Episodic future thinking principally involves vivid mental simulation of future experiences, whereas semantic future thinking (SFT) may be broadly defined as general knowledge regarding the future (Atance and O’Neill, 2001; Schacter et al., 2007). As an example, the episodic/semantic distinction is captured by the difference between envisioning specific events occurring during a forthcoming vacation (episodic) versus having knowledge regarding a particularly scenic spot (semantic). Given that EFT projects the self into pre-experiencing future events, engagement in EFT may activate brain regions involved in prospective thinking (Schacter et al., 2007), thereby inducing a future orientation (Atance and O’Neill, 2001). Hence, the delay discounting-reducing effect of EFT results from the ability of prospective images to enhance either considerations of distant rewards (Benoit et al., 2011) or efforts to determine the value of delayed outcomes (Kurth-Nelson et al., 2012) in intertemporal choices. Prior research has demonstrated that EFT can reduce the tendency to discount the future (e.g., Peters and Büchel, 2010; Cheng et al., 2012; Daniel et al., 2013a,b). For example, Peters and Büchel (2010) showed that engagement in EFT induced a lower discounting rate in the context of intertemporal choice. Similarly, Cheng et al. (2012) demonstrated that participants engaging in prospective imagery showed an increased focus on future gains, in the context of financial decision-making, compared with control participants. Daniel et al. (2013a) provided experimental evidence that overweight and obese women engaging in EFT showed lower discounting rates. Daniel et al. (2013b) also showed that EFT was effective in reducing discounting tendencies in both obese and lean individuals. EFT interventions may decrease the discounting tendency by activating a future-oriented mindset (Schacter et al., 2007; Cheng et al., 2012; Chiou and Wu, 2017).

Self-theorists suggest that prospective thinking pertaining to the ideal self (i.e., an ideal version of one’s future self; Markus and Nurius, 1986) can facilitate self-regulatory efforts (Carver and Scheier, 1999). Individuals may be motivated to engage in goal-directed behaviors that aid in achieving the desired state when envisaging the ideal self (Higgins, 1996). In other words, the ideal self may motivate one to engage in self-regulatory behavior to achieve desirable aspects of the self (Freitas and Higgins, 2002). However, SFT involves only semantic knowledge about the future (Atance and O’Neill, 2001). Atance and O’Neill (2001) pointed out that the relationship between implementation intentions and goal attainment is mediated by EFT, but not by SFT. For instance, participants who were instructed merely to list positive aspects of their desired future failed to commit to goals (Oettingen, 1996, 2000; Oettingen et al., 2001). Compared with participants whose only goal was to take a vitamin C pill (SFT manipulation), those who were required to commit themselves to taking the pill at a particular time and place every day (EFT manipulation) missed fewer doses after both 10 days and 3 weeks (Sheeran and Orbell, 1999).

In principle, the ideal self involves images of a desired future (Boyatzis and Akrivou, 2006). This image is the ideal version of one’s future self (Markus and Nurius, 1986). Being primed with the ideal self may induce a future-oriented mindset, which facilitates considerations of future consequences and thereby reduces delay discounting (Chen and Vazsonyi, 2011; Cheng et al., 2012). Thus, thinking about one’s ideal self may activate a mindset that includes a future orientation, and which, thereby, reinforces tendencies toward delayed gratification. Two recent studies showed that interacting with a virtual lower-weight self (i.e., the ideal self of participants who were intending to lose weight) induced lower discounting rates (Kuo et al., 2016), while imagining future life events involving a non-smoking self (i.e., the ideal self of smokers intending to quit or reduce smoking) reduced the tendency toward discounting (Chiou and Wu, 2017). Moreover, EFT about positive events was associated with increased clarity about temporal considerations and more vivid pre-experiencing than was EFT about negative events (D’Argembeau and Van der Linden, 2004). A fundamental tendency shared by most individuals is the preferential processing of information that conveys a positive view of themselves (Taylor and Brown, 1988; Baumeister, 1998). Given that the ideal self represents a positive, desirable version of the future self, and given that the generation of possible future events is influenced by self-relevant goals (Johnson and Sherman, 1990), engagement in EFT about the future ideal self should be more likely to induce a future orientation than engagement in EFT about the future more generally. Individuals may generate images of positive or negative life events when engaging in EFT about the future in general. Taken together, the aforementioned theoretical considerations and findings suggest that an EFT intervention focusing on the ideal self would be more effective in reducing the discounting tendency than would SFT and EFT about the future more generally.

From the perspective of the association between delay discounting and crime, delinquency entails a typical intertemporal choice dilemma involving the immediate gains of delinquent behavior versus distant costs (e.g., criminal record, loss of studentship or vocation, or social ostracism). Delinquent acts may tempt people into focusing on short-term gains, such as money, sexual gratification, or excitement, but incur greater long-term costs (Hirschi, 2004). The extent to which individuals focus on the remote costs of delinquency is crucial in determining whether they will choose to commit a delinquent act or abstain from crime (Wilson and Herrnstein, 1985). Hence, the tendency to discount the future can lead to delinquency by interfering with the ability to think through the future consequences of delinquent behavior (Moffitt, 1993).

In a study by Gottfredson and Hirschi (1990), people who showed greater discounting had a reduced ability to defer gratification, leading them to engage in criminal activity. In a similar vein, Nagin and Pogarsky (2003) proposed that individuals with high discounting rates are less deterred by the delayed cost of their behaviors. Moreover, neuroimaging evidence has shown that both delay discounting (Harris et al., 2013; Hare et al., 2014) and morally questionable behavior (Greene and Paxton, 2009; Greene and Haidt, 2012) are associated with activity in prefrontal cortical regions implicated in self-control. Given that delinquency results from the inability to make deliberate, trade-off-based decisions that weigh immediate gains against the distant costs of delinquent behavior, any reduction in the tendency toward discounting induced by EFT (i.e., imagining future life events experienced by the ideal self) may in turn lower the tendency toward delinquent commitment.

Building on prior work showing that EFT can induce lower discounting rates (Peters and Büchel, 2010; Cheng et al., 2012; Daniel et al., 2013a,b; Chiou and Wu, 2017), and on the notion that a lower tendency toward discounting is associated with decreased involvement in delinquency (Gottfredson and Hirschi, 1990; Moffitt, 1993; Pratt and Cullen, 2000; Nagin and Pogarsky, 2003), we conducted two experiments to test whether engaging in prospective imagery, to pre-experience the life events that would be experienced by the ideal self, leads to lower discounting rates, thereby reducing the likelihood of making delinquent choices (Experiment 1) or cheating (Experiment 2).

A total of 90 undergraduate students (42 females and 48 males; mean age = 20.9 ± 1.4 years) enrolled at a public university in southern Taiwan participated in this experiment for extra course credit. The sample size was determined by calculating the number of participants required to satisfy the omnibus F-test (number of groups = 3) under the following conditions: α = 0.05; ω² = 0.10 (a small association; Cohen, 1988); and power (1 - β) = 0.80 (Kirk, 2012; p. 925).

Upon arrival at the laboratory, the participants were informed that they would be assisting with the piloting of several unrelated tasks that would be used in future studies. After providing written informed consent, every three participants were randomly assigned to one of the three study conditions (EFT, SFT, and control) via a permuted-block randomization schedule. Under the EFT condition, participants were instructed to write down three aspects of their ideal selves (e.g., physical, social, moral, and psychological) and then listed three life events that would occur “if the desirable aspects of the self are realized in 1 year’s time.” The 1-year period was identical to that of a subsequent delay-discounting task. The participants were then asked to close their eyes and to try to imagine the events they had listed as specifically and vividly as possible (i.e., to imagine the setting and sequence of the events, the persons and objects that would be present, etc.; D’Argembeau and Van der Linden, 2004). They had 1 min to pre-experience each life event mentally. In contrast with the EFT condition, participants under the SFT condition were instructed to write about only three events related to aspects of their ideal selves (e.g., physical, social, moral, and psychological) that would occur within 1 year. Therefore, participants in the SFT group generated representations of only their ideal selves. They did not engage in prospective imagery to mentally pre-experience the life events of the ideal self. Each participant in the SFT condition was given the same amount of time as the matched participant in the EFT condition. In accordance with the yoking procedure, the mean time spent performing the manipulation task was approximately equal between the EFT and SFT conditions. Participants under the control condition completed an unrelated questionnaire (e.g., Freitas et al., 2004; Chiou et al., 2013; Chiou and Wu, 2017; the Big Five Inventory in the current study, John and Srivastava, 1999) requiring approximately 6–10 min of their time.

The thinking manipulation task was followed by a delay discounting task, in which participants indicated their preferences in relation to winning the lottery; they made nine binary choices between either receiving $120 immediately or varying amounts of money ($113, $120, $137, $154, $171, $189, $206, $223, and $240) (Hardisty and Weber, 2009) in 1 year. To increase involvement in the discounting task, participants were told that one of them would receive a free gift certificate to a major online retailer (Joshi and Fast, 2013). They were further informed that “what you would receive is determined by random selection of one of the nine choices you make.” We calculated the discount rate, k, by employing the hyperbolic-discounting formula (see Hardisty and Weber, 2009). Larger values of k imply greater temporal discounting.

The final task was a questionnaire comprising six delinquent choice scenarios, which were presented according to how often they might be encountered in real life (van Gelder et al., 2013; Donner et al., 2014). For example, a scenario pertaining to the purchase of stolen goods online read as follows:

Imagine the following: You want the latest mobile phone but are short on cash. A fellow student forward you a link to an Internet vendor who is selling the latest models of mobile phones that ‘fell off a shipping truck.’ The mobile phones are very attractively priced. How likely is it that you would buy a potentially stolen mobile phone?

The other scenarios concerned cyberbullying (posting hurtful information online), fraud, illegal downloading, purchasing counterfeit goods, and theft. Participants recorded their responses on a 7-point scale (from very unlikely to very likely). The responses to the six scenarios were averaged to form a delinquent choice index (α = 0.86; M = 3.86, SD = 1.27). Higher scores on this index reflect a tendency to make more delinquent choices. At the end of the experiment, a probe for suspicion was administered to each participant. None of the participants accurately guessed as to how the three tasks were related.

As predicted, the discounting tendency was associated with the thinking manipulation [F_(2,87) = 5.807, p = 0.004, ηp2 = 0.118; Table 1]. Follow-up contrasts showed that participants in the EFT group discounted less than did those in the SFT (t = -2.449, p = 0.016, Cohen’s d = 0.63) and control (t = -3.277, p = 0.002, d = 0.85) groups. The discounting rate did not differ between the SFT and control groups (t = -0.828, p = 0.41). In addition, the discounting rate did not differ between male (M = 0.56, SD = 0.24) and female (M = 0.52, SD = 0.25) participants [F_(1,84) = 0.645, p = 0.424]. The interaction of condition by participant sex was not significant [F_(2,84) = 0.329, p = 0.721].

As shown in Table 1, the delinquent-choice index differed among the three study conditions [F_(2,87) = 3.815, p = 0.026, ηp2 = 0.081]. Participants under the EFT condition scored significantly lower on the delinquent-choice index than did those under the SFT (t = -2.187, p = 0.031, d = 0.56) and control (t = -2.555, p = 0.012, d = 0.66) conditions. In addition, the delinquent-choice index did not differ between male (M = 3.89, SD = 1.36) and female (M = 3.83, SD = 1.17) participants [F_(1,84) = 0.052, p = 0.819]. The experimental effect did not interact with participant sex [F_(2,84) = 0.698, p = 0.501].

Moreover, we examined whether delay discounting mediated the link between the thinking manipulation and the delinquency tendency. As the SFT and control groups did not differ in the dependent measures, these two groups were combined as the reference group for the dummy variable (1 = EFT, 0 = SFT, and control). A bootstrap analysis (Preacher and Hayes, 2004) showed that the indirect effect was significant [B = -0.54, SE = 0.18, 95% bias-corrected confidence interval (CI): -0.97 to -0.22; bootstrap resamples = 5000; Figure 1]. Engagement in EFT predicted the discounting rate (B = -0.17, SE = 0.05, t = -3.312, p = 0.001), the discounting rate predicted the delinquent-choice index (B = 3.27, SE = 0.47, t = 6.987, p < 0.001), and the connection between engagement in EFT and the delinquent-choice index (B = -0.75, SE = 0.27, t = -2.751, p = 0.007) was no longer significant (B = -0.21, SE = 0.23, t = -0.897, p = 0.372) when we controlled for the discounting rate. Thus, the mediation analysis results suggest that lower discounting, induced by engaging in EFT to envisage life events of the ideal self, leads to a reduced tendency toward virtual delinquent choices.

The results indicated that engaging in EFT pertaining to the ideal self led to lower discounting, which did not appear to be the case with SFT. Among our participants, envisaging future life events experienced by the ideal self influenced the tendency to make delinquent choices in imaginary scenarios. It should be noted that the inclusion of an SFT condition ruled out the possibility that differences in the dependent measures between the SFT and EFT conditions were due simply to thinking ahead. In other words, engagement in SFT about the ideal self was not sufficient to produce the reduction effects on discounting and delinquent choices. Furthermore, the control group was not subject to any thinking manipulation. Whether merely engaging in episodic thinking (ET) about one’s current experiences would also be sufficient to produce a similar reduction in delinquency remains unknown. Thus, a control ET condition and an ET condition pertaining to the present self were included in Experiment 2. Additionally, given that the tendency toward delinquent choices relied on a self-report measure and lacked a behavioral indicator, a behavioral measure was also employed in Experiment 2.

In total, 90 undergraduate students (45 females and 45 males; mean age = 20.2 ± 2.0 years) were recruited via posters from a public university in southern Taiwan. The sample size was determined according to the number of participants required to satisfy the omnibus F-test (number of groups = 3) under the following conditions: α = 0.05, ω² = 0.10, and power = 0.80. Thus, the required sample size was 90 (Kirk, 2012; p. 925).

Every three same-sex participants were randomly assigned to one of the three thinking conditions (EFT pertaining to the ideal self, control ET, and ET pertaining to the present self). The sex ratios were identical between the groups. Under the EFT ideal-self condition, participants engaged with the same thinking exercise (i.e., identical procedures and materials) that was employed in Experiment 1. Under the control ET condition, participants were instructed to read about three life events taken from the personal blog of a travel writer, and then engaged in ET to pre-experience non-personal autobiographical details mentally (e.g., Daniel et al., 2013a). Under the ET present-self condition, participants were instructed to list three routine life events that occurred in their present-day life, and then engaged in ET to re-experience these events. The mean time spent performing the ET task was approximately equal among the three study conditions. At the end of the thinking exercise, participants rated their simulations in terms of contextual detail and vividness of imagery using Likert-type scales (1 = not at all, 7 = very much).

After the thinking exercise, each participant completed an ostensibly unrelated survey measuring discounting tendency. The discounting measure was identical to that of Experiment 1. Under the auspices of pilot testing a future experiment, the participants were then given a worksheet of 20 number matrices; each matrix consisted of 12 three-digit numbers (e.g., 4.69; see Mazar et al., 2008). Participants were given 5 min to find the two numbers in each matrix that equaled 10.00 when summed. The time limit was insufficient to allow solving of all 20 matrices (Mazar et al., 2008; Gino et al., 2009). Participants were instructed that they would earn NT $20 (approximately US $0.66) for each correct solution.

A stopwatch notified the participants when 5 min had passed. Participants folded their worksheets and placed them in a recycling box positioned in the corner of the room. A specific three-digit number in the example matrix was matched to each participant to allow us to examine their actual performance. Participants then counted how many matrices they had solved and paid themselves for their performance. An envelope containing 20 NT $20 coins was placed on the desk before the participants arrived; they were instructed to take one coin for every matrix that they had solved correctly. The self-pay arrangement was designed to provide an opportunity for cheating. The number of coins taken by each participant represented their self-scored performance. The overstated performance (i.e., the indicator of cheating) was calculated as the difference between self-scored performance and actual performance.

There were no group differences in ratings of contextual detail [F_(2,87) = 2.095, p = 0.129] and vividness [F_(2,87) = 1.321, p = 0.272]. Table 2 shows that the discounting rate differs among the three study conditions [F_(2,87) = 4.868, p = 0.01, ηp2 = 0.101]. Planned contrasts indicated that participants engaging in EFT pertaining to the ideal self were less likely to discount the future than were those in the control ET (t = -2.522, p = 0.013, d = 0.65) and ET pertaining to the present self (t = -2.852, p = 0.005, d = 0.74) groups. The discounting rate did not differ between the latter two groups (t = -0.329, p = 0.743). As shown in Table 2, there were no significant differences in actual performance on the matrix task across the three conditions [F_(2,87) = 1.698, p = 0.189]. However, self-scored performance varied as a function of study condition [F_(2,87) = 8.279, p = 0.001, ηp2 = 0.16]. Participants engaging in EFT pertaining to the ideal self claimed to have correctly solved less matrices than did those under the control ET (t = -3.47, p = 0.001, d = 0.90) and the ET pertaining to the present self (t = -3.575, p < 0.001, d = 0.92) conditions. Self-scored performance did not differ between our two control conditions (t = -0.105, p > 0.05).

Moreover, we found that cheating was associated with the experimental condition [F_(2,87) = 3.583, p = 0.032, ηp2 = 0.076]. Table 2 shows that the mean difference between self-scored and actual performance (i.e., the overstated performance) was significantly smaller among participants under the EFT ideal-self condition than among those under the control ET (t = -2.174, p = 0.032, d = 0.56) and ET present-self (t = -2.439, p = 0.017, d = 0.63) conditions. The overstated performance did not differ between the two control conditions (t = -0.245, p > 0.05). A similar pattern was observed with regard to the percentage of participants who overstated their performance with the aim of obtaining undeserved reward [χ²_{(2, N = 90)} = 7.216, p = 0.021, Nagelkerke R² = 0.105]. Two dummy variables (EFT ideal-self vs. control ET; ET present-self vs. control ET) were created for our three study conditions, treating the control ET condition as the reference group. A logistic regression showed that participants under the EFT ideal-self condition were less likely to cheat in the matrix task (20.0%, 6 out of 30) than were those under the control ET condition (46.7%, 14 out of 30; B = -1.25, SE = 0.59, p = 0.032, Wald = 4.585, Odds Ratio = 0.29, 95% CI: 0.09–0.90); however, the likelihood of cheating did not differ between the ET present-self condition (50.0%, 15 out of 30) and the control ET condition (B = 0.13, SE = 0.52, p = 0.796).

Additionally, we examined whether delay discounting mediated the link between the thinking manipulation and cheating. As the control EFT and ET present-self groups did not differ in both the discounting rate and level of cheating, these two groups were combined as the reference group for the dummy variable (1 = EFT ideal-self, 0 = control ET and ET present-self). The relationship between engagement in EFT pertaining to the ideal self and the likelihood of cheating (B = -1.32, SE = 0.52, Z = -2.516, Wald = 6.331, p = 0.012) was no longer significant (B = -6.97, SE = 10.61, Z = -0.657, Wald = 0.431, p = 0.511) after controlling for the discounting rate. A bootstrap analysis showed that the indirect effect was significant (B = -9.72, SE = 3.32, 95% bias-corrected CI: -15.20 to -2.29; bootstrap resamples = 5000). Thus, the mediation analysis results suggest that lower discounting, induced by engaging in EFT with a focus on the ideal self, leads to refraining from cheating.

In summary, the results of Experiment 2 were consistent with those of Experiment 1, and provide support for the hypothesized connection between engagement in EFT pertaining to the ideal self and a reduced tendency toward delinquency. The mediation analysis suggested that EFT pertaining to the ideal self reduces the likelihood of cheating by shifting the time perspective of intertemporal decision-making, that is, a lower tendency toward discounting. However, the size of the indirect effect in the mediation model was moderate. It should be noted that other intervening variables (e.g., self-control, Baumeister et al., 2007) may also have affected the observed association between EFT about the ideal self and the tendency toward cheating.