Sage Journals: Discover world-class research

Abstract

Prior research linking anxiety and depression to risky decision-making has yielded inconsistent findings. This may reflect differing definitions of risk, characterized as outcome variance or potential negative consequences. This cross-sectional experimental study investigated affective distress and aimed to determine whether choices were driven by aversion to riskier (higher variance) outcomes, sensitivity to negative outcomes, or a combination of the two. A sample of adults (N = 526, aged 18-60) completed a risky decision-making task involving choices between probabilistic gambles with positive and/or negative outcomes. An affective distress component was obtained from self-report measures of anxiety and depression using principal component analysis. When losses were possible, choice behavior reflected risk aversion and heightened sensitivity to differences in expected value; however, this was attenuated with increasing levels of affective distress, resulting in less reliable selection of mathematically advantageous options. In gain-only lotteries, choices were consistent with risk seeking and reduced sensitivity to differences in expected value, regardless of affective distress. These findings suggest that decision-making and risk attitudes are substantially influenced by the presence or absence of potential loss. Affective distress alters choice behavior specifically when loss is present, suggesting that loss may play a more prominent role than variance in risky decision-making for anxiety and depression.

Keywords

affective distress decision-making loss aversion risk valence

Affective disorders such as anxiety and depression are among the most prevalent mental health conditions globally (World Health Organization, 2022), impacting approximately 298 million and 193 million people, respectively (Santomauro et al., 2021). Anxiety is primarily characterized by excessive worry and heightened sensitivity to threat, and depression is defined by persistent low mood and reduced sensitivity to reward. Both anxiety and depression involve common features including general distress, and comorbidity between them is high (Clark & Watson, 1991). Heightened intolerance of uncertainty, wherein unknown future events elicit aversive emotional responses, has been identified as a transdiagnostic feature, impacting individuals with subclinical and clinical presentations of anxiety and/or depression (Carleton, 2016; McEvoy & Mahoney, 2012). Furthermore, altered decision-making processes, namely over-valuation of bad outcomes and under-valuation of good outcomes, have been observed in individuals with affective disorders (Paulus & Yu, 2012). However, current understanding of value-based decision components underlying these trends remains somewhat contentious due to conflicting perspectives and variability in the definition of risk (Aven, 2012; Mata et al., 2018; Schonberg et al., 2011). Some argue that differences in decision patterns are driven by an aversion to risk, in terms of uncertain (i.e., probabilistic and/or indeterminate) outcomes (Charpentier et al., 2017; Hartley & Phelps, 2012), whereas others suggest that aversion to potential loss plays a more prominent role (Xu et al., 2020). The present study used a novel adaptation of an established risky decision task (Levy et al., 2010; Ruderman et al., 2016; Tymula et al., 2013) to clarify differential contributions of risk and loss to value-based decision-making across the spectrum of affective distress symptoms in the general population.

Conceptualization of risk plays an essential role in what we understand about affective distress and decision-making. Risk, or first-order uncertainty, is typically defined as any scenario in which potential outcomes and their probabilities are known, but the eventual outcome is indeterminate (Bach & Dolan, 2012). This is distinct from second-order uncertainty, which involves uncertainty about the decision environment in addition to outcomes, such as ambiguity, where some element of potential outcomes (e.g., value and/or probability) is unknown or obscured. According to economic definitions, the objective degree of risk is quantified by the variance of potential outcomes relative to the expected value (EV; the sum of the probability-weighted potential outcomes; Markowitz, 1952). Larger variance reflects greater disparity between potential outcomes and therefore higher risk. In contrast, psychological approaches define risk in terms of negative consequences, such as harm or loss (Schonberg et al., 2011). Accordingly, the degree of risk is measured by the magnitude and probability of undesirable outcomes (Botelho et al., 2023). While both models involve uncertainty regarding eventual outcomes, they differ in their methods of quantifying risk, leading to distinct interpretations of which element makes it aversive (Hertwig et al., 2018). For the purposes of the present study, we adopt the economic definition of risk as variance, and account for the psychological definition through examination of discomfort with loss. Clarifying these distinctions is critical because theoretical differences map directly onto competing explanations of why individuals with affective distress may deviate from optimal choice.

The process of decision-making is informed by assessing and comparing the values of available options (Rangel et al., 2008). A rational decision-maker would be expected to choose the option with the most favorable EV, reflecting attempts to maximize positive outcomes and minimize negative outcomes (Bishop & Gagne, 2018); however, it is well established that human behavior often deviates from such expectations as choices are guided by subjective rather than objective valuation (Camerer, 1989). The ‘risk-as-feelings’ hypothesis offers a potential explanation, highlighting that emotional responses, rather than cognitive assessments of risk, often drive decisions involving uncertainty (Loewenstein et al., 2001). Under this theory, uncertainty creates an internalized negative state that influences decision-making rather than relying on the objective EV of options. Anxiety and depression are characterized by increased negative affect, which may exacerbate the ‘risk-as-feelings’ effect when faced with potential negative outcomes (Phelps et al., 2014). Anxiety and mood disorders have been associated with elevated appraisals of the likelihood and severity of negative future events (Butler & Mathews, 1983), with depressed individuals additionally anticipating fewer positive outcomes (Miranda & Mennin, 2007). Affective states have been shown to distort the valuation of probabilistic outcomes through heightened sensitivity to low-probability negative outcomes, diminished responsiveness to reward magnitude, and altered weighting of gains versus losses (Bishop & Gagne, 2018; Gu et al., 2017; Paulus & Yu, 2012).

Attitudes toward risk are commonly evaluated using value-based decision-making tasks that manipulate the properties of potential outcomes of options, such as their magnitude, probability, or valence (positive or negative; Botelho et al., 2023). For example, consider a pair of gambles with one option offering a 20% chance of winning $20 and an 80% chance of receiving nothing, and the other offering a 60% chance of winning $10 and a 40% chance of losing $5. Neither option is mathematically advantageous, as they both have an EV of $4; however, the mechanisms driving choice can be inferred from the option selected. Discomfort with variance would likely result in selection of the second gamble, while preference to avoid the possibility of loss would lead to selection of the first gamble. There are numerous studies examining the relationship of anxiety and depression with risky decision-making, but a primary limitation of these studies is their focus on gain outcomes, which does not allow for examining gain/loss asymmetry (valence; e.g., Chung et al., 2017; FeldmanHall et al., 2016; Hagiwara et al., 2022). Anxiety is typically associated with a preference for certain or low-variance options, even when that preference results in lower reward (Hartley & Phelps, 2012). Such findings have emerged across diverse samples including induced (Raghunathan & Pham, 1999) and subclinical anxiety populations (Schmidt et al., 2018).

There is evidence to suggest that individuals with anxiety exhibit increased discomfort with variance when loss is possible. According to the reflection effect (Kahneman & Tversky, 1979), people tend to be risk averse for gains, avoiding probabilistic gains in favor of guaranteed reward, and risk seeking for losses, preferring probabilistic losses which present the possibility of avoiding unfavorable outcomes. Contrary to these expectations, Galván and Peris (2014) found that individuals with anxiety selected significantly fewer risky choices under loss relative to the control group, preferring certain options, consistent with risk aversion in a context that would predict risk seeking. This may indicate that those with anxiety find not knowing whether a negative outcome will occur more aversive than the loss itself. However, the comparison between probabilistic and guaranteed outcomes conflates tolerance for outcome variance and preference for certainty because selecting the guaranteed option eliminates variance altogether. Accordingly, sensitivity to variance can be better evaluated in the absence of guaranteed options. In a study presenting choices between pairs of probabilistic gambles, each offering a 50% chance of resulting in a gain or loss, variance was manipulated by adjusting the magnitudes of potential outcomes (Giorgetta et al., 2012). Anxious individuals selected the ‘safer’ lower variance gambles more frequently than healthy controls. Although this behavior may be driven by heightened discomfort with variance when loss is possible, it cannot be distinguished from desire to minimize loss because higher variance options also involved larger potential losses. In addition, both paradigms attempted to isolate variance preferences by presenting choices between options of equal EV; however, this approach precludes evaluation of rational value-based choice as when value is held constant, it is not possible to determine whether choices are driven by altered sensitivity to variance or potential loss.

In an influential study, Charpentier et al. (2017) aimed to quantify the relative contributions of risk aversion and loss aversion in clinical anxiety using a prospect theory modeling approach (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992). Participants made binary choices involving gain lotteries (certain gain versus a 50% chance of winning a larger gain or nothing) and mixed lotteries (certain $0 versus a 50% chance of winning or losing some monetary amount). Beyond evaluating choice proportions, Charpentier and colleagues (2017) conducted computational modeling to estimate risk and loss parameters. Relative to healthy controls, individuals with anxiety exhibited increased risk aversion, but not loss aversion. However, it is important to note that these parameters represent subjective adjustments to value informing choice selection rather than discomfort with variance or loss. Under prospect theory, risk aversion is operationalized as diminishing sensitivity to value, which describes how differences between larger outcomes are perceived as smaller than equivalent differences between smaller outcomes (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992). Furthermore, loss aversion specifically represents how the magnitude of losses are weighted relative to equivalent gains. Additionally, the fundamentally different formats of lotteries potentially compromised comparability. Specifically, mixed gambles were more akin to accept/reject approaches given the certain alternative was $0 whereas gain lotteries presented a clear certain versus probabilistic choice. These analytic approaches and methodological features limit the strength of the conclusion that risk aversion, rather than an exaggerated weighting of losses, underlies anxiety-related decision biases. Consequently, it remains unclear whether the presence of loss increases discomfort with variance relative to gain contexts or whether anxious individuals exhibit a generalized discomfort with variance regardless of valence.

As real-world decision-making rarely includes a guaranteed option, paradigms involving choices between risky alternatives varying in both EV and variance offer more ecologically valid inference (Holt & Laury, 2002; Wright et al., 2012, 2013). Wright and colleagues (2012, 2013) demonstrated that the pattern of choice preferences across gains and losses reversed when participants could avoid risk by selecting a guaranteed option compared with when they were required to choose between two risky options, underscoring the importance of task structure in interpreting attitudes toward risk. Complementary evidence shows that potential losses increase attention to EV differences, enhancing discrimination between advantageous and disadvantageous options rather than reflecting increased risk seeking or loss-averse behavior (Yechiam & Hochman, 2013). This heightened attentiveness to potential loss information may be especially pronounced in individuals with affective distress symptoms given that heightened sensitivity to punishment is characteristic of both anxiety and depression (Katz et al., 2020).

The Present Study

The present study aims to investigate decision-making under first-order uncertainty by examining the contribution of distinct aversive features identified in economic and psychological theories of risk, and how these interact with affective distress symptomatology. To achieve this, we implemented a binary choice paradigm including pairs of probabilistic lotteries with gain only, loss only, or mixed monetary outcomes. If behavior is driven by a general discomfort with variance, we would expect to observe avoidance of riskier (higher variance) options regardless of whether potential outcomes are gains or losses, even when the EV of the safer (lower variance) option is less favorable ( $H_{a}$ ). Alternatively, if discomfort with undesirable outcomes/loss guides decisions, choices involving potential loss should reflect sensitivity to EV, favoring the mathematically optimal option even if the variance is higher ( $H_{b}$ ). A third possibility is that both mechanisms contribute, such that discomfort with variance is present across all lotteries but amplified in the presence of loss. We predict that individuals with elevated affective distress will demonstrate stronger avoidance of riskier options in general, but particularly so when gambles involve loss, consistent with evidence of elevated intolerance of uncertainty and sensitivity to punishment.

Method

Participants

Data were collected from 627 participants recruited via Amazon’s Mechanical Turk (MTurk) in September 2018. Subjects between 18 and 60 years old residing in the United States were eligible to participate. Following exclusions (see Data Screening), the final sample consisted of 526 participants (254 male, 271 female, 1 non-binary) aged 19 to 60 (M = 35.78, SD = 10.40). Participants completed one of three conditions (see Experimental Conditions): reward-only lotteries (gain-only), reward-only and punishment-only lotteries in separate blocks (gain/loss separate [G/L-separate]) and mixed reward and punishment lotteries (G/L-mixed), which were comprised of 190, 185 and 151 participants, respectively. All participants provided informed consent. This study was approved by the Human Research Ethics Committee of the University of Melbourne (ID 1852118).

There is currently no single standardized analytical procedure for calculating the statistical power of binomial generalized linear mixed models involving complex random-effects structures and multiple interaction terms (Kumle et al., 2021; Mathieu et al., 2012). Accordingly, we conducted a sensitivity analysis for an ANCOVA, with lottery type as a between-subjects factor, and affective distress and $Δ E V$ (described in Data Analysis) as covariates, using GPower (Faul et al., 2009). This indicated that the sample of 526 participants would provide 80% power to detect a small-to-medium effect (f = 0.15) with nominal $α$ = .05. This analysis likely constitutes a conservative estimate of power given that linear mixed-effects analyses allow examination of choice behavior without aggregating across trials, thereby retaining within-participant variability across trials and supporting greater statistical precision (Brown, 2021; Matuschek et al., 2017).

Measures

Psychopathology Questionnaires

Several reliable and well-validated questionnaires were administered to assess psychopathology. Cronbach’s alpha was calculated from the items of each subscale and/or scale, to assess the internal consistency of the present sample.

Generalized Anxiety Disorder Questionnaire (GAD-7; Spitzer et al., 2006)

The GAD-7 is a 7-item scale designed to measure the core symptoms of Generalized Anxiety Disorder per the DSM-5 (American Psychiatric Association, 2013). Participants reported the extent to which they experienced each symptom over a retrospective period of 2 weeks using a Likert scale ranging from 0 (not at all) to 3 (nearly every day). Previous literature has indicated a cut-off score of 10 is able to identify probable cases of generalized anxiety disorder with 89% sensitivity and 82% specificity (Spitzer et al., 2006). Internal consistency in the present study was excellent ( $α$ = .93).

Patient Health Questionnaire Depression Module (PHQ-9; Kroenke et al., 2001)

The PHQ-9 is a 9-item scale designed to measure the nine depression criteria as per the DSM-5 (American Psychiatric Association, 2013; Kroenke et al., 2001). Participants reported how much they had been bothered by each problem over the past two weeks using a Likert scale ranging from 0 (not at all) to 3 (nearly every day). A cut-off score of 10 has been shown to identify individuals likely to meet diagnostic criteria for major depressive disorder at a sensitivity and specificity of 88% (Kroenke et al., 2001). Internal consistency in the present study was excellent ( $α$ = .92).

Depression, Anxiety and Stress Scale (DASS-21; Lovibond & Lovibond, 1995)

The DASS is a 21-item scale designed to measure negative affect (depression), physiological arousal (anxiety), and general anxiety/tension (stress). Participants reported how much each statement applied to them over the past week using a Likert scale ranging from 0 (did not apply at all) to 3 (applied very much or most of the time). Internal consistency in the present study was excellent for Depression ( $α$ = .94), Anxiety ( $α$ = .90), and Stress ( $α$ = .90).

Repetitive Negative Thought Questionnaire (RNTQ; Stout et al., 2022)

The RNTQ is a 22-item scale designed to measure trans-diagnostic and disorder-specific repetitive negative thought. Participants indicated how typical each statement regarding thoughts or behaviors was to them on a Likert scale ranging from 1 (not at all typical) to 6 (extremely typical). Elevated RNTQ scores have been observed in individuals with probable anxiety and/or depressive disorders (Stout et al., 2022). The RNTQ includes four subscales: Worry, Interference, Brooding and Pessimistic Fixation. Internal consistency in the present study was excellent for all subscales (α ≥ .88).

Intolerance of Uncertainty Scale – Short Form (IUS-SF; Carleton et al., 2007)

The IUS-SF is a 12-item scale designed to measure one’s aversion to the uncertainty of a negative event occurring. Participants rated how much they agreed with each statement on a Likert scale from 1 (not at all characteristic of me) to 5 (entirely characteristic of me). Internal consistency in the present study was excellent for the total score ( $α$ = .94), Inhibitory Anxiety ( $α$ = .91), and Prospective Anxiety ( $α$ = .90).

Experimental Procedures

Decision Under Ambiguity and Risk Task

The DART was developed to expand on the principles of a similar task (Levy et al., 2010). In contrast to the previously used version, wherein participants were only presented with one option at a time, to be compared to an option that was not visually presented (i.e., a fixed reference), the DART required participants to choose between two lotteries displayed side-by-side on each trial (which reduced memory load as compared to the original version). The choice options were displayed in the form of ‘urns’, filled with a mix of red and blue marbles representing the probability of winning or losing monetary amounts displayed above each urn (Figure 1). More specifically, each visible row of marbles corresponded to approximately 10% of the contents of the urn. Gamble pairs consisted of a fixed urn that remained unchanged within a block and a variable urn that presented variations in both probability and magnitude of potential outcomes as a function of the lottery type. Participants were advised that a marble would be drawn from their selected urn. Blue marbles corresponded to positive outcomes while red marbles corresponded to negative outcomes.
Figure 1.
Schematic of risky trials from the decision under ambiguity and risk task.

Experimental Conditions

Participants were randomly assigned to one of three conditions reflecting lottery types distinguished by the valence of potential outcomes. In the gain-only condition, participants were presented with reward-only lotteries that could result in monetary gain but not loss. The G/L-separate condition consisted of both reward-only lotteries (G/L-reward; identical to those presented in the gain-only condition) and punishment-only lotteries (G/L-punishment), where money could be lost but not gained. The G/L-mixed condition contained mixed lotteries whereby drawing a blue marble resulted in monetary gain but drawing a red marble resulted in monetary loss.

The fixed urn was always a risky lottery (i.e.,both the outcome amount and probability of potential outcomes were known). On half the total trials, the variable urn was also risky, with the probability of winning being equally balanced between 10%, 20%, 30%, and 40%. In the other half of the trials, the variable urn was an ambiguous lottery in which the potential outcomes were known but their probabilities were partially occluded. Note that this manipulation resulted in two fundamentally different forms of uncertainty. In each condition, half of the blocks presented consisted of risky trials (first-order uncertainty) while the other half consisted of ambiguous trials (second-order uncertainty). The current study is solely focused on risk, with ambiguity to be the subject of a separate publication.

We aimed to present a total of 80 risky trials in each condition. Due to a technical error, the block of reward-only trials in the gain-only condition was presented twice, allowing participants to review their prior responses, and resulting in data from only 40 trials being retained. The G/L-separate condition consisted of one block of 40 risky reward-only trials and one block of 40 risky punishment-only trials. In the G/L-mixed condition, one block of 80 risky mixed trials was presented. We counterbalanced the side on which the fixed and variable options were presented. Presentation of the gamble pairs and selection of an urn was not limited by time. Within each condition, blocks were presented in a randomized order. The presentation of trials within each block was also randomized.

DART Parameters

In addition to the probabilities of the choice options, we also manipulated the potential winnings. For reward-only lotteries (presented in gain-only and G/L-separate condition) the fixed gamble constituted a 50% chance of winning +$0.50 while the variable option constituted a chance of winning the following amounts: +$0.25, +$0.50, +$0.90, +$1.70, +$3.25. For punishment-only lotteries (presented in the G/L-separate condition) the fixed gamble comprised a 50% chance of losing -$0.25, while the variable gamble could result in losing the following amounts: -$0.15, -$0.25, -$0.45, -$0.85, -$1.65. In the G/L-mixed condition, the fixed option involved a 50% chance of winning +$0.50 and a 50% chance of losing -$0.25 while the variable gamble constituted a chance of winning or losing the following outcomes: +$0.25/-$0.15, +$0.50/-$0.25, +$0.90/-$0.45, +$1.70/-$0.85, +$3.25/-$1.65. Potential loss outcomes were half the magnitude of potential gain outcomes to account for the tendency for losses to be weighted as twice as much as equivalent gains (Tversky & Kahneman, 1992). This method has been employed in similar economic decision-making tasks (e.g., Ernst et al., 2014). Potential outcomes and their associated probabilities for all risky urns across lottery types are provided in Supplemental Table S1.

Monetary Incentive

In addition to the guaranteed payment of $2.50, participants could receive a bonus ranging from $0 to $3.25 (up to an additional 130% of the guaranteed reimbursement), depending on their choices and simulated outcomes. Participants were informed that bonuses would be calculated by simulating the outcome of a gamble they had selected in a trial chosen at random. A random number was computed and compared to the probability of winning. Selecting a trial for implementation and using actual monetary payment ensured that participants treated every trial as if it were real, because they did not know which trial would be selected (cf. Camerer & Mobbs, 2017). Simulated outcomes were determined by generating a random number which was then compared to the probability of winning associated with the selected gamble. For participants allocated to the G/L-separate condition, one trial from each of the lottery types (reward-only and punishment-only) was simulated. The difference between the outcomes of the gambles was calculated and added as a bonus if it was greater than 0. For participants allocated to the G/L-mixed or gain-only conditions, one trial was simulated, and the value of the outcome was added to their payment if it was greater than 0.

Procedure

Participants were directed to Qualtrics from MTurk via a hyperlink. First, they were asked to answer demographic questions and self-report psychopathology questionnaires. They then viewed instructions for the experimental task and were given four practice trials with feedback. Participants were then randomly assigned to one of the three conditions described above (gain-only, G/L-separate, G/L-mixed) and completed the task. The median study duration was approximately 20 min (M = 25.62, SD = 16.49). Participants were paid electronically via MTurk following the experiment.

Data Analysis

Data Screening

Several data cleaning procedures were applied following data collection. Questionnaire data from 21 participants were discarded due to incomplete or missing responses on the experimental task. Based on the assumption that each item of the questionnaires would take at least 1s to answer, and that the minimum amount of time the behavioral task could be completed in was 300 s (5 min), responses from six participants who completed the study in 400 s or less were excluded. Further, an inconsistency score was calculated from DASS responses using a method analogous to the procedure employed by Conners et al. (1999). In an independent dataset from n = 39,775 participants (OpenPsychometrics, 2019), responses for four item pairs were found to be highly correlated (r > .6), indicating that the response to one question in the pair should be very similar to the response on the other question in the pair. The inconsistency index consisted of the sum of the absolute difference in responses within the item pairs. A simulation of 10,000 random responses from 5,000 participants each indicated that 95% of responses resulted in inconsistency scores of 5 or less (i.e., a score greater than 5 occurred only 5% of the time in random data). Accordingly, we excluded data from 30 participants with scores exceeding this value. To examine whether participants were attentive during the risky choice task, we separately analyzed the choices in the subset of trials in which both the fixed and the variable gambles offered identical outcomes, but the probability of the variable option was less advantageous (i.e. the fixed gamble was the clear logical choice). Overall, there were six of these trials in the gain-only condition and 12 such trials in the other two conditions. The proportion of fixed gamble choices in the aforementioned trials represented a ‘logic score’. Data from 44 participants with a logic score at or below chance were excluded, consistent with decision quality analyses performed by Ruderman et al. (2016). Differences in the number of exclusions by condition were not significant.

Psychopathology Scales

As we anticipated high intercorrelations among the psychopathology variables, we conducted a principal component analysis (PCA) to reduce the data (Suhr, 2005). This method was selected over averaging items or exploratory factor analysis as PCA ensures the contribution of each subscale toward the composite score reflects the actual variance. PCA indicated a single component (see Results), which was standardized and used in analyses to explore the influence of psychopathology on decision-making (described below). Additionally, to determine whether individuals in our sample were likely to meet diagnostic criteria for anxiety and depression, clinical groups were created using established cut-off scores.

DART Performance

For the DART, we first calculated the expected value (EV) for each gamble. The EV was calculated drawing on Cumulative Prospect Theory (CPT; Tversky & Kahneman, 1992) and expected utility theory (Von Neumann & Morgenstern, 1944):
$E V = \sum_{i = 1}^{n} p_{i} x_{i}$
where $p_{i}$ is the probability of the outcome, and $x_{i}$ is the objective value of the outcome. Further, we calculated the degree of risk for each option, defined as variance ( $V A R$ ) as per financial decision theory (Bossaerts, 2010; Markowitz, 1952):
$V A R = \sum_{i = 1}^{n} {(x_{i} - E V)}^{2} p_{i} .$

Note that EV,* $p_{i}$ , and $x_{i}$ are defined as above. The complete set of risky urns and their corresponding EV and VAR values are reported in Supplemental Table S1. In order to examine behavior in relation to variance, within each gamble pair, the option with the higher VAR value was labelled ‘riskier’ while the option with the lower VAR value was labelled ‘safer’. Based on this classification, a measure of rational choice behavior was derived by calculating the relative EV ( $Δ E V = {E V}_{S a f e r} - {E V}_{R i s k i e r}$ ) for each trial. Negative ΔEV values indicated the riskier option was the optimal choice while positive values indicated the safer option was optimal. The magnitude of ΔEV was indicative of the degree to which an option was favored.

Choice Behavior

We modelled selection of the riskier option via generalized linear mixed-effects models with a binomial distribution and logit link function, which transforms probabilities to the log-odds scale for estimation.

We began with a Null model, which included only a random effect that allowed intercepts to vary for participants. The null model presented the possibility that no variable systematically predicted choice and served as a comparison point for models including fixed effects.

To compare trials with and without gain or loss outcomes, the G/L-separate condition was split into its component lottery types, henceforth referred to as G/L-reward and G/L-punishment. Lottery types differed in the valence/s of potential outcomes, resulting in systematic variation in ΔEV. Accordingly, to improve interpretability and correct for this variation, cluster-based mean centering was implemented (Sommet & Morselli, 2017). This method separates within-cluster effects from between-cluster variation so that sensitivity to ΔEV is evaluated within each specific lottery type, preventing confounding effects due to differences in the distribution of ΔEV between lottery types.

A Base model was constructed that included task variables (ΔEV, lottery type) and their interaction as fixed effects, with participant (id) and ΔEV as random effects, allowing participants to have random intercepts and for slopes to vary by ΔEV:
$B a s e m o d e l : Riskier Choice \sim Δ EV + lottery type + Δ EV x lottery type + (1 + Δ EV | id) .$

Individual differences in psychopathology were examined by adding the psychopathology PCA component and its interactions with task variables as fixed effects to the Base model equation to construct the Psychopathology model:
$P s y c h o p a t h o l o g y m o d e l : Riskier Choice \sim Δ EV + lottery type + psychopathology + Δ EV x lottery type + Δ EV x psychopathology + lottery type x psychopathology + Δ EV x lottery type x psychopathology + (1 + Δ EV | id) .$

Model fit of the Null, Base and Psychopathology models were compared via evaluation of Akaike information criterion (AIC), which penalizes model complexity (Burnham & Anderson, 2004). The percentage of variance explained by fixed and random effects were also examined via theoretical pseudo $R^{2}$ , calculated using the ‘MuMIn’ package, as appropriate for binomial data (Bartoń, 2023). Omnibus fixed-effect significance tests were obtained using Wald type III Analysis of Variance (ANOVA) tests using the ‘car’ package (Fox & Weisberg, 2019).

As lottery types were not ordinal, there was no logical reference point. Accordingly, we applied sum coding, which constrains the coefficients to sum to zero. Under this coding scheme, the intercept represents the grand-mean log-odds of selecting the riskier option across all lottery types when the continuous variables are held constant at their centered means (i.e.,0). The fixed effects model estimates (β) therefore indicate changes in log-odds relative to this grand mean. For continuous predictors, coefficients represent linear trends (slopes), specifying the change in log-odds associated with a one-unit increase in the predictor, holding remaining predictors constant. Because coefficients under sum coding do not directly yield condition-specific predictions, to aid interpretability, we report estimated marginal means, which are identical to the sum of the model intercept and model coefficient of the predictor/s of interest. Predicted probabilities were obtained by transforming the fitted log-odds from the model using the inverse-logit function.

Index of Risk Attitude

We derived a model-based estimate of the indifference point (IP), an index of the ΔEV at which selection of riskier and safer gambles are equally probable. While ΔEV indicates which gamble is mathematically favorable according to their objective EVs, the IP signifies subjective equivalence based on choice behavior. Risk attitude can be inferred by comparing subjective and objective equivalence. For instance, subjective and objective evaluation are aligned at IP = 0, as neither option is favored mathematically or behaviorally, reflecting a risk-neutral attitude. A more negative IP indicates a stronger mathematical advantage of the riskier option is required for the predicted probability to exceed 50%, consistent with risk aversion. Conversely, a more positive IP denotes risk seeking, as selection of the riskier option exceeds chance at a point where it is not objectively advantageous, revealing greater willingness to tolerate variance even in the absence of more favorable outcomes.

Statistical Analyses

Data pre-processing and analyses of choice behavior were implemented within the statistical package R (v4.3.3; R Core Team, 2024), using RStudio (v2023.06.0; Posit team, 2023). Mixed effects modelling was conducted via the lme4 package (Bates et al., 2015) and estimated marginal means were obtained with the emmeans package (Lenth, 2023). Statistical analyses of PCA, t-tests, ANOVA, ANCOVA, post-hoc comparisons and Pearson’s correlations were performed with the statistical program JASP (v 0.18.3; JASP Team, 2024). For transparency and to support interpretation of the results, data and annotated analysis scripts are available at osf.io/cvshu.

Analyses of overall task effects were conducted followed by investigation of individual differences in psychopathology. Statistics are reported for the best fitting model. Where there were significant effects, we conducted post-hoc analyses. We implemented pairwise comparisons and linear trend analysis for categorical and continuous variables, respectively. Where there were interactions, we used combinations of pairwise and linear contrasts, as appropriate. All statistical inferences were conducted on the log-odds scale. The Benjamini-Hochberg method was used for controlling the false discovery rate (FDR) and was chosen to balance the risk of Type I and Type II errors. For post-hoc pairwise comparisons, we applied Bonferroni correction due to its conservative nature.

Results

Demographics

All sociodemographic data are described in Supplemental Table S2. Approximately 75% of the sample identified as Caucasian. Most of the sample (52.47%) had completed a college/university degree or higher education. Less than 25% of the sample had never been enrolled in a college course. More than 80% of the sample was employed and approximately 70% of participants earned between $25,001 and $100,000. Roughly 45% of participants were married and an additional 45% were single or had never been married. No significant differences by DART condition were found (p’s $\geq$ .318), indicating that demographic characteristics were comparable across conditions and therefore unlikely to confound interpretation of the individual differences in choice behavior reported below.

Psychopathology Scales

We administered the GAD-7, PHQ-9, DASS, RNTQ and IUS. There were no differences in any of the questionnaire scales or subscales by DART condition (gain-only, G/L-separate, G/L-mixed), controlling for multiple comparisons. Descriptive statistics of questionnaires are provided in Supplemental Materials (Table S3). All psychopathology questionnaires were moderately-to-highly intercorrelated (see Supplemental Table S4). PCA with promax rotation indicated a single component that accounted for 69.31% of the variance. Inclusion of a second component accounted for an additional 9.52% of the variance, which did not meaningfully contribute to the total proportion of variance explained. As such, the single component solution was selected for analysis (see scree plot, Supplemental Figure S1). Scale loadings on the first component are provided in Supplemental Table S5. Given the high loading for all questionnaires, and that the contribution of depression and anxiety measures were equivalent, the component was labelled Affective Distress (AD).

Clinical cut-off scores on the GAD-7, PHQ-9 and DASS were used to categorize individuals to assess the percentage of our sample likely to meet diagnostic criteria. Approximately 30% of the sample was likely to have an anxiety disorder and approximately 40% of the sample was likely to have a depressive disorder, with significantly higher AD scores observed in the groups likely to meet diagnostic criteria (see Supplemental Table S6).

Choice Behavior

Generalized linear mixed-effects models were used to estimate the log-odds of selecting the riskier (higher variance) gamble over the safer (lower variance) gamble. This approach enabled trial-level analysis of choice behavior in relation to both the variance and the EV of the lotteries, indicating whether choice behavior aligned with selection of the objectively advantageous option within each gamble pair (i.e., selection of the option with the higher EV). We compared the Null model (conceptually stating that no systematic effects were present), the Base model (including ΔEV and the lottery types as fixed effects of interest), and the Psychopathology model (additionally including the PCA-derived loadings on the affective distress component as a main effect, as well as all possible interaction effects). Model comparisons using the AIC showed that the best-fitting model, with the highest percentage of variance explained, was the Psychopathology model (see Supplemental Table S7). All main and interaction effects were significant in both the Base and Psychopathology models, as determined by Type III Wald $χ^{2}$ tests (see Table 1). Model estimates were largely comparable across the two models (see Supplemental Table S8), indicating that core task effects remained significant after accounting for psychopathology, though the inclusion of AD modified the magnitude of several interactions, consistent with a moderating rather than independent role. All subsequent results report the estimated marginal means of each fixed effect in the Psychopathology model, equivalent to the log-odds of selecting the riskier gamble obtained from model estimates.
Table 1.
Omnibus Test of Effects for the Base and Psychopathology Mixed-Effect Generalized Linear Models

Effect df $χ^{2}$

Base Psychopathology

Intercept 1 168.94* 204.06*

Lottery Type 3 1600.19* 1534.01*

ΔEV 1 1114.56* 1188.57*

Lottery Type × ΔEV 3 1204.65* 1224.25*

AD 1 - 27.27*

Lottery Type × AD 3 - 139.70*

ΔEV × AD 1 - 33.04*

Lottery Type × ΔEV × AD 3 - 147.28*

Note. Relative Expected Value $(Δ E V) = {E V}_{S a f e r} - {E V}_{R i s k i e r} .$ The Base model consisted of task effects only while the Psychopathology model additionally included a measure of Affective Distress (AD), which was calculated via principal components analysis from several scales measuring symptomatology of anxiety and depression.

* p < .001.

DART Effects

First, we examined the main effect of lottery type, which captures differences in the selection of the riskier (higher variance) option across the lottery types (gain-only, G/L-reward, G/L-punishment, and G/L-mixed) when holding ΔEV and AD constant at their means. The predicted probability of selecting the riskier gamble was lowest in G/L-punishment lotteries, followed by G/L-mixed, and highest in gain-only and G/L-reward lotteries (see Figure 2A). Pairwise comparisons indicated significant differences in gamble selection between all lottery types (all p’s < .001), except between gain-only and G/L-reward lotteries (p = .214). That is, lottery types involving potential losses were associated with significantly lower selection of the riskier option than lottery types involving potential gains.
Figure 2.
Main effects of lottery type and relative expected value on the predicted probability of selecting the riskier gamble.

The main effect of ΔEV reflects systematic variation in choice behavior as a function of ΔEV, averaged across lottery types and holding AD constant. Because ΔEV was calculated by subtracting the EV of the riskier option from the EV of the safer option, negative values correspond to trials where the riskier option was optimal, whereas positive values indicate that the safer option was optimal. A significant linear trend was identified such that as ΔEV increased by one unit, the log-odds of selecting the riskier gamble decreased (β = -6.37, z = -34.47, p < .001). This negative linear trend represents sensitivity to ΔEV with predicted choice generally aligning with selection of the gamble with the higher EV, as illustrated in Figure 2B. That is, the predicted probability of choosing the riskier option was higher when the riskier option was mathematically advantageous, but lower when the safer (lower variance) option was advantageous. The largest changes in predicted choice occurred where gambles within a pair were close in EV (i.e., near 0). The predicted probability of selecting the riskier option fell below the 50% threshold despite having a slightly higher EV than the safer option (IP = -0.12, 95% confidence interval [CI] [-0.17, -0.06]), indicating very slight overall risk aversion.

The interaction effect between lottery type and ΔEV evaluates whether the association between ΔEV and selection of the riskier gamble differs across the lottery types, when AD is held constant. Significant negative ΔEV linear trends were present for all lottery types, but varied in magnitude (see Table 2). Larger negative estimates indicate greater sensitivity to ΔEV, corresponding to steeper slopes when transformed into predicted probabilities, as illustrated in Figure 3. Pairwise linear comparisons revealed that sensitivity to ΔEV was greater in lotteries involving loss outcomes (G/L-punishment and G/L-mixed) than in reward-only lotteries (gain-only and G/L-reward; p’s < .001). There were no significant differences in sensitivity to ΔEV between G/L-punishment and G/L-mixed (p = .064) or between G/L-reward and gain-only lotteries (p = .064). That is, sensitivity to differences in EV between options was stronger when losses were possible, yielding larger changes in predicted choice as ΔEV varied than in lotteries without potential loss. Similarly, examination of the IPs (see Table 2) revealed differences in risk attitudes across lottery types. In lotteries with potential losses, negative IPs indicated risk aversion; however, in lotteries without loss, positive IPs signified risk seeking, such that selection of the riskier option exceeded chance even when the safer option was slightly more advantageous. The IP for G/L-punishment was significantly lower than the IPs of all other lottery types (p’s < .001), suggesting increased risk aversion in the absence of potential gains. Similarly, the IP for G/L-mixed lotteries was significantly lower than in reward-only lotteries (p’s < .001). In contrast, there was no significant difference between the IPs of reward-only lotteries (p = .140); however, the wide 95% CIs indicated imprecision in the population-level IP estimate, spanning values consistent with moderate risk aversion to moderate risk seeking, indicating that risk attitude in reward-only contexts was less precisely estimated than in loss contexts.
Table 2.
Psychopathology Model Estimates for the Lottery Type × Relative Expected Value Interaction Effect

Lottery type $β$ 95% CI for $β$ SE z IP 95% CI for IP

Gain-Only -2.68* [-3.25, -2.11] 0.29 -9.22 0.24 [-0.08, 0.56]

G/L-Reward -1.91* [-2.45, -1.38] 0.27 -7.06 0.27 [-0.20, 0.73]

G/L-Punishment -10.90* [-11.63, -10.18] 0.37 -29.45 -0.26 [-0.33, -0.20]

G/L-Mixed -9.98*** [-10.65, -9.31] 0.34 -29.27 -0.13 [-0.19, -0.06]

Note. Relative Expected Value ( $Δ E V$ ) = ${E V}_{S a f e r} - {E V}_{R i s k i e r} .$ CI = Confidence Interval; IP = indifference point, representing the ΔEV value at which participants were equally likely to choose either of the options within a gamble pair; G/L = Gain/Loss, G/L-Reward and G/L-Punishment were presented to participants allocated to the G/L-separate condition. $β$ estimates represent the change in the log-odds associated with a one unit increase in ΔEV within each lottery type.

*** p < .001.

Figure 3.
Sensitivity to relative expected value varies according to the valence of potential outcomes.

Effects of Psychopathology and Interactions with the DART

The main effect of AD indicates how selection of a riskier (higher variance) gamble varies as a function of AD when averaging across lottery types and holding ΔEV constant. A one unit increase in AD was associated with an increase in the log-odds of selecting the riskier gamble ( $β$ = 0.26, z = 5.22, p < .001). Although the predicted probability of selecting the riskier option increased as AD increased, it remained below 50% across levels of AD (see Figure 4A). This pattern is consistent with diminishing strength of risk aversion across increasing levels of AD. For those with higher levels of AD (>2 SD) predicted choice approached the 50% threshold that would represent risk neutral attitudes.
Figure 4.
Predicted probability of selecting the riskier gamble as a function of affective distress and its interactions with lottery type and relative expected value.

The interaction effect between AD and lottery type assesses whether the relationship between AD and selection of the riskier gamble varied as a function of lottery type when holding ΔEV constant. Significant positive linear trends were identified in G/L-punishment ( $β$ = 0.91, z = 9.29, p < .001) and G/L-mixed lotteries ( $β$ = 0.18, z = 2.00, p = .045). In contrast, AD did not significantly impact choice behavior in gain-only (p = .107) or G/L-reward lotteries (p = .371). For loss gambles, the predicted probability of choosing the riskier gamble increased with higher levels of AD; however, consistently remained below 50% (see Figure 4B). This pattern suggests that risk aversion was persistent in the context of loss, but reduced in strength across increasing levels of AD. Conversely, in lottery types without potential loss, the probability of selecting the riskier option was higher than 50%, indicative of risk seeking that was not significantly modified by AD. Notably, at higher levels of AD, 95% CIs broadened, particularly in G/L-punishment lotteries, reflecting reduced precision in these estimates but no change in the direction of effects.

The interaction effect between ΔEV and AD examines whether sensitivity to ΔEV varies as a function of AD when averaging across lottery types. As AD increased, the linear trend of ΔEV weakened, becoming less negative, relative to the main effect of ΔEV ( $β$ = 1.00, z = 5.75, p < .001). That is, those with low AD levels showed relatively large shifts in choice preference with changes in ΔEV, especially near ΔEV = 0 (where the EVs of the options within a gamble pair were equivalent), indicating adaptive choice adjustments, whereas those with higher AD demonstrated relatively more gradual choice adjustments reflecting reduced sensitivity to differences in EV between options (see Figure 4C). Visual inspection of the predicted probabilities plot showed the IPs corresponding to higher levels of AD exhibited a rightward shift toward 0, consistent with lessened risk aversion.

Finally, the three-way interaction between lottery type, ΔEV and AD evaluates whether the impact of AD on sensitivity to ΔEV differs across lottery types. As AD increased, the magnitude of the ΔEV linear trend decreased in G/L-punishment ( $β$ = 2.91, z = 8.97, p < .001) and G/L-mixed ( $β$ = 0.62, z = 1.96, p = .050) lotteries. In contrast, AD did not significantly moderate sensitivity to ΔEV in gain-only or G/L-reward lotteries (p’s $\geq$ .132). This indicates that higher levels of AD were associated with diminished sensitivity to differences in EV in lotteries involving potential loss, particularly in the absence of potential gain, but not in lotteries without loss (see Figure 5). Visual inspection of IPs reflected a progressive decline in risk aversion across increasing levels of AD in the context of loss. In the absence of loss, predicted choice behavior appeared consistent with risk seeking regardless of AD.
Figure 5.
Moderating effect of affective distress on sensitivity to relative expected value as a function of lottery type

Discussion

The present study investigated drivers of value-based decision-making under first-order uncertainty and examined whether affective distress symptoms influenced how variance and loss contributed to decision-making. We contrasted two theoretically motivated accounts and hypothesized that behavior driven by economic conceptualizations of risk would involve avoidance of riskier (higher variance) gambles, independent of the valence of potential outcomes ( $H_{a}$ ), whereas behavior driven by risk as defined in psychological theory would result in participants avoiding negative outcomes through selection of mathematically advantageous options ( $H_{b}$ ). We also predicted that individuals with elevated affective distress would show heightened discomfort with variance, particularly when loss was possible. Our findings partially supported these predictions. We found some evidence supporting the economic theory of risk in that selection of the riskier gamble generally required a more favorable EV relative to the safer option, reflecting mild risk aversion; however, this did not align with complete avoidance of riskier options. Distinct patterns emerged across lottery types with choice behavior heavily influenced by the presence or absence of potential loss lending support to the psychological theory of risk. In lotteries with potential loss, choice behavior reflected modest risk aversion, but was highly sensitive to differences in EV, resulting in reliable selection of the optimal choice regardless of variance while the opposite pattern was observed in lotteries with only potential gains. Contrary to our expectation of amplified avoidance of riskier gambles, individuals with higher affective distress showed greater willingness to select the riskier gambles than individuals with lower affective distress scores and slightly decreased sensitivity to EV differences in loss contexts as demonstrated by less consistent selection of the optimal gamble. These findings suggest a more nuanced relationship between affective distress and decision-making than would be predicted by discomfort with either variance or loss alone, and highlight the importance of considering valence when characterizing behavior under first-order uncertainty.

Asymmetry in behavior between gambles with and without loss is an established finding in economic decision-making tasks. We found that the probability of selecting the riskier gamble was lowest in lotteries with potential loss, while in lotteries with gain outcomes participants were more willing to select the riskier gamble. In lotteries without potential loss, this occurred even when the safer option was the rational choice, indicative of a greater tolerance of variance (i.e., risk seeking behavior). Although prospect theory predicts risk aversion for gains and risk seeking for losses (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992), reversed choice preferences have been observed when risk (variance) cannot be avoided through selection of a guaranteed option (Wright et al., 2012). In the present study, the point at which gambles involving potential loss reached subjective equivalence appeared to function as a threshold, whereby there was a strong shift aligned with selection of the optimal gamble. As such, this pattern suggests that loss enhanced sensitivity to differences in value, rather than reflecting uniform discomfort with variance. This interpretation aligns with Yechiam and Hochman (2013), who argued that potential loss results in increased sensitivity to option values, leading to more reliable selection of optimal choices. Our findings further align with von Gaudecker et al. (2011) who varied the probabilities on a set of lotteries, thereby systematically manipulating EV and variance, and found that the riskier option had to be more favorable to be selected in the presence of potential loss, while higher tolerance of variance was demonstrated when only gains were possible. Collectively, these findings suggest that choice behavior may be contingent on task structure, including the presence or absence of certainty or loss, which may be of particular importance in attempting to understand individual differences in psychopathology. By demonstrating these patterns in a large, normative sample before exploring affective distress variables, we establish a baseline understanding of decision behavior in our paradigm, which allows us to more precisely characterize how affective symptoms alter these processes.

Although the influence of affective distress was relatively small, individual differences in psychopathology accounted for additional variability in participants’ choices beyond the influence of task features. In general, when the objective EVs of options were equal, such that a risk-neutral individual would be expected to be indifferent between them, we found that increasing symptoms were associated with a greater probability of selecting the riskier gamble (although the predicted probability remained below chance across levels of affective distress). Examination across lottery types revealed this pattern was only significant in the presence of loss and was particularly evident in the absence of potential gain. Furthermore, while there was no relationship between affective distress and risky choice selection in gain lotteries, the predicted probability remained above chance, further demonstrating the risk-seeking attitude revealed in the task effects. While participants were generally sensitive to differences in EV in lotteries with loss, relative to individuals with lower levels of affective distress, those with elevated affective distress demonstrated diminished sensitivity to EV. This resulted in less reliable selection of the advantageous gamble. Higher affective distress was associated with reduced selection of the riskier gamble when it was objectively optimal. When evaluated in isolation, this may appear to be consistent with elevated discomfort with variance; however, selection of riskier gamble was simultaneously more likely when the safer option was advantageous. This pattern may indicate that affective distress disrupts the calibration of risk assessment in relation to potential losses, leading to decision-making that is less guided by objective evaluation of outcomes.

Individuals with higher affective distress may have struggled to integrate key decision variables (e.g., EV and variance) in the presence of loss, opting for higher probabilities of avoiding loss or obtaining a lower magnitude loss (e.g., preferring a 90% chance of losing 15c over the option with a 50% chance of losing 25c), even when such choices were less optimal. Interpretation of our findings is bolstered by the appraisal-tendency framework (Lerner & Keltner, 2000) and the risk-as-feelings hypothesis (Loewenstein et al., 2001). Both theories underscore the role of emotions in risk and decision-making, with the appraisal-tendency-framework focusing on the effects of specific emotions and the risk-as-feelings hypothesis emphasizing the conflict between emotions and cognitive evaluation. In other words, the presence of loss may have decreased rational appraisal in favor of affective appraisal, resulting in suboptimal choices. Our findings align with established distinctions in the decision-making literature between integral affect (emotion directly relevant to a decision) and incidental affect (background emotional states unrelated to the decision at hand). While integral affect can serve as an adaptive heuristic for guiding decisions (Slovic et al., 2007), the affective distress measured in our study represents incidental affect (i.e., how an individual is feeling unrelated to the task at hand) that appears to interfere with optimal decision-making. This interference aligns with well-documented evidence that incidental emotions can bias judgment processes (Lerner et al., 2015; Schwarz & Clore, 1983). Our findings demonstrate how persistent affective distress symptoms interact with loss contexts to produce specific patterns of suboptimal choice behavior. For individuals with higher levels of affective distress, rather than simply increasing risk aversion as might be expected, incidental negative emotions appear to disrupt EV estimation processes specifically when loss is possible. This suggests that the everyday decision-making challenges faced by individuals with elevated anxiety and depression may be most pronounced in contexts involving potential negative outcomes, where emotional states may cloud rational evaluation of options. In the real world, this means people with elevated affective distress might be more prone to taking harmful risks or avoiding necessary risks, impacting their well-being, relationships, and economic stability. Underlying affect may have a potentially more persistent influence on decision-making than decision-related components and may be particularly insidious when threat, harm, or loss is possible.

Our results also speak to ongoing debates about the mechanisms underlying altered risky decision-making in anxiety and related forms of psychopathology. Charpentier and colleagues (2017) modelled choices within a prospect theory framework and found that anxiety broadly enhanced risk aversion (diminishing sensitivity to value), independent of loss aversion (weighting of losses relative to equivalent gains). The present study quantified risk as degree of variance, and examined discomfort with potential loss, finding that altered risk attitudes in affective distress are specifically modulated by the presence of potential loss, particularly in the absence of potential gains. The conflicting findings are likely to be a result of differences in methodological and analytic approaches. The selective influence of affective distress on choices in the context of loss reflects a pattern that cannot be captured by a single diminishing sensitivity parameter because such aggregation may obscure distortions in decisions about losses that do not generalize to decisions about gains. These findings further demonstrate that modelling loss-aversion, which requires mixed gambles where potential gain and loss outcomes can be directly compared, may not be sufficient to evaluate how attitudes toward loss interact with risky decision-making and individual differences in affective distress.

Our dimensionality reduction of psychopathology measures resulted in one component, Affective Distress, that did not differentiate between anxiety and depression. This single component is congruent with the suggestion that common psychiatric disorders are characterized by an underlying General Psychopathology dimension, also referred to as the p factor, which is associated with symptom severity and functional impairment (Caspi et al., 2014). Additionally, the use of an online, non-clinical sample resulted in access to a large sample for notable statistical power that was well suited for identification of individual differences in symptomology and behavior (Gillan & Daw, 2016; Kraemer et al., 2004). More distinct patterns of behavior may have been found if comorbidity of disorders was accounted for; however, given the suggestion that concurrence of anxiety and depression is more common than pure cases of each disorder (Preisig et al., 2001), our findings may have greater ecological validity than those using case-control comparisons. Recent evidence suggests that elevated risk aversion may be specifically altered in individuals with anxiety (Sporrer et al., 2024). In a sample of patients with major depressive disorder induced worry increased risk aversion (as defined in prospect theory) only in those with comorbid generalized anxiety symptoms, with similar baseline levels of risk and loss aversion across patient and control groups. Future research should examine whether comparable findings are observed outside of the prospect theory framework, in paradigms where uncertainty cannot be avoided by rejecting probabilistic options and that also include conditions with and without the possibility of gains.

Our findings should be interpreted in light of study limitations. Our gambles featured equivalent gamble outcomes across lottery types (i.e., the potential gain and loss outcomes for mixed gambles were identical to the potential gain and loss outcomes in reward and punishment gambles, respectively), but this resulted in an imbalanced range of EV and VAR values. Future research should implement standardized EV and VAR values across lottery types even though this may result in gamble outcomes that are not matched across lottery types. This approach may aid in directly comparing the contribution of risk as conceptualized by economic and psychological theories. Due to a programming error, participants allocated to the gain-only condition received half the number of trials as participants in the other conditions. They also were able to review their previous answers when making subsequent decisions. This may have led to inconsistencies in findings between the reward-only gambles. Alternatively, exposure to G/L-punishment lotteries may have influenced subsequent choice behavior in G/L-reward lotteries contributing to inconsistencies with gain-only lotteries (Schneider et al., 2016). Although evidence suggests that exposure to gains and losses can influence behavior when participants receive feedback following a choice (e.g., Giorgetta et al., 2012; Shapiro et al., 2020), our task did not present feedback. We additionally note that participants were also presented with blocks of ambiguous trials to be the subject of a separate manuscript. The current study was not sufficiently powered to conclusively examine potential order effects; however, previous studies employing the task from which the DART was derived have neither found evidence of order effects between alternate presentations of gain and loss blocks (e.g., Pushkarskaya et al., 2017), nor reported order effects in relation to risky and ambiguous trials (e.g., Jia et al., 2020; Ruderman et al., 2016). Nonetheless, we acknowledge that it is possible that behavior was impacted by exposure to different combinations of valence and types of uncertainty. Finally, our investigation was limited to economic choices, but it would be beneficial to expand examination of choice behavior to non-monetary domains such as social risk (see Moutoussis et al., 2021). Future research could more directly investigate the extent to which prior experience with different decision-making contexts influences risk preferences, such as when social influence is present (Mislavsky & Gaertig, 2022).

This study advances the understanding of how affective distress influences decision-making under first-order uncertainty. Using a data-driven dimensionality reduction approach, we uncovered variation in affective distress symptoms in the general population that were relevant for economic choice behavior. In the presence of loss, although participants generally demonstrated sensitivity to differences in value to select mathematically advantageous choices, this tendency was comparatively diminished in individuals with higher levels of affective distress. In contrast, affective distress did not influence choice selection in the absence of potential loss. Our findings underscore the role of incidental emotions in shaping decision-making, aligning with psychological theories like the appraisal-tendency framework and the risk-as-feelings hypothesis, which highlight how emotions can override rational cognitive processes in risk assessment.

Supplemental Material

Supplemental Material - Disentangling Components of Suboptimal Risky Decision-Making in Affective Distress

Supplemental Material for Disentangling Components of Suboptimal Risky Decision-Making in Affective Distress by Kiran Sutcliffe, Sarah M. Tashjian, Nicholas T. Van Dam in Journal of Experimental Psychopathology.

Effect	df	$χ^{2}$
Intercept	1	168.94***	204.06***
Lottery Type	3	1600.19***	1534.01***
ΔEV	1	1114.56***	1188.57***
Lottery Type × ΔEV	3	1204.65***	1224.25***
AD	1	-	27.27***
Lottery Type × AD	3	-	139.70***
ΔEV × AD	1	-	33.04***
Lottery Type × ΔEV × AD	3	-	147.28***

Lottery type	$β$	95% CI for $β$	SE	z	IP	95% CI for IP
Gain-Only	-2.68***	[-3.25, -2.11]	0.29	-9.22	0.24	[-0.08, 0.56]
G/L-Reward	-1.91***	[-2.45, -1.38]	0.27	-7.06	0.27	[-0.20, 0.73]
G/L-Punishment	-10.90***	[-11.63, -10.18]	0.37	-29.45	-0.26	[-0.33, -0.20]
G/L-Mixed	-9.98***	[-10.65, -9.31]	0.34	-29.27	-0.13	[-0.19, -0.06]

Footnotes

Acknowledgments

We acknowledge Aboriginal and Torres Strait Islander people of the unceded land on which we work, learn and live. We pay respect to Elders past, present and future, and acknowledge the importance of Indigenous knowledge in the Academy.

ORCID iD

Kiran Sutcliffe

Ethical Considerations

This study was approved by the Human Research Ethics Committee of the University of Melbourne (ID 1852118).

Consent to Participate

All participants provided informed consent electronically as indicated by selecting the statement “I consent to participate in this study” displayed via Qualtrics.

Author Contributions

Kiran Sutcliffe: Formal analysis; Methodology, Visualization, Writing – original draft; Writing – review & editing

Sarah M. Tashjian: Supervision; Writing – review & editing

Nicholas T. Van Dam: Conceptualization; Investigation, Methodology; Supervision; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: KS is supported by an Australian Government Research Training Program (RTP) Scholarship. NTVD is supported by the Contemplative Studies Centre at the University of Melbourne, established via a philanthropic gift from the Three Springs Foundation, Pty, Ltd. SMT is supported by an Australian National Health and Medical Research Council (NHMRC) Investigator Grant (2033400) and a Brain and Behavior Research Foundation Young Investigator Award (30788).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data and code are openly available in the Open Science Framework at .

Supplemental Material

Supplemental material for this article is available online.

References

American Psychiatric Association . (2013). Diagnostic and statistical manual of mental disorders (5th ed.). American Psychiatric Association. https://doi.org/10.1176/appi.books.9780890425596

Aven

(2012). The risk concept—Historical and recent development trends. Reliability Engineering & System Safety, 99, 33–44. https://doi.org/10.1016/j.ress.2011.11.006

Bach

D. R.

Dolan

R. J.

(2012). Knowing how much you don’t know: A neural organization of uncertainty estimates. Nature Reviews Neuroscience, 13(8), 572–586. https://doi.org/10.1038/nrn3289

Bartoń

(2023). MuMIn: Multi-Model Inference. R package version 1.47.5. https://CRAN.R-project.org/package=MuMIn

Bates

Mächler

Bolker

Walker

(2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1–48. https://doi.org/10.18637/jss.v067.i01

Bishop

S. J.

Gagne

(2018). Anxiety, Depression, and Decision Making: A Computational Perspective. Annual Review of Neuroscience, 41(1), 371–388. https://doi.org/10.1146/annurev-neuro-080317-062007

Bossaerts

(2010). Risk and risk prediction error signals in anterior insula. Brain Structure and Function, 214(5–6), 645–653. https://doi.org/10.1007/s00429-010-0253-1

Botelho

Fernandes

Campos

Seixas

Pasion

Garcez

Ferreira-Santos

Barbosa

Maques-Teixeira

Paiva

T. O.

(2023). Uncertainty deconstructed: Conceptual analysis and state-of-the-art review of the ERP correlates of risk and ambiguity in decision-making. Cognitive, Affective, & Behavioral Neuroscience, 23(3), 522–542. https://doi.org/10.3758/s13415-023-01101-8

Brown

V. A.

(2021). An introduction to linear mixed-effects modeling in R. Advances in Methods and Practices in Psychological Science, 4(1), Article 2515245920960351. https://doi.org/10.1177/2515245920960351

10.

Burnham

K. P.

Anderson

D. R.

(2004). Multimodel Inference: Understanding AIC and BIC in Model Selection. Sociological Methods & Research, 33(2), 261–304. https://doi.org/10.1177/0049124104268644

11.

Butler

Mathews

(1983). Cognitive processes in anxiety. Advances in Behaviour Research and Therapy, 5(1), 51–62. https://doi.org/10.1016/0146-6402(83)90015-2

12.

Camerer

C. F.

(1989). An experimental test of several generalized utility theories. Journal of Risk and Uncertainty, 2(1), 61–104. https://doi.org/10.1007/BF00055711

13.

Camerer

Mobbs

(2017). Differences in behavior and brain activity during hypothetical and real choices. Trends in Cognitive Sciences, 21(1), 46–56. https://doi.org/10.1016/j.tics.2016.11.001

14.

Carleton

R. N.

(2016). Into the unknown: A review and synthesis of contemporary models involving uncertainty. Journal of Anxiety Disorders, 39, 30–43. https://doi.org/10.1016/j.janxdis.2016.02.007

15.

Carleton

R. Nicholas

Norton

M. A. Peter J.

Asmundson

Gordon J. G.

(2007). Fearing the unknown: A short version of the Intolerance of Uncertainty Scale. Journal of Anxiety Disorders 21(1):105−117. https://linkinghub.elsevier.com/retrieve/pii/S088761850600051X, https://doi.org/10.1016/j.janxdis.2006.03.014

16.

Caspi

Houts

R. M.

Belsky

D. W.

Goldman-Mellor

S. J.

Harrington

Israel

Meier

M. H.

Ramrakha

Shalev

Poulton

Moffitt

T. E.

(2014). The p factor: One general psychopathology factor in the structure of psychiatric disorders? Clinical Psychological Science, 2(2), 119–137. https://doi.org/10.1177/2167702613497473

17.

Charpentier

C. J.

Aylward

Roiser

J. P.

Robinson

O. J.

(2017). Enhanced risk aversion, but not loss aversion, in unmedicated pathological anxiety. Biological Psychiatry, 81(12), 1014–1022. https://doi.org/10.1016/j.biopsych.2016.12.010

18.

Chung

Kadlec

Aimone

J. A.

McCurry

King-Casas

Chiu

P. H.

(2017). Valuation in major depression is intact and stable in a non-learning environment. Scientific Reports, 7(1), Article 44374. https://doi.org/10.1038/srep44374

19.

Clark

L. A.

Watson

(1991). Tripartite model of anxiety and depression: Psychometric evidence and taxonomic implications. Journal of Abnormal Psychology, 100(3), 316–336. https://doi.org/10.1037//0021-843x.100.3.316

20.

Conners

C. K.

Erhardt

Sparrow

E. P.

(1999). Conners’ adult ADHD rating scales: Technical manual. Multi Health Systems (MHS).

21.

Ernst

Plate

R. C.

Carlisi

C. O.

Gorodetsky

Goldman

Pine

D. S.

(2014). Loss aversion and 5HTT gene variants in adolescent anxiety. Developmental Cognitive Neuroscience, 877–85. https://doi.org/10.1016/j.dcn.2013.10.002

22.

Faul

Erdfelder

Buchner

Lang

A. G.

(2009). Statistical power analyses using G* Power 3.1: Tests for correlation and regression analyses. Behavior research methods, 41(4), 1149–1160. https://doi.org/10.3758/BRM.41.4.1149

23.

FeldmanHall

Glimcher

Baker

A. L.

Phelps

E. A.

(2016). Emotion and decision-making under uncertainty: Physiological arousal predicts increased gambling during ambiguity but not risk. Journal of Experimental Psychology: General, 145(10), 1255–1262. https://doi.org/10.1037/xge0000205

24.

Fox

Weisberg

(2019). An R companion to applied regression. Third Sage publications. https://www.john-fox.ca/Companion/

25.

Galván

Peris

T. S.

(2014). Neural correlates of risky decision making in anxious youth and healthy controls. Depression and Anxiety, 31(7), 591–598. https://doi.org/10.1002/da.22276

26.

Gillan

C. M.

Daw

N. D.

(2016). Taking psychiatry research online. Neuron, 91(1), 19–23. https://doi.org/10.1016/j.neuron.2016.06.002

27.

Giorgetta

Grecucci

Zuanon

Perini

Balestrieri

Bonini

Sanfey

A. G.

Brambilla

(2012). Reduced risk-taking behavior as a trait feature of anxiety. Emotion, 12(6), 1373–1383. https://doi.org/10.1037/a0029119

28.

Broster

L. S.

Jiang

Yang

Luo

Y.-J.

(2017). Trait anxiety and economic risk avoidance are not necessarily associated: Evidence from the framing effect. Frontiers in Psychology, 8, 92. https://doi.org/10.3389/fpsyg.2017.00092

29.

Hagiwara

Mochizuki

Chen

Lei

Hirotsu

Matsubara

Nakagawa

(2022). Nonlinear Probability Weighting in Depression and Anxiety: Insights From Healthy Young Adults. Frontiers in Psychiatry, 13, Article 810867. https://doi.org/10.3389/fpsyt.2022.810867

30.

Hartley

C. A.

Phelps

E. A.

(2012). Anxiety and decision-making. Biological Psychiatry, 72(2), 113–118. https://doi.org/10.1016/j.biopsych.2011.12.027

31.

Hertwig

Wulff

D. U.

Mata

(2018). Three gaps and what they may mean for risk preference. Philosophical Transactions of the Royal Society B, 374(1766), Article 20180140. https://doi.org/10.1098/rstb.2018.0140

32.

Holt

C. A.

Laury

S. K.

(2002). Risk aversion and incentive effects. American Economic Review, 92(5), 1644–1655. https://doi.org/10.1257/000282802762024700

33.

JASP Team . (2024). JASP (Version 0.18.3). [Computer software].

34.

Jia

Furlong

Gao

Santos

L. R.

Levy

(2020). Learning about the Ellsberg Paradox reduces, but does not abolish, ambiguity aversion. PLOS ONE, 15(3), Article e0228782. https://doi.org/10.1371/journal.pone.0228782

35.

Kahneman

Tversky

(1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263. https://doi.org/10.2307/1914185

36.

Katz

B. A.

Matanky

Aviram

Yovel

(2020). Reinforcement sensitivity, depression and anxiety: A meta-analysis and meta-analytic structural equation model. Clinical Psychology Review, 77, Article 101842. https://doi.org/10.1016/j.cpr.2020.101842

37.

Kraemer

H. C.

Noda

O’Hara

(2004). Categorical versus dimensional approaches to diagnosis: Methodological challenges. Journal of Psychiatric Research, 38(1), 17–25. https://doi.org/10.1016/S0022-3956(03)00097-9

38.

Kroenke

Spitzer

R. L.

Williams

J. B. W.

(2001). The PHQ-9: Validity of a brief depression severity measure. Journal of General Internal Medicine, 16(9), 606–613. https://doi.org/10.1046/j.1525-1497.2001.016009606.x

39.

Kumle

Võ

M. L.-H.

Draschkow

(2021). Estimating power in (generalized) linear mixed models: An open introduction and tutorial in R. Behavior Research Methods, 53(6), 2528–2543. https://doi.org/10.3758/s13428-021-01546-0

40.

Lenth

R. V.

(2023). emmeans: Estimated marginal means, aka least-squares means. R Package Version 1.8.9. https://CRAN.R-project.org/package=emmeans

41.

Lerner

J. S.

Keltner

(2000). Beyond valence: Toward a model of emotion-specific influences on judgement and choice. Cognition & Emotion, 14(4), 473–493. https://doi.org/10.1080/026999300402763

42.

Lerner

J. S.

Valdesolo

Kassam

K. S.

(2015). Emotion and Decision Making. Annual Review of Psychology, 66(1), 799–823. https://doi.org/10.1146/annurev-psych-010213-115043

43.

Levy

Snell

Nelson

A. J.

Rustichini

Glimcher

P. W.

(2010). Neural Representation of Subjective Value Under Risk and Ambiguity. Journal of Neurophysiology, 103(2), 1036–1047. https://doi.org/10.1152/jn.00853.2009

44.

Loewenstein

G. F.

Weber

E. U.

Hsee

C. K.

Welch

(2001). Risk as feelings. Psychological Bulletin, 127(2), 267–286. https://doi.org/10.1037/0033-2909.127.2.267

45.

Lovibond

P. F.

Lovibond

S. H.

(1995). The structure of negative emotional states: Comparison of the Depression Anxiety Stress Scales (DASS) with the Beck Depression and Anxiety Inventories. Behaviour Research and Therapy, 33(3), 335–343. https://doi.org/10.1016/0005-7967(94)00075-U

46.

Markowitz

(1952). Portfolio selection. Journal of Finance, 7(1), 77–91. https://doi.org/10.1111/j.1540-6261.1952.tb01525.x

47.

Mata

Frey

Richter

Schupp

Hertwig

(2018). Risk Preference: A View from Psychology. Journal of Economic Perspectives, 32(2), 155–172. https://doi.org/10.1257/jep.32.2.155

48.

Mathieu

J. E.

Aguinis

Culpepper

S. A.

Chen

(2012). Understanding and estimating the power to detect cross-level interaction effects in multilevel modeling. Journal of Applied Psychology, 97(5), 951–966. https://doi.org/10.1037/a0028380

49.

Matuschek

Kliegl

Vasishth

Baayen

Bates

(2017). Balancing Type I error and power in linear mixed models. Journal of memory and language, 94, 305–315. https://doi.org/10.1016/j.jml.2017.01.001

50.

McEvoy

P. M.

Mahoney

A. E. J.

(2012). To Be Sure, To Be Sure: Intolerance of Uncertainty Mediates Symptoms of Various Anxiety Disorders and Depression. Behavior Therapy, 43(3), 533–545. https://doi.org/10.1016/j.beth.2011.02.007

51.

Miranda

Mennin

D. S.

(2007). Depression, Generalized Anxiety Disorder, and Certainty in Pessimistic Predictions about the Future. Cognitive Therapy and Research, 31(1), 71–82. https://doi.org/10.1007/s10608-006-9063-4

52.

Mislavsky

Gaertig

(2022). Combining probability forecasts: 60% and 60% is 60%, but likely and likely is very likely. Management Science, 68(1), 541–563. https://doi.org/10.1287/mnsc.2020.3902

53.

Moutoussis

Garzón

Neufeld

Bach

D. R.

Rigoli

Goodyer

Bullmore

Guitart-Masip

Dolan

R. J.

Fonagy

Jones

Hauser

Romero-Garcia

St Clair

Vértes

Whitaker

Inkster

Prabhu

Ooi

Kievit

(2021). Decision-making ability, psychopathology, and brain connectivity. Neuron, 109(12), 2025–2040.e7. https://doi.org/10.1016/j.neuron.2021.04.019

54.

OpenPsychometrics . (2019). Depression Anxiety Stress Scale. [Data set]. https://openpsychometrics.org/_rawdata/DASS_data_21.02.19.zip

55.

Paulus

M. P.

A. J.

(2012). Emotion and decision-making: Affect-driven belief systems in anxiety and depression. Trends in Cognitive Sciences, 16(9), 476–483. https://doi.org/10.1016/j.tics.2012.07.009

56.

Phelps

E. A.

Lempert

K. M.

Sokol-Hessner

(2014). Emotion and decision making: Multiple modulatory neural circuits. Annual Review of Neuroscience, 37(1), 263–287. https://doi.org/10.1146/annurev-neuro-071013-014119

57.

Posit team . (2023). RStudio: Integrated Development Environment for R. Posit Software. PBC. https://www.posit.co/

58.

Preisig

Merikangas

K. R.

Angst

(2001). Clinical significance and comorbidity of subthreshold depression and anxiety in the community. Acta Psychiatrica Scandinavica, 104(2), 96–103. https://doi.org/10.1034/j.1600-0447.2001.00284.x

59.

Pushkarskaya

Tolin

Ruderman

Henick

Kelly

J. M.

Pittenger

Levy

(2017). Value-based decision making under uncertainty in hoarding and obsessive-compulsive disorders. Psychiatry Research, 258, 305–315. https://doi.org/10.1016/j.psychres.2017.08.058

60.

R Core Team . (2024). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. https://www.R-project.org/

61.

Raghunathan

Pham

M. T.

(1999). All negative moods are not equal: Motivational influences of anxiety and sadness on decision making. Organizational Behavior and Human Decision Processes, 79(1), 56–77. https://doi.org/10.1006/obhd.1999.2838

62.

Rangel

Camerer

Montague

P. R.

(2008). A framework for studying the neurobiology of value-based decision making. Nature Reviews Neuroscience, 9(7), 545–556. https://doi.org/10.1038/nrn2357

63.

Ruderman

Ehrlich

D. B.

Roy

Pietrzak

R. H.

Harpaz-Rotem

Levy

(2016). Posttraumatic stress symptoms and aversion to ambiguous losses in combat veterans: PTSD and aversion to ambiguous losses. Depression and Anxiety, 33(7), 606–613. https://doi.org/10.1002/da.22494

64.

Santomauro

D. F.

Mantilla Herrera

A. M.

Shadid

Zheng

Ashbaugh

Pigott

D. M.

Abbafati

Adolph

Amlag

J. O.

Aravkin

A. Y.

Bang-Jensen

B. L.

Bertolacci

G. J.

Bloom

S. S.

Castellano

Castro

Chakrabarti

Chattopadhyay

Cogen

R. M.

Collins

J. K.

Ferrari

A. J.

(2021). Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic. The Lancet, 398(10312), 1700–1712. https://doi.org/10.1016/S0140-6736(21)02143-7

65.

Schmidt

Kanis

Holroyd

C. B.

Miltner

W. H. R.

Hewig

(2018). Anxious gambling: Anxiety is associated with higher frontal midline theta predicting less risky decisions. Psychophysiology, 55(10), Article e13210. https://doi.org/10.1111/psyp.13210

66.

Schneider

Kauffman

Ranieri

(2016). The effects of surrounding positive and negative experiences on risk taking. Judgment and Decision Making, 11(5), 424–440. https://doi.org/10.1017/S1930297500004538

67.

Schonberg

Fox

C. R.

Poldrack

R. A.

(2011). Mind the gap: Bridging economic and naturalistic risk-taking with cognitive neuroscience. Trends in Cognitive Sciences, 15(1), 11–19. https://doi.org/10.1016/j.tics.2010.10.002

68.

Schwarz

Clore

G. L.

(1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45(3), 513–523. https://doi.org/10.1037/0022-3514.45.3.513

69.

Shapiro

M. S.

Price

P. C.

Mitchell

(2020). Carryover of domain-dependent risk preferences in a novel decision-making task. Judgment and Decision Making, 15(6), 1009–1023. https://doi.org/10.1017/S1930297500008202

70.

Slovic

Finucane

M. L.

Peters

MacGregor

D. G.

(2007). The affect heuristic. European Journal of Operational Research, 177(3), 1333–1352. https://doi.org/10.1016/j.ejor.2005.04.006

71.

Sommet

Morselli

(2017). Keep calm and learn multilevel logistic modeling: A simplified three-step procedure using Stata, R, Mplus, and SPSS. International Review of Social Psychology, 30(1), 203–218. https://doi.org/10.5334/irsp.90

72.

Spitzer

R. L.

Kroenke

Williams

J. B. W.

Löwe

(2006). A brief measure for assessing Generalized Anxiety Disorder: The GAD-7. Archives of Internal Medicine, 166(10), 1092–1097. https://doi.org/10.1001/archinte.166.10.1092

73.

Sporrer

J. K.

Johann

Chumbley

Robinson

O. J.

Bach

D. R.

(2024). Induced worry increases risk aversion in patients with generalized anxiety. Behavioural Brain Research, 474, Article 115192. https://doi.org/10.1016/j.bbr.2024.115192

74.

Stout

Z. E.

Fletcher

Sanderson

W. C.

Van Dam

N. T.

(2022). Development and validation of the Repetitive Negative Thoughts Questionnaire (RNTQ). PsyArXiv. [Preprint]. https://doi.org/10.31234/osf.io/3c5z7

75.

Suhr

Principal component analysis vs. Exploratory factor analysis (2005, April 10-13). SUGI 30 Proceedings, Philadelphia, PA, United States https://support.sas.com/resources/papers/proceedings/proceedings/sugi30/203-30.pdf.

76.

Tversky

Kahneman

(1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323. https://doi.org/10.1007/bf00122574

77.

Tymula

Rosenberg Belmaker

L. A.

Ruderman

Glimcher

P. W.

Levy

(2013). Like cognitive function, decision making across the life span shows profound age-related changes. Proceedings of the National Academy of Sciences, 110(42), 17143–17148. https://doi.org/10.1073/pnas.1309909110

78.

Von Neumann

Morgenstern

(1944). Theory of games and economic behavior. (pp. xviii–625). Princeton University Press.

79.

von Gaudecker

H.-M.

Van Soest

Wengström

(2011). Heterogeneity in risky choice behavior in a broad population. American Economic Review, 101(2), 664–694. https://doi.org/10.1257/aer.101.2.664

80.

World Health Organization . (2022). World mental health report: Transforming mental health for all. World Health Organization.

81.

Wright

N. D.

Symmonds

Dolan

R. J.

(2013). Distinct encoding of risk and value in economic choice between multiple risky options. NeuroImage, 81, 431–440. https://doi.org/10.1016/j.neuroimage.2013.05.023

82.

Wright

N. D.

Symmonds

Hodgson

Fitzgerald

T. H. B.

Crawford

Dolan

R. J.

(2012). Approach–Avoidance Processes Contribute to Dissociable Impacts of Risk and Loss on Choice. The Journal of Neuroscience, 32(20), 7009–7020. https://doi.org/10.1523/JNEUROSCI.0049-12.2012

83.

Van Dam

N. T.

van Tol

M.-J.

Shen

Cui

Qin

Aleman

Fan

Luo

(2020). Amygdala–prefrontal connectivity modulates loss aversion bias in anxious individuals. NeuroImage, 218, Article 116957. https://doi.org/10.1016/j.neuroimage.2020.116957

84.

Yechiam

Hochman

(2013). Losses as modulators of attention: Review and analysis of the unique effects of losses over gains. Psychological Bulletin, 139(2), 497–518. https://doi.org/10.1037/a0029383

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.71 MB

Effect	df	$χ^{2}$
Effect	df	Base	Psychopathology
Intercept	1	168.94***	204.06***
Lottery Type	3	1600.19***	1534.01***
ΔEV	1	1114.56***	1188.57***
Lottery Type × ΔEV	3	1204.65***	1224.25***
AD	1	-	27.27***
Lottery Type × AD	3	-	139.70***
ΔEV × AD	1	-	33.04***
Lottery Type × ΔEV × AD	3	-	147.28***