other experiments.
3.1 Results
The second-left column in Table 1 presents the main results of Experiment 1b. It shows that the mean WIM score in the Narrow condition (M = 0.117, sd = 0.18) was significantly greater than the mean score in the Broad condition (M = -0.001, sd = 0.18; t(38) = 2.03, p = 0.049) indicating that participants in the Narrow condition misreported more correctly solved questions than did participants in the Broad condition. As in Experiment 1a, to test whether the increases in participants’ reports in the two experimental conditions were above chance level we tested whether their WIM scores were significantly greater than zero. Once again, the WIM score was significantly greater than zero in the Narrow condition (t(19) = 2.85, p = 0.010), but not in the Broad condition (t(19) = −0.25, p = 0.981) indicating that, as in Experiment 1a, only participants in the Narrow condition cheated. Finally, to verify that the experimental manipulation in Experiment 1b yielded the same pattern as in Experiment 1a, we compared the experiments by an ANOVA model with WIM as the dependent factor and experiment (1a, 1b) and condition (Narrow, Broad) as between-subjects factor. The analysis revealed a significant intercept (F(1,96) = 9.697, p = 0.002, η2p = 0.092) showing that on average participants in both experiments reported more correct answers in the second, for pay stage than in the practice stage. A significant effect of Condition (F(1,96) = 10.61, p = 0.002, η2p = 0.10) reflects the fact that participants in the Narrow condition showed greater increase than participants in the Broad condition. Importantly, neither Experiment (F(1,96) = 0.305, p = 0.582, η2p =0.00) nor the interaction between Experiment and Condition F(1,96) = 0.22, p = 0.638, η2p =0.00) yielded significant results, indicating that with respect to our main experimental manipulation, the two experiments yield similar pattern of results. Thus, we can rule out hindsight bias as an alternative explanation.
Finally, as in Experiment 1a it is worth contrasting participants’ actual payment to the expected one. As illustrated in Figure 2, relative to the expected payoff, participants in the narrow perspective condition earned 6.25 NIS more than expected ( M = 20.80 sd = 9.79 vs. 27.05 sd = 9.45 expected and actual respectively). In contrast, participants in the broad perspective condition earned 0.5 less than expected (M = 23.95 sd = 9.46 vs 23.45 sd = 8.79 expected and actual respectively). Repeated measures ANOVA with participants’ actual and expected payment as a within subject variable and condition as a between subjects variable revealed a marginally significant main effect of payment (F(1,38) = 3.65, p = 0.064, η2p =0.09) and more importantly an interaction between the experimental conditions and payment (F(1,38) = 5.03, p = 0.031, η2p =0.12) indicating that the difference between the expected and actual payment was greater in the narrow condition than in the Broad condition.
Both Experiments 1a and 1b provided participants with monetary incentives to misreport the correctness of their answers. Would the observed effect of perspective still hold without such incentives? Experiment 1c addressed this question.
4 Experiment 1c: The effect of monetary incentive on dishonest behavior
The method of Experiment 1c was again identical to that of Experiment 1a, except that participants were awarded a flat fee of 20 NIS, regardless of their reported success.
Participants: Forty-eight students participated in the experiment (24 males and 24 females). The participants were recruited through signs inviting students to participate in an interesting experiment for a monetary reward. They were randomly assigned to one of the two experimental conditions, with the goal of assigning an equal number of males and females to each condition. None of the participants took part in any of our other experiments.
4.1 Results
The next-to-last column of Table 1 presents the results of Experiment 1c. Again, there was a significant difference in WIM scores. The mean WIM score under the Narrow condition (M = 0.004, sd = 0.10) was significantly higher than the mean WIM score under the Broad condition (M = −0.062, sd = 0.12; t(46) = 2.136, p = 0.038). The mean WIM score under the Narrow condition was not significantly different from zero (t(23) = 0.212, p = 0.834), whereas whereas the mean WIM score under the Broad condition was significantly lower than zero (t(23) = −2.599, p = 0.016). Although this difference was not predicted, it may indicate that unethical behavior in the practice stage was diminished by the broad manipulation in the actual game stage. Still the difference between choice procedures was also evident here. It is worth noting that the overall change in reported success was lower in both conditions than in Experiments 1a and 1b. An ANOVA with WIM as the dependent factor and Experiment (1a and 1b vs. 1c) and Condition (Narrow, Broad) as between-subjects factors revealed a significant main effect of Experiment (F(1,144) = 9.557, p = 0.002, η 2p =0.062), a significant main effect of Condition, (F(1,144)=11.34, p = 0.001, η 2p = 0.073) and no significant interaction between the two (F(1,144) = 0.466, p = n.s, η 2p = 0.003) indicating that a narrow perspective increases dishonesty in the presence of a monetary incentive.
Finally, as in previous the experiments we contrasted participants’ actual payment to the expected one. As illustrated in Figure 2, relative to the expected payment, participants in the narrow perspective condition earned 0.96 NIS less than expected (M = 19.67 NIS sd = 5.10 vs. M = 18.71 sd = 5.61 expected and actual respectively). Participants in the broad perspective condition earned 1.91 NIS less than expected (M = 19.79 sd = 5.13 vs. M = 17.88, sd = 6.97 expected and actual respectively). Repeated measures ANOVA with participants’ actual and expected payment as a within subject variable and condition as a between subjects variable revealed a significant main effect of payment (F(1,46) = 5.51, p = 0.023, η2p = 0.11) and more importantly a non significant interaction between the experimental conditions and payment (F(1,46) = 0.61, p = 0.438, η2p = 0.05) indicating that under the flat fee condition the experimental conditions did not have a differential effect on participants’ earnings
In sum, with no monetary incentive, the overall level of reported success dropped significantly between the practice stage and the actual game stage. Furthermore, to the extent that some misreporting occurred even in the practice stage (as indicated by higher reported scores in the practice stage relative to the preliminary test), our findings suggest that engaging in planning reduced this propensity below that of the practice stage, even when no monetary incentive was offered.
5 Discussion
Taken together the results of the three experiments provide a coherent picture with regard to ethical behavior: The results of Experiment 1a showed that a narrow perspective led to less ethical behavior than a broad perspective. The results of Experiment 1b replicated these results and ruled out hindsight bias as an alternative explanation. Finally, the results of Experiment 1c indicated, not surprisingly perhaps, that the presence of some external motivation—monetary in our case—might be a pre-requisite for the emergence of unethical behavior.
6 Experiment 2: The effect of priming
The results of our first three experiments suggested that individuals are more likely to act dishonestly when the choice procedure allows them to segregate their decisions. We hypothesized that the initial requirement to choose the number of easy and difficult questions in the Broad conditions would trigger an integrative broad perspective, in contrast with the narrow perspective in the Narrow conditions. Although the possibility to behave unethically existed in both conditions, the initial requirement to consider the whole set of questions evoked a broad perspective, and thus affected participants’ ethicality. In Experiment 2, we tested the generality of the broad vs. narrow perspective effect by applying another manipulation—one of priming. All participants performed the same trivial-pursuit style game under the Narrow choice procedure, but some participants were primed to adopt a broad, high-level perspective and others were primed to adopt a narrow, low-level perspective.
6.1 Method
Experiment 2 also employed the Is that the answer you had in mind? trivia game used in the Narrow condition of Experiment 1a. We employed a shorter version that consisted of only 10 questions in each stage. The priming manipulation was introduced before the second, real-game stage, involved and was varied on three levels: Broad, Narrow 1Q and Narrow 10Q. All participants received a sheet of paper and read the following: “Sometime in the near future we plan to include additional questions on geography—a topic you will not be asked about in the current experiment. We would appreciate it if you could spare a couple of moments of your time to answer the following question(s).” Participants in the Broad condition were asked to rate their knowledge of geography on a 10-point scale ranging from poor knowledge to excellent knowledge. Participants in the Narrow 1Q condition were asked to answer one multiple-choice question concerning the name of a European capital. Participants in the Narrow 10Q condition were asked to answer 10 multiple-choice general knowledge questions on geography. We expected broad condition priming to evoke a different perspective than the two forms of narrow condition priming, because estimating one’s own knowledge is a higher, more abstract process than answering a specific question. Under all priming conditions, the participants chose the difficulty level of the upcoming question (easy or difficult) before each trial (as in the Narrow condition of Experiment 1a).
Participants: Ninety-four students participated in the experiment (32 in the Broad, 30 in theNarrow 1Q and 32 in the Narrow 10Q conditions). The participants were recruited through signs inviting students to participate in an interesting experiment for a monetary reward. They were randomly assigned to one of three experimental conditions, with the goal of assigning an equal number of males and females to each condition. None of the participants took part in any of our other experiments.
6.2 Results
The increase in levels of reported success was the same for the Narrow 1Q and the Narrow 10Q conditions. The WIM score in the Narrow 1Q condition was 0.17, whereas the mean score in the Narrow 10Q condition was 0.18, t(60) = 0.11, p = 0.916). Because there was hardly any difference between the conditions, we combined the two Narrow conditions and compared performance in them to that in the Broad condition.
As in the previous experiments, our main dependent measure was the change in reported success from the first (practice) stage, to the actual game stage. The last column of Table 1 shows that the mean improvement (WMI) in the Narrow priming conditions (M = 0.178, sd = 0.28) was larger than that in the Broad priming condition (M = 0.04, sd = 0.19). A comparison of the WMI in the Narrow and Broad conditions revealed the difference between conditions was significant (t(85) = 2.11, p = 0.038; a test for unequal variances was required). In addition, as in Experiments 1a-1c, the overall improvement in the Narrow priming conditions was significantly different from zero (t(61) = 5.01, p = 0.001), whereas the overall improvement in the Broad condition was not significantly greater than zero (t(31) = 1.16, p = 0.252). These results are consistent with those of Experiment 1 and provide further evidence that people’s perspectives toward their actions affect the ethical choices that they make.
Finally, as in Experiments 1a-1c we contrasted participants’ actual payment to the expected one. As illustrated in Figure 2, relative to the expected payment, participants in the Narrow priming condition earned 4.49 NIS more than expected (M = 11.11 sd = 6.42 vs. M = 15.60 sd = 5.31 expected and actual respectively). Participants in the Broad priming condition earned 1.21 NIS more than expected(M =11.13 sd = 5.21 vs. M = 12.34 sd = 5.02 expected and actual respectively). Repeated measures ANOVA with participants’ expected and actual payment as a within subject variable and condition as a between subjects variable revealed a significant main effect of payment (F(1,92) = 14.68, p < 0.000, η2p = 0.14) and a significant interaction between the experimental conditions and payment (F(1,92) = 4.81, p = 0.031, η2p = 0.05) indicating that the difference between the expected and actual payment was greater in the narrow priming condition than in the Broad priming condition.
7 General discussion
Our research examined the effect of people’s perspectives on their dishonest behavior. Our main hypothesis was that dishonest behavior is more likely to occur under narrow perspectives, when people consider each choice they make in isolation, rather than under broad perspectives, when people consider the aggregate consequences of their choices. Experiment 1a—in which we introduced a new indirect measure of assessing dishonest behavior, the Weighted Improvement Measure (WIM)—provided initial evidence in support of our hypothesis. Its results demonstrated that dishonest behavior is more pervasive when people can segregate their choices (e.g., when they make sequential choices), than when they plan ahead, making an initial aggregate choice. This result was evident although the actual decision to be dishonest was made separately for every question under both perspective manipulations. Experiment 1b tested whether the effect observed in Experiment 1a was due to a hindsight bias, the “I knew it all along” feeling that is so common in retrospect. Although participants were required to write down their answers before reporting their success or failure, the results replicated the findings of Experiment 1a, thus ruling out an explanation in terms of hindsight bias. Experiment 1c tested whether the effect of perspective on ethical behavior occurs only when there is an external motivation to engage in unethical behavior. The results showed that, although overall removing the monetary incentive diminished unethical behavior, the effects of perspective on this level persisted even in the absence of such incentives. Finally, Experiment 2 manipulated perspectives through priming. The results of this manipulation yielded a similar pattern of results to those obtained in Experiments 1a-1c. In addition, the results of this experiment indicate that the tendency to report higher than expected success under narrow perspective is not an artifact of making sequential choices.
Our research contributes to the growing literature on the psychology of dishonest behavior by identifying perspective as an important factor affecting the extent of such behavior. Current theories of unethical behavior can accommodate the role of perspective. For example, in line with bounded ethicality theory (Reference Chugh, Bazerman, Banaji, Moore, Cain, Loewenstein and BazermanChugh et al., 2005; Reference Bazerman and TenbrunselBazerman & Tenbrunsel, 2011), we submit that violations of ethicality are more likely to go undetected under narrow perspective. Furthermore, segregation, prompted by a narrow perspective, decreases the chances of considering the cumulative implications of one’s unethical behavior, hence its effect on unethical behavior is compatible with the predictions of self maintenance theory (Reference Mazar, Amir and ArielyMazar et al. 2008). The current research also contributes to the literature on joint vs. separate evaluations, by showing that people tend to behave more unethically in a situation that allows them to segregate their ethical choices.
Dishonest behavior seems to involve many social concerns (Reference Aquino and ReedAquino & Reed, 2002). The advantage of the paradigm we developed is that participants can behave dishonestly without the risk of being exposed. A drawback of this paradigm is that we cannot tell whether any of the 242 students who participated in our experiments actually cheated on any single question. Nevertheless, we can indirectly assess the extent of an individual’s dishonest behavior by comparing his/her performance in the training stage, in which there is no monetary incentive to cheat, with performance in the actual game stage, in which there is a monetary incentive to cheat. Our conclusions are therefore restricted to the group level. We believe, however, that this restriction does not prevent us from testing (and corroborating) the research hypotheses. In other words, the fact that a greater than expected rate of improvement was observed under conditions that promoted a narrow perspective but not under conditions that promoted a broad perspective supports our hypothesis.
The results of our research suggest that perspective may be induced by manipulations of choice procedure or by priming. This opens opportunities for real-world applications. For instance, organizations may try to reduce the “borrowing” of office supplies by workers simply by asking them to estimate in advance the supplies they will need over a certain period of time. And managers and workers may use other procedures that induce a broader view of employees’ roles to promote ethical behavior.
Our findings suggest that a broad perspective promotes ethical behavior partly because as an aggregate, one’s unethical choices are likely to exceed one’s ethical standards. But the question whether there are situations in which the broad perspective actually promotes unethical behavior remains open. For example, if people realize that their aggregate choices still fall below an ethical norm, then would that induce, rather than inhibit future unethical behavior? Research on moral licensing, which shows that the choice to behave morally is a balancing act between the desire to maintain a positive self concept and the cost of doing so (e.g., Reference Sachdeva, Iliev and MedinSachedeva, Iliev & Medin, 2009; Reference Shalvi, Eldar and Bereby-MeyerShalvi, 2012; Reference Shalvi, Dana, Handgraaf and De DreuShalvi, Handgraaf & De Dreu, 2011), suggests that taking the broader perspective could, at times, also have a detrimental effect. We leave this question open for future research.
Finally, in the current research we demonstrated that considering a decision in the context of other decisions leads to fewer dishonest acts than when the decision is considered in isolation. A related, but different effect that has been extensively researched is the effect of joint vs. separate evaluation of options. This research shows that when people evaluate one option at a time as they do in separate evaluations they often arrive at different decisions than when they evaluate several options simultaneously as they do in joint evaluations (Reference Bazerman, Tenbrunsel and Wade-BenzoniBazerman, Tenbrunsel & Wade-Benzoni, 1998); Reference Bazerman, Moore, Tenbrunsel, Wade-Benzoni and BlountBazerman, Moore, Tenbrunsel, Wade-Benzoni & Blount, 1999; Reference Irwin, Slovic, Lichtenstein and McClellandIrwin, Slovic, Lichtenstein & McClelland, 1993; Reference Ritov and BaronRitov & Baron, 2010; Reference Slovic, Finucane, Peters, MacGregor, Gilovich, Griffin and KahnemanSlovic, Finucane, Peters & MacGregor, 2002;). The main account that has been proposed to explain the joint/separate evaluation rests on the comparative nature of the joint evaluation, making the evaluability of certain attributes clearer. By contrast, the effect of perspective that we studied, considering each decision in the context of other, similar decisions, turns the DM’s attention not to the comparison between different decisions, but rather to the overall pattern that emerges.
8 Conclusions
People like to think that their ethical standards are firmly rooted in inherent values, and that they have a fixed threshold beyond which any behavior would be unacceptable. The current research provides new evidence that the threshold of unacceptable behavior is not fixed. Rather, it depends on the perspective through which people view their actions. Actions that are made under a broad perspective tend to be evaluated by more stringent standards than actions that are made under a narrow perspective.
Open access
