Regret-Based Decision-Making Style Acts as a Dispositional Factor in Risky Choices

Abstract

People who anticipate the potential regret of one’s decisions are believed to act in a more risk-averse manner and, thus, display fewer risk-taking behaviors across many domains. We conducted two studies to investigate whether individual differences in regret-based decision-making (a) reflect a unitary cognitive-style dimension, (b) are stable over time, and (c) predict later risk-taking behavior. In Study 1, 332 participants completed a regret-based decision-making style scale (RDS) to evaluate its psychometric qualities. In Study 2, participants (N = 119) were tested on two separate occasions to assess the association between RDS and risk-taking. At Time 1, participants completed the RDS, as well as trait measures of anxiety and depression. One month later, they completed the Balloon Analogue Risk Task (BART) and state mood (Positive/Negative affect) scales. The RDS had a sound unidimensional factorial structure and was stable over time. Further, higher reported RDS scores were significantly associated with less risk-taking on the BART, holding other variables constant. These studies suggest that individual differences in regret-based decision-making may lead to a more cautious approach to real-world risk behaviors.

Keywords

Cognitive style decision-making Balloon Analogue Risk Task individual differences regret risk negative affectivity anxiety depression

Introduction

The decisions that we face in everyday life often lack a clear-cut superior choice and ultimately involve trade-offs. For instance, choosing where to go on vacation, accepting or rejecting a job offer, includes options that may have both positive and negative qualities ascribed to them. Moreover, the decision itself involves uncertainty: One might not know the likelihood that he or she will be satisfied with one’s choice. Although the choices may vary in content, research suggests that stable patterns of decision styles may guide individuals’ choices across domains. For instance, some individuals are inclined toward analyzing the pros and cons of each decision they make; others have an immediate feeling of what to do; and still, others try to make the choice that minimizes future negative feelings in the event that one’s decision did not turn out as expected (i.e., avoiding future regrets).

Despite research that has implicated regret as an important emotional input in decision-making processes, dispositional tendencies in regret-based decision-making have received less attention (George & Dane, 2016; Lerner, Li, Valdesolo, & Kassam, 2015; Phillips, Fletcher, Marks, & Hine, 2016; Wang, Highhouse, Lake, Petersen, & Rada, 2017). Construct validity for a personality disposition should be supported by different types of evidence. First, a scale should include a sample of items that adequately conceptualize the construct. Second, there must be some stability of personality ratings over time. Third, these ratings must be predictive of relevant behaviors. Before empirically examining these requirements, we briefly frame regret style within the literature on cognitive and decision styles. Then, we motivate why risky choices may be an ideal domain within which one can test predictive hypotheses essential for establishing regret as disposition.

Cognitive style and decision-making

Broadly, cognitive style describes the habitual ways in which people process information in everyday situations (Kozhevnikov, 2007). In the management science and organizational psychology literature in particular (Kozhevnikov, Evans, & Kosslyn, 2014), cognitive styles have evolved more specifically into “decision-making styles,” which represent the degree to which individuals tend to construe a decision problem, search for information, and choose for alternative options in a relatively stable way across time and situations (Phillips et al., 2016; Wang et al., 2017).

Several dispositional decision-making styles have been proposed (Betsch & Iannello, 2010; Hamilton, Shih, & Mohammed, 2016; Pacini & Epstein, 1999; Scott & Bruce, 1995; Sjöberg, 2003). One prominent model of decision styles has been inspired by dual-process theories (Evans & Stanovich, 2013), which suggests an interplay of two systems that facilitate either intuitive, associative reasoning (System 1) or reflective, deliberative processing (System 2). Accordingly, an intuitive decision maker tends to make choices based on immediate feelings about the key elements of a decision, whereas a reflective decision maker may be more likely to evaluate the options and their attributes thoroughly and to make rule-based decisions (Kozhevnikov et al., 2014; Phillips et al., 2016; Wang et al., 2017).

However, research has acknowledged that two primary decision-making styles may not adequately account for diversity in how individuals typically approach decisions. For example, Scott and Bruce (1995) proposed a typology, which incorporated dependent, avoidant, and spontaneous styles. Similarly, Mann, Burnett, Radford, and Ford (1997) considered buck-passing, procrastination, and hypervigilance as additional constructs to consider. More recently, Leykin and DeRubeis (2010) developed five additional styles, namely, confident, spontaneous, anxious, regretful (or brooding), and respected. Therefore, although intuition and reflection are both prominent examples of the habitual ways in which people make their decisions, they cannot be considered exhaustive of all decision styles.

Regret-based decision-making

Regret is a comparison-based negative emotion of self-blame that people experience when they realize or imagine that their present situation would have been better if they had decided differently in the past (Pieters & Zeelenberg, 2007; Zeelenberg & Pieters, 2007). By definition, regret is inextricably intertwined with decision-making and has received considerable empirical attention in research that has examined how emotion impacts decision processes (George & Dane, 2016; Lerner et al., 2015). Regret regulation theory suggests that experienced regret is unique from other emotions in that it included elements of counterfactual thinking (i.e., considering alternative courses of actions and their outcomes) and assumed personal agency for one’s choices (Pieters & Zeelenberg, 2007; Zeelenberg & Pieters, 2007). In addition, regret can be experienced about past (i.e., realizing that a discarded option was better than a freely chosen one) or future decisions (i.e., imagining that an available choice could be better than another that one would like to pick). Therefore, dispositional tendencies to consider regret may influence decision-making at multiple stages of a decision process. Specifically, predecisional regret has a high informative valence in the appraisal of risks and benefits (Brewer, DeFrank, & Gilkey, 2016). In contrast, postdecisional regret plays an essential role in learning from the past, in turn updating the informative value of anticipated regret for future decisions (DeWall, Baumeister, Chester, & Bushman, 2014).

Consistent with regret regulation theory, Nygren (2000) posited that some people could be more inclined toward protecting themselves against regret feelings in decision-making, while others could be less so. A regret-based decision maker is expected to choose the option that reduces regret or to avoid making decisions due to the fear of regret (Nygren, 2000). To assess this style, Nygren and White (2002, 2005) used the Decision-Making Style Inventory (DMI), which also measured the tendency to use intuition and reflection. Different studies supported the validity of the regret style. First, although the regret style was positively associated with reflection and negatively with intuition, the correlations were small (Nygren & White, 2002). Therefore, regret style did not overlap with the two main styles of decision-making, a finding that supports the independence of the construct. Second, individuals reporting higher levels of regret style also held related beliefs about their decision-making tendencies, such as experiencing difficulties making choices (Djulbegovic et al., 2015; Rim, Turner, Betz, & Nygren, 2011; Shaffer, Tomek, & Hulsey, 2014; Turner, Rim, Betz, & Nygren, 2012), and expressing self-doubt about one’s problem-solving and decision-making ability (Morera et al., 2006; Nygren & White, 2002). In addition, other DMI studies suggested that individuals who tend to focus on regret tend to be more indecisive or avoidant when making choices (Elaydi, 2006; Leykin & DeRubeis, 2010; Morera et al., 2006; Nygren & White, 2002). Last, regret style was positively associated with affective traits such as harm avoidance, anxiety, depression, and neuroticism (Dewberry, Juanchich, & Narendran, 2013; Leykin & DeRubeis, 2010; Nygren & White, 2002, 2005).

Regret style and risk-taking

By definition, regret would not be experienced if one knew what the outcomes of each possible outcome would be with 100% certainty, assuming that an individual had an ordered preference of desired outcomes. However, a majority of choices that we face each day involve uncertain or risky outcomes. From a behavioral economics perspective, risk-taking involves choosing an available option with a higher degree of variability than its alternatives (e.g., in a decision that involves a choice that offers $100 for sure and one that offers a 50% chance to win $200, otherwise $0). It follows that the presence of uncertainty increases the possibility that an undesired outcome may occur. Hence, risk evaluation processes are likely to be part of the nomological network of the regret-based decision-making style construct.

Several studies support this assertion. For instance, anticipating regret for the consequences of not using a condom predicted greater condom use intentions among adolescents, young adults, and drug users (Richard, van der Pligt, & de Vries, 1996; Smerecnik & Ruiter, 2010; van Empelen, Kok, Jansen, & Hoebe, 2001). Likewise, anticipated regret predicted college students intention to limit alcohol consumption and less binge-drinking in the past (Cooke, Sniehotta, & Schüz, 2007), as well as childbearing age women intention to abstain from drinking while pregnant (Vézina-Im & Godin, 2011). Anticipated regret was also a protective factor for adolescent smoking initiation (Conner, Sandberg, McMillan, & Higgins, 2006) and predicted adult daily smokers cessation intentions (Janssen, Waters, Van Osch, Lechner, & De Vries, 2014; Lazuras, Chatzipolychroni, Rodafinos, & Eiser, 2012). Moreover, anticipated regret for speeding (i.e., fine) was associated with lower reported driving speeds (Elliott, Thomson, Robertson, Stephenson, & Wicks, 2013) and for using prohibited substances to enhance sport performance with intention to dope (Lazuras, Barkoukis, Mallia, Lucidi, & Brand, 2017). Anticipated regret has also been linked to both risk perception and intentions to partake in a variety of gambling activities (Li et al., 2010). Although the type of gamble (e.g., sport bets, horse racing, card games) moderated the relationship between anticipated regret and intention to gamble, anticipated regret was correlated with higher risk perception, and both these variables were uniquely associated with intention to gamble (Li et al., 2010).

By contrast, regret regulation theory maintains that people strive to avoid regret and predicts risk-seeking or risk aversion depending on which option protects the most from feedback on the foregone outcome (Pieters & Zeelenberg, 2007; Zeelenberg & Pieters, 2007). For instance, in a classic study, in which risky and safe gambles were offered to the decision maker, participant’s risk attitude was based on which of the two options he or she expected to receive feedback, regardless of either outcome valence (gains vs. losses) or risk level (Zeelenberg, Beattie, Van Der Pligt, & De Vries, 1996). Likewise, there was a higher propensity to buy lottery tickets if a winner could have been identified, at least roughly, based on the postcode of the area of residence than if it was impossible for the losers to trace the winner (e.g., designated by a printed code on a ticket that no one knows; Zeelenberg & Pieters, 2004). Moreover, anticipated regret for knowing that the winner was in one’s neighborhood of residence but one has not participated in the lottery mediated this propensity. Therefore, risk-seeking for gains was the most frequent choice in situations, in which participants anticipated to have a feedback on a discarded risky option (Zeelenberg et al., 1996; Zeelenberg & Pieters, 2004).

Other studies (Tochkov, 2009, 2012) also examined how anticipated regret affected risk preferences using a computerized laboratory task in groups of social and problem gamblers. Although social gamblers were more regret-averse than risk-averse, problem gamblers more strongly preferred the proposed risky alternatives and were less guided by anticipated regret in their choices (Tochkov, 2009, 2012). Furthermore, social gamblers were more accurate than problem gamblers in predicting their regret feelings in a gambling task, although the negative mood biased their anticipated regret.

Research hypotheses and overview of the studies

In this study, we aimed to conceptualize the construct validity of a regret-based decision-making style better. To address this issue, we devised two independent studies, testing the requirements of stability, content, and predictive validity that would support the existence of this construct. Study 1 examined the dimensionality and stability of regret style using selected items tapping into the hypothesized dispositional style. One primary question was the degree to which predecisional and postdecisional regret reflect a unitary construct. Then, we examined whether regret-style ratings remain constant or change over a one-month period. Study 2 examined the predictive validity of the regret-based decision-making style. Over a one-month interval, we assessed regret style (T1) and performance on a frequently used behavioral risky decision-making task, the Balloon Analogue Risk Task (BART; Lejuez et al., 2002). In this task, individuals are asked to pump a computerized balloon, in which they gain points for each pump. With each pump, the balloon gets bigger, but also does the potential that it will burst, which would subsequently result in the loss of all winnings for that trial. Participants make sequential choices between a sure outcome (i.e., their current level of winnings) and a risky gamble (i.e., making an additional pump, but with an increased chance to lose all winnings). Our previous research has found a positive effect of anticipated regret on risk aversion on BART (Panno, Lauriola, & Pierro, 2015). In this study, we examine whether regret style can be predictive of risky choices on this task. Because of their reported associations with both regret and risk perceptions, we also assessed and controlled for affective traits such as harm avoidance, anxiety, depression, and emotional stability (T1) and current state positive and negative affect (T2) to strengthen our test of unique predictive validity.

Study 1: Dimensionality and stability of regret style

Participants and procedure

A total of 332 participants took the Regret Decision Style (RDS) scale. We collected data from two different samples. Undergraduate students at Sapienza University of Rome (N = 143; M_age = 24.66 years, SD = 2.52 years; 69% women) participated in this study for partial course credits and completed the scale in small-group sessions of about eight people in a laboratory room. Fifty-nine participants (M_age = 24.26 years, SD = 1.32 years; 81% women) also completed the RDS one month apart. Community participants (N = 189; M_age = 31.97 years, SD = 11.11 years; 58% women) took the RDS scale as a part of a larger study on individual differences in decision-making. This group voluntarily participated and was recruited during leisure time nearby museums, archaeological, or art sites in the city center. The student and the community groups had statistically different mean age, t(213.297) = 8.75, p < .001, equal variance not assumed, and gender composition, χ²(1) = 36.44; p < .001. The local ethical review board at Blinded for Review approved the study.

Measures

Regret style

Using a standard forward/backward translation procedure, we translated the DMI¹ (Nygren, 2000) into Italian. Because the original DMI regret-style subscale comprised 15 items, the RDS used in this study was abridged to make it less time-consuming for use in laboratory experiments and larger test batteries. In so doing, we reviewed item-total correlations obtained from an unpublished data set in which we administered the full DMI.² The items that correlated the most with the total score were short-listed, and their content was examined by the first and second authors to ensure content preservation. As a result of this process, nine items were retained. Regarding content coverage, six items described postdecisional regret and three items described predecisional regret (Figure 1).

Figure 1.

Confirmatory factor analysis model of the Regret Decision Style scale. Note: The circle represents a latent variable. Squares represent measured indicator variables. Numbers attached to the arrows going from the latent variable to the measured indicators are standardized factor loadings. Numbers attached to other arrows are error terms variances. The text to the right of each measured indicator describes the content of each RDS item. The text in square brackets specifies whether the content refers to predecisional or postdecision regret. The text in round brackets specifies the item number in the Decision-Making Style Inventory (Nygren, 2000).

Statistical analyses

Single group analysis

A confirmatory factor analysis model with a single regret-style latent variable and nine indicator variables was fitted to the data. For comparison, we also fitted a two-factor model with predecision and postdecision regret latent variables. We used EQS 6.1 in all analyses (Bentler, 2004). Specifically, the maximum likelihood robust method was used to estimate the model parameters, as the data significantly violated the assumption of multivariate normality (Mardia’s normalized coefficient = 11.44). Since any factor model could be rejected, based on the inspection of the model chi-square, we used both absolute (i.e., the root mean square error of approximation, RMSEA) and relative fit indexes (i.e., the comparative fit index, CFI; and the Bentler–Bonett non-normed fit index, NNFI) to assess the model fit. For RMSEA, values lower than .05, between .05 and .08, and higher than .10 indicate good, acceptable, and poor model fit, respectively. For CFI and NNFI, values greater than or equal to .90 support an acceptable model fit; values higher than .95 are strongly recommended (Byrne, 2006).

Multiple group analyses

In the single group analysis, we merged data from two different samples to have an acceptable sample size for a CFA study. The single group analysis assumed that we had drawn the student sample and the community sample from the same population. However, the student and the community sample had significantly a different gender and composition. Through multigroup factor analyses, we tested whether the regret-style factor model fitted to the whole sample could be generalized to different subpopulations. According to Byrne (2008), an initial step requires only that the same number of factor and factor loadings be the same across groups, namely configuration invariance. This analysis, which represents the most lenient form of invariance, also serves as a baseline onto which future analyses can be compared, each imposing more stringent forms of invariance. Typically, primary interest focuses on the factor loadings invariance because the tests for the equivalence of error variances and covariances are typical “overly restrictive” (Byrne, 2006, p. 227).

Results

Dimensionality of the regret style

Although the single-factor model was statistically significant, SBχ²= 75.54; df = 27; p < .001, there was an acceptable fit between the model and the data, NNFI = .930; CFI = .947; RMSEA = .075 (90% confidence interval (CI) = .055, .095). Because the RDS included both predecision and postdecision regret items, we also fitted a two-factor model reflecting specific item contents. This model, NNFI = .930; CFI = .947; RMSEA = .076 (90% CI = .056, .096), was not statistically different from the single-factor model, ΔSBχ²= 0.94; df = 1; p = .33. Therefore, the parameter added to model the two latent variables was redundant and the single-factor model can be accepted just as well. As one can see from Figure 1, no item had a standardized factor loading less than .50 on the single latent variable, and the coefficients ranged from .51 to .74. Noteworthy, there were two postdecisional regret statements (i.e., nos. 2 and 5) and two predecisional regret items (nos. 3 and 9) among the indicator variables with the highest loadings on the latent regret-style factor. This finding reinforced the view that both predecisional and postdecisional regret are strongly tied to how a decision maker habitually construes a decision problem. As a whole, these results supported the unidimensionality of the RDS, which is a necessary condition for construct validity.

Following Byrne (2008), we started a multigroup factor analysis by evaluating the equality of configuration between students and community participants. This analysis yielded a slightly suboptimal fit relative to the single sample analysis but still overall acceptable, SBχ²= 123.91; df = 54; p < .001; CFI = .937; NNFI = .916; RMSEA = 0.086 (90% CI = .065, .105). After that, we constrained the factor loading to be equal between-groups. This analysis yielded acceptable fit indexes, SBχ²= 132.75; df = 62; p < .001; CFI = .936; NNFI = .926; RMSEA = 0.080 (90% CI = .061, .099), and more importantly, fit the data as equally well as the configural invariance model, ΔSBχ²= 5.12; df = 8; p = .74. Finally, we attempted to test the equality of error variances across groups. Although such stringent test of invariance still yielded an acceptable fit, SBχ²= 170.87; df = 71; CFI = .919; NNFI = .908; RMSEA = 0.089 (90% CI = .072, .106), it resulted in a significant worse fit than the loading invariance model, ΔSBχ²= 54.67; df = 9; p = .010. Taken together, the multigroup analyses established the configural and metric invariance of the RDS.

Stability and reliability of regret style

The nine-item RDS attained fairly high-reliability coefficients. The model-based reliability coefficient ω resulting from the single group analysis was .87, while it was .88 and .86, respectively, for students and community participants. Recall that 59 students completed the RDS twice. We used this subsample to examine the stability of the regret-style construct and to assess the test–retest reliability of the RDS scores. The total scores were M = 41.52 (SD = 9.84) and M = 41.66 (SD = 9.53) for test and retest occasions, respectively. A paired sample t test turned out to be statistically insignificant, t(58) = − 0.17; p = .86, and the 4-week test–retest correlation was r = .81. These findings indicated that the decision style was a relatively stable characteristic of the person and strongly supported the reliability of the RDS.

In sum, this study showed that predecision and postdecision regret items tend to merge into a unidimensional latent construct, namely the regret style. Furthermore, the RDS was invariant at the metric level across the student and community samples, an important prerequisite for examining the relationships of regret style with other constructs across different subpopulations. Consistent with our dispositional hypothesis, RDS scores were highly stable over a one-month period. In the next study, we investigated whether the regret style could predict a behavioral outcome in a risky decision-making task.

Study 2: Regret style predicts risky choices

Participants and procedure

One hundred nineteen undergraduate students (68% women) participated in this study for partial course credits. Participants mean age was 21.32 years (SD = 3.47 years). Participants signed up for two separate studies spaced one month apart. On the first occasion, the participants took a large, self-report battery, which included the RDS, the Hospital Anxiety and Depression Scale (HADS), and other scales unrelated to the goals presented in this study. We administered the questionnaires in small-group sessions of about eight people in a laboratory room. We also recorded participants’ gender and age on this occasion. One month later, the participants took the BART. The Positive and Negative Affect Schedule (PANAS) was also administered just before the task to control for individual differences in mood state. Participants individually took the risk task in a 3 m × 3 m experimental room using a desktop computer. After the experiment, the students were debriefed and rewarded for their participation. The experimenters were blind to participant scores on RDS, HADS, and PANAS.

Measures

Regret style

Same as in Study 1.

Positive and negative affect

The PANAS (Terracciano, McCrae, & Costa, 2003) comprises 20 items. Participants were instructed to rate how they felt “right now” on a scale from 1 (very slightly or not at all) to 5 (extremely). Positive and negative mood scores were obtained by adding the 10 positive and 10 negative mood items (α values = .79 and .86 for positive and negative mood, in this sample).

Anxiety and depression

The HADS (Iani, Lauriola, & Costantini, 2014) comprises 14 items, each rated on a 4-point scale. It provides two subscale scores with higher scores indicating greater levels of anxiety and depression (α values = .77 and .54 for anxiety and depression, in this sample).

Risk taking

The BART (Lejuez et al., 2002) is a 30-trial, computer-based measure of risk-taking propensity. On each trial, a small simulated balloon is presented, and the participants can inflate it by pressing a pump button, thereby earning €.05, which is accrued in a temporary bank. The balloon can explode any time after each pump, and if this happens, sums of money in the temporary bank are lost. The participant can decide to transfer the money from the temporary bank to the permanent bank, thus ending the trial. The primary dependent measure on the BART is the average number of pumps across trials, excluding trials in which the balloon exploded, hereafter referred to as the Average Adjusted Pump (AAP). The AAP score is widely considered to be a valid measure of risk-taking tendencies across a range of clinical and nonclinical contexts (for review, see Lauriola, Panno, Levin, & Lejuez, 2014).

Statistical analysis

We carried out a hierarchical multiple regression analysis, regressing AAP on regret style, anxiety, depression, positive affect, and negative affect. Because of our interest in determining the degree to which regret style uniquely accounts for the variance in risk-taking, we first fitted a regression model with mood states (Step 1), then anxiety and depression (Step 2), and finally, added regret style (Step 3) to the model. We interpreted a significant change in R² as support for the unique predictive power of regret style, beyond that observed with mood state variables entered on Stage 1 because of their temporal proximity with the dependent variable, and more dispositionally stable measures of anxiety and depression entered on Stage 2 because of their direct link with the negative emotionality/neuroticism trait. To assess the shared and unique contributions of each predictor, we compared the zero-order correlations to partial correlations of mood states, depression, anxiety, and regret style with risk-taking while controlling for the other predictors.

Results and discussion

The results of Step 1 indicate that negative and positive mood accounted for about 4% of the variance in risk-taking on BART. Despite the fact that the analysis only approached to statistical significance, F(2, 115) = 2.39, p = .096, the inspection of model’s coefficients showed that positive mood negatively predicted risk-taking on BART (Table 1). Participants who were in a positive mood just before taking the task were indeed more conservative in their choices. Introducing trait anxiety and depression on Step 2 explained an additional 10% of the variation in risk-taking; ΔR² was significant, F(2, 113) = 6.32, p = .003. Anxiety, but not depression, predicted less risk-taking (i.e., fewer pumps) on BART (Table 1). For Step 3, the addition of regret style to the regression model explained an additional 4% of the variation in risk-taking. This led to a significant ΔR², F(1, 112) = 6.46, p = .012. When all five independent variables were included in the analysis, neither depression nor negative state mood were significant predictors of risk-taking (Table 1). State positive mood uniquely explained 8% of the variation in risk-taking. Both trait anxiety and regret style uniquely accounted for about 5% of the variance in the dependent variable. Together, the five independent variables accounted for about 18% of the variance in risk-taking, F(5, 112) = 5.04, p < .001. Noteworthy, the partial correlation for regret style was about as large as the zero-order correlation assessing the gross effect of that style on risk-taking (Table 1). In sum, our results demonstrate that decision makers who are more likely to take regret feelings into account were more risk-averse than those who did not consider regret as much. More broadly, the association of regret style with the decision outcome was not spuriously inflated by negative mood, either in state or in trait form.

Table 1.

Summary table of multiple regression analysis.

Step		B	Standard error	β	t	p	Zero order	Partial	R ²	R² change
1	(Constant)	47.83	8.11						.04^†	.04^†
	Negative Affect	−0.13	0.73	−0.02	−0.18	.86	−.04	−.02
	Positive Affect	−1.74	0.81	−0.20	−2.15	.03	−.20	−.20
2	(Constant)	71.10	11.29						.14**	.10**
	Negative Affect	0.43	0.72	0.05	0.59	.56	−.04	.06
	Positive Affect	−2.06	0.79	−0.23	−2.61	.01	−.20	−.24
	Depression	−0.36	1.23	−0.03	−0.29	.77	−.07	−.03
	Anxiety	−3.34	0.99	−0.31	−3.36	.00	−.29	−.30
3	(Constant)	76.54	11.23						.18***	.05*
	Negative Affect	1.26	0.78	0.16	1.62	.11	−.04	.15
	Positive Affect	−2.35	0.78	−0.27	−3.02	.00	−.20	−.27
	Depression	0.47	1.25	0.04	0.38	.71	−.07	.04
	Anxiety	−2.43	1.03	−0.23	−2.35	.02	−.29	−.22
	Regret Decision Style	−0.44	0.17	−0.28	−2.54	.01	−.26	−.23

Note: N = 119; ***p < .001; **p < .01; *p < .05; ^†p = .096.

Discussion

Although intuitive and reflective decision styles have been extensively studied as trait-like dimensions with predictive validity in real life decisions (Kozhevnikov et al., 2014; Phillips et al., 2016; Wang et al., 2017), less is known about regret-based decision-making style, despite the fact that minimizing regret is the most studied emotion in decision-making (George & Dane, 2016; Lerner et al., 2015). Is regret-based decision-making a trait-like construct with predictive validity in risky decision-making? The results of two independent studies suggest that the answer to this question is overall positive.

Study 1 demonstrated that a single latent variable measured by predecisional and postdecisional regret items was an acceptable fit, whereas a two-factor model with separate latent variables for each type of regret was an less parsimonious classification of regret style. When people were asked to report their typical way to make their decisions, the two aspects were strongly interrelated. Consistent with an emotion-as-feedback approach, postdecisional regret may help the decision maker to learn from experience and to update the recovery of anticipated regret feeling useful for future decisions (DeWall et al., 2014). Risky choices that end up in negative outcomes may engender counterfactual thinking on how the outcome could have been different if one had made a different choice, in turn, it may elicit, the painful experience of regret. This emotional information, stored in memory, is believed to become salient again in the form of anticipated regret when one has to make future risky decisions and might contribute to deciding more cautiously.

By definition, a dispositional construct is expected to be relatively stable over time and able to predict behavior. In this regard, Study 1 provided preliminary evidence that regret-based decision-making scores did not change over a one-month period. Also, Study 2 reinforced the conclusion regarding the stability of the trait, showing that regret scores predicted observable risky choices in a behavioral risk-task administered one month later. Noteworthy, prior regret-style research was entirely based on cross-sectional designs. Therefore, although people endorsing a regret style exhibited decision difficulty, indecisiveness, decision avoidance, and self-doubting concerning decision-making ability (Djulbegovic et al., 2015; Elaydi, 2006; Leykin & DeRubeis, 2010; Morera et al., 2006; Nygren & White, 2002; Rim et al., 2011; Shaffer et al., 2014; Turner et al., 2012), the finding that regret style was stable over time (Study 1) and predicted observable choice in a risk-taks (Study 2) is novel and extends the literature in a meaningful way.

Because regret style has been shown to be associated with negative affective traits like harm avoidance, anxiety, depression, and neuroticism (Dewberry et al., 2013; Leykin & DeRubeis, 2010; Nygren & White, 2002, 2005), one might argue that a negative emotionality might predict risk aversion as well as regret style. Despite the fact that trait-anxiety assessed one month before the task also predicted subsequent risk aversion, regret style was also significant controlling for anxiety. Therefore, it is tempting to speculate that our findings might reflect unique processes through which negative affectivity and regret might motivate risk-averse choices. Even though regret and anxiety are both negative emotions, their role in predicting risk-taking on BART was likely very specific. Consistent with regret theory (Pieters & Zeelenberg, 2007; Zeelenberg & Pieters, 2007), the regret-style measure used in this study predicted risk aversion because there was an element of agency in pumping the balloon, while the money accumulated in the temporary bank might have triggered counterfactual thoughts as if the balloon had exploded. Because regret regulation theory maintains that people strive to avoid regret and predicts risk-seeking or risk aversion depending on which option protects the most from feedback on the foregone outcome, these results suggest that individuals reporting greater regret-style orientation approached the BART as if the foregone outcome was the money accumulated in the temporary bank. In addition, consistent with an adult temperament perspective on risk-taking (Lauriola & Weller, 2018), we found that anxiety predicted risk aversion. This finding resonates with past research that has reported robust associations between traits related to negative affectivity and risk perceptions (e.g., Chauvin, Hermand, & Mullet, 2007; Sween, Ceschi, Tommasi, Sartori, & Weller, 2017; Weller & Tikir, 2011). In this context, trait-anxiety may be associated with elevated perceptions of dangers associated with the likelihood that the balloon would explode, thereby motivating a reaction of cognitive, and physical, withdrawal from the aversive stimuli.

More generally, our second study paralleled the extant literature on regret and risk aversion real-world domains (Conner et al., 2006; Cooke et al., 2007; Elliott et al., 2013; Janssen et al., 2014; Lazuras et al., 2012, 2017; Li et al., 2010; Richard et al., 1996; Smerecnik & Ruiter, 2010; van Empelen et al., 2001; Vézina-Im & Godin, 2011) as well as our previous research on anticipated regret and BART (Panno, Lauriola & Pierro, 2015). As such, it seems reasonable to conclude that regret style might be associated with real-world risk-taking as well as it was predictive of risky choices in our laboratory.

Because emotions unrelated to the actual decision (i.e., incidental emotions) may bias the decision process (Lerner et al., 2015) and interfere with one’s experience of regret (Tochkov, 2012), we controlled for negative and positive mood states in regression analysis. Regret style predicted risk aversion on BART above and beyond incidental emotions. However, one unanticipated finding was that positive mood was the single best predictor of risk aversion on BART. We interpreted this finding in a mood maintenance framework (George & Dane, 2016). Specifically, individuals in a positive mood tend to be more optimistic in risky situations, but notwithstanding this, they are also less willing to risk, especially if possible losses are tangible and valuable. However, it is worth acknowledging that other accounts are possible, such as affect generalization approaches, that make different predictions, namely positive mood may increase risk-taking due to an optimistic appraisal of probabilities (Lerner et al., 2015).

Limitations and future directions

Notwithstanding methodological strengths (e.g., longitudinal design, behavioral observations), this study has several limitations that provide directions for future research. First, both studies involved a one-month interval between two measurement occasions. Therefore, the degree to which the stability coefficient of regret style fully resembles that of personality traits remain an open research question. Future research should expand on these findings using research designs involving more time points, longer periods, and psychometric methods aimed to disentangle state and trait components in regret-style scores (e.g., latent state-trait models). Second, although there is evidence that risk-taking behavior is rooted in personality (Lauriola & Weller, 2018), it is also domain specific (Weber & Blais, 2006). Accordingly, one might argue that the predictive relation of regret-based decision-making with the BART is not warranted to generalize to other risk domains (e.g., social, financial, or ethical). Studies might then investigate to what extent individual differences in regret-based decision-making generalize to other types of risk tasks or predict real-world risk-taking in at least two different domains. It would be interesting to include in future studies variables describing risky behaviors of the participant’s everyday life that span across a variety of domains, including, but not limited to, financial, health, social, and ethical domains. Third, the samples used in both studies are nonprobabilistic, and hence, we must temper any generalizations across a broader population. A goal of future research would be to examine the measurement invariance of the RDS with more robust samples, and the predictive relations assessed for regret style and risk aversion should be cross-validated using more diverse samples.

Conclusion

The results of this study ruled in three main characteristics expected for personality dispositions, such as showing an empirically robust and unidimensional factorial structure, being relatively stable over time, and being able to predict choice behavior. As such, our study sought to clarify whether habitually taking regret into account can be thought as a trait-like dimension with predictive validity on risk-taking behavior in laboratory settings. Furthermore, our findings also show promise for studies oriented toward investigating how dispositional regret can predict real-world risk-taking behavior.

Footnotes

Author Contributions

The authors discussed the contents of this article together. Marco Lauriola elaborated the research hypotheses, devised the methodological content, and analyzed the data. Angelo Panno developed the theoretical framework and collected the data. Joshua A. Weller conferred with other authors about theoretical and empirical aspects of the studies and provided a significant contribution to the interpretation and discussion of research findings. The final version of the article was written by M. Lauriola, A. Panno, and J. A. Weller.

Acknowledgments

The authors thank Giulia Santoro, Correggia Angela, and Valentina Di Rago for their help in collecting data and references.

Article Notes

References

Bentler

P. M.

(2004) EQS 6: Structural equation program manual, Encino, CA: Multivariate Software, Inc.

Betsch

Iannello

(2010) Measuring individual differences in intuitive and deliberate decision-making styles: A comparison of different measures. In: Glockner

Witteman

(eds) Foundations for tracing intuition: Challenges and methods, London, England: Psychology Press., pp. 251–267. Retrieved from https://publicatt.unicatt.it/handle/10807/33801.

Brewer

N. T.

DeFrank

J. T.

Gilkey

M. B.

(2016) Anticipated regret and health behavior: A meta-analysis. Health Psychology 35(11): 1264–1275. doi:10.1037/hea0000294.sup.

Byrne

B. M.

(2006) Structural equation modeling with EQS: Basic concepts, applications, and programming, Mahwah, NJ: Lawrence Erlbaum Associates.

Byrne

B. M.

(2008) Testing for multigroup equivalence of a measuring instrument: A walk through the process. Psicothema 20(4): 872–882.

Chauvin

Hermand

Mullet

(2007) Risk perception and personality facets. Risk Analysis 27(1): 171–185. doi:10.1111/j.1539-6924.2006.00867.x.

Conner

Sandberg

McMillan

Higgins

(2006) Role of anticipated regret, intentions and intention stability in adolescent smoking initiation. British Journal of Health Psychology 11(1): 85–101. doi:10.1348/135910705X40997.

Cooke

Sniehotta

Schüz

(2007) Predicting binge-drinking behaviour using an extended tpb: Examining the impact of anticipated regret and descriptive norms. Alcohol and Alcoholism 42(2): 84–91. doi:10.1093/alcalc/agl115.

DeWall

C. N.

Baumeister

R. F.

Chester

D. S.

Bushman

B. J.

(2014) How often does currently felt emotion predict social behavior and judgment? A meta-analytic test of two theories. Emotion Review 8(2): 136–143. doi:10.1177/1754073915572690.

10.

Dewberry

Juanchich

Narendran

(2013) Decision-making competence in everyday life: The roles of general cognitive styles, decision-making styles and personality. Personality and Individual Differences 55(7): 783–788. doi:10.1016/j.paid.2013.06.012.

11.

Djulbegovic

Beckstead

Elqayam

Reljic

Kumar

Paidas

Djulbegovic

(2015) Thinking styles and regret in physicians. PLoS One 10(8): e0134038. doi:10.1371/journal.pone.0134038.

12.

Elaydi

(2006) Construct development and measurement of indecisiveness. Management Decision 44(10): 1363–1376. doi:10.1108/00251740610715696.

13.

Elliott

M. A.

Thomson

J. A.

Robertson

Stephenson

Wicks

(2013) Evidence that changes in social cognitions predict changes in self-reported driver behavior: Causal analyses of two-wave panel data. Accident Analysis and Prevention 50: 905–916. doi:10.1016/j.aap.2012.07.017.

14.

Evans

J. S. B. T.

Stanovich

K. E.

(2013) Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science 8(3): 223–241. doi:10.1177/1745691612460685.

15.

George

J. M.

Dane

(2016) Affect, emotion, and decision making. Organizational Behavior and Human Decision Processes 136: 47–55. doi:10.1016/j.obhdp.2016.06.004.

16.

Hamilton

Shih

S. I.

Mohammed

(2016) The development and validation of the rational and intuitive Decision Styles Scale. Journal of Personality Assessment 98(5): 523–535. doi:10.1080/00223891.2015.1132426.

17.

Iani

Lauriola

Costantini

(2014) A confirmatory bifactor analysis of the Hospital Anxiety and Depression Scale in an Italian community sample. Health and Quality of Life Outcomes 12(84): 1–8. doi:10.1186/1477-7525-12-84.

18.

Janssen

Waters

E. A.

Van Osch

Lechner

De Vries

(2014) The importance of affectively-laden beliefs about health risks: The case of tobacco use and sun protection. Journal of Behavioral Medicine 37(1): 11–21. doi:10.1007/s10865-012-9462-9.

19.

Kozhevnikov

(2007) Cognitive styles in the context of modern psychology: Toward an integrated framework of cognitive style. Psychological Bulletin 133(3): 464–481. doi:10.1037/0033-2909.133.3.464.

20.

Kozhevnikov

Evans

Kosslyn

S. M.

(2014) Cognitive style as environmentally sensitive individual differences in cognition. Psychological Science in the Public Interest 15(1): 3–33. doi:10.1177/1529100614525555.

21.

Lauriola, M., & Weller, J. (2018). Personality and risk: Beyond daredevils – Risk taking from a temperament perspective. In M. Raue, E. Lermer, & B. Streicher (Eds.), Psychological aspects of risk and risk analysis: Theory, models, and applications. New York: Springer.

22.

Lauriola

Panno

Levin

I. P.

Lejuez

C. W.

(2014) Individual differences in risky decision making: A meta-analysis of sensation seeking and impulsivity with the Balloon Analogue Risk Task. Journal of Behavioral Decision Making 27(1): 20–36. doi:10.1002/bdm.1784.

23.

Lazuras

Barkoukis

Mallia

Lucidi

Brand

(2017) More than a feeling: The role of anticipated regret in predicting doping intentions in adolescent athletes. Psychology of Sport and Exercise 30: 196–204. doi:10.1016/j.psychsport.2017.03.003.

24.

Lazuras

Chatzipolychroni

Rodafinos

Eiser

J. R.

(2012) Social cognitive predictors of smoking cessation intentions among smoker employees: The roles of anticipated regret and social norms. Addictive Behaviors 37(3): 339–341. doi:10.1016/j.addbeh.2011.11.008.

25.

Lejuez

C. W.

Read

J. P.

Kahler

C. W.

Richards

J. B.

Ramsey

S. E.

Stuart

G. L.

Brown

R. A.

(2002) Evaluation of a behavioral measure of risk taking: The Balloon Analogue Risk Task (BART). Journal of Experimental Psychology: Applied 8(2): 75–84. doi:10.1037/1076-898X.8.2.75.

26.

Lerner

J. S.

Valdesolo

Kassam

K. S.

(2015) Emotion and decision making. Annual Review of Psychology 66(1): 799–823. doi:10.1146/annurev-psych-010213-115043.

27.

Leykin

DeRubeis

R. J.

(2010) Decision-making styles and depressive symptomatology: Development of the Decision Styles Questionnaire. Judgment and Decision Making 5(7): 506–515. doi:10.1016/j.cognition.2008.05.007.

28.

Zhou

Sun

Rao

L. L.

Zheng

Liang

Z. Y.

(2010) Anticipated regret, risk perception, or both: Which is most likely responsible for our intention to gamble? Journal of Gambling Studies 26(1): 105–116. doi:10.1007/s10899-009-9149-5.

29.

Mann

Burnett

Radford

Ford

(1997) The Melbourne decision making questionnaire: An instrument for measuring patterns for coping with decisional conflict. Journal of Behavioral Decision Making 10(1): 1–19. doi:10.1002/(SICI)1099-0771(199703)10:1<1::AID-BDM242>3.0.CO;2-X.

30.

Morera

O. F.

Maydeu-Olivares

Nygren

T. E.

White

R. J.

Fernandez

N. P.

Skewes

M. C.

(2006) Social problem solving predicts decision making styles among US Hispanics. Personality and Individual Differences 41(2): 307–317. doi:10.1016/j.paid.2005.08.016.

31.

Nygren, T. E. (2000). Development of a measure of decision making styles to predict performance in a dynamic J/DM task. Paper presented at the 41st Psychonomic Society Meeting, New Orleans, LA.

32.

Nygren

T. E.

White

R. J.

(2002) Assessing individual differences in decision making styles: Analytical vs. intuitive. Proceedings of the Human Factors and Ergonomics Society Annual Meeting 46(12): 953–957. doi:10.1177/154193120204601204.

33.

Nygren

T. E.

White

R. J.

(2005) Relating decision making styles to predicting selfefficacy and a generalized expectation of success and failure. Proceedings of the Human Factors and Ergonomics Society Annual Meeting 49(3): 432–434. doi:10.1177/154193120504900346.

34.

Pacini

Epstein

(1999) The relation of rational and experiential information processing styles to personality, basic beliefs, and the ratio-bias phenomenon. Journal of Personality and Social Psychology 76(6): 972–987. doi:10.1037/0022-3514.76.6.972.

35.

Panno, A., Lauriola, M., & Pierro, A. (2015). Regulatory mode and risk-taking: The mediating role of anticipated regret. PlosOne, 10(11), e0143147. https://doi.org/10.1371/journal.pone.0143147.

36.

Phillips

W. J.

Fletcher

J. M.

Marks

A. D. G.

Hine

D. W.

(2016) Thinking styles and decision making: A meta-analysis. Psychological Bulletin 142(3): 260–290. doi:10.1037/bul0000027.

37.

Pieters

Zeelenberg

(2007) A theory of regret regulation 1.1. Journal of Consumer Psychology 17(1): 29–35. doi:10.1207/s15327663jcp1701_6.

38.

Richard

van der Pligt

de Vries

(1996) Anticipated regret and time perspective: Changing sexual risk-taking behavior. Journal of Behavioral Decision Making 9: 185–199. doi:10.1002/(SICI)1099-0771(199609)9:3<185::AID-BDM228>3.0.CO;2-5.

39.

Rim

H. B.

Turner

B. M.

Betz

N. E.

Nygren

T. E.

(2011) Studies of the dimensionality, correlates, and meaning of measures of the maximizing tendency. Judgment and Decision Making 6(6): 565–579. Retrieved from http://journal.sjdm.org/11/11316/jdm11316.html.

40.

Scott

S. G.

Bruce

R. A.

(1995) Decision-making style: The development and assessment of a new measure. Educational and Psychological Measurement 55(5): 818–831. doi:10.1177/0013164495055005017.

41.

Shaffer

V. A.

Tomek

Hulsey

(2014) The effect of narrative information in a publicly available patient decision aid for early-stage breast cancer. Health Communication 29(1): 64–73. doi:10.1080/10410236.2012.717341.

42.

Sjöberg, L. (2003, March 1). Intuitive vs. analytical decision making: Which is preferred? Scandinavian Journal of Management, 19(1), 17–29. doi:10.1016/S0956-5221(01)00041-0.

43.

Smerecnik

C. M. R.

Ruiter

R. A. C.

(2010) Fear appeals in HIV prevention: The role of anticipated regret. Psychology, Health and Medicine 15(5): 550–559. doi:10.1080/13548506.2010.498888.

44.

Sween

Ceschi

Tommasi

Sartori

Weller

(2017) Who is a distracted driver? Associations between mobile phone use while driving, domain-specific risk taking, and personality. Risk Analysis 37(11): 2119–2131. doi:10.1111/risa.12773.

45.

Terracciano

McCrae

R. R.

Costa

P. T.

(2003) Factorial and construct validity of the Italian Positive and Negative Affect Schedule (PANAS). European Journal of Psychological Assessment 19(2): 131–141. doi:10.1027//1015-5759.19.2.131.

46.

Tochkov

(2009) The effects of anticipated regret on risk preferences of social and problem gamblers. Judgment and Decision Making 4(3): 227.

47.

Tochkov

(2012) No regrets? Mood and the anticipation of emotions in problem gambling. International Gambling Studies 12(1): 39–53. doi:10.1080/14459795.2011.611525.

48.

Turner

B. M.

Rim

H. B.

Betz

N. E.

Nygren

T. E.

(2012) The maximization inventory. Judgment and Decision Making 7(1): 48–60. doi:10.1016/j.cognition.2008.05.007.

49.

van Empelen

Kok

Jansen

M. W. J.

Hoebe

C. J. P. A.

(2001) The additional value of anticipated regret and psychopathology in explaining intended condom use among drug users. AIDS Care—Psychological and Socio-Medical Aspects of AIDS/HIV 13(3): 309–318. doi:10.1080/09540120120043964.

50.

Vézina-Im

L. A.

Godin

(2011) Psychosocial determinants of intention to abstain from drinking alcohol while pregnant among a sample of women of childbearing age. Addiction Research and Theory 19(2): 128–137. doi:10.3109/16066359.2010.512107.

51.

Wang

Highhouse

Lake

C. J.

Petersen

N. L.

Rada

T. B.

(2017) Meta-analytic Investigations of the Relation Between Intuition and Analysis. Journal of Behavioral Decision Making 30(1): 15–25. doi:10.1002/bdm.1903.

52.

Weber

E. U.

Blais

A.-R.

(2006) A Domain-Specific Risk-Taking (DOSPERT) scale for adult populations. Judgement and Decision Making 1(1): 33–47. doi:10.1037/t13084-000.

53.

Weller

J. A.

Tikir

(2011) Predicting domain-specific risk taking with the HEXACO personality structure. Journal of Behavioral Decision Making 24(2): 180–201. doi:10.1002/bdm.677.

54.

Zeelenberg

Beattie

Van Der Pligt

De Vries

N. K.

(1996) Consequences of regret aversion: Effects of expected feedback on risky decision making. Organizational Behavior and Human Decision Processes 65(2): 148–158. doi:10.1006/obhd.1996.0013.

55.

Zeelenberg

Pieters

(2004) Consequences of regret aversion in real life: The case of the Dutch postcode lottery. Organizational Behavior and Human Decision Processes 93(2): 155–168. doi:10.1016/j.obhdp.2003.10.001.

56.

Zeelenberg

Pieters

(2007) A theory of regret regulation 1.0. Journal of Consumer Psychology 17(1): 3–18. doi:10.1207/s15327663jcp1701_3.