Abstract
Adolescent decision-making has been characterized as risky, and a heightened reward sensitivity may be one of the aspects contributing to riskier choice-behavior. Previous studies have targeted reward-sensitivity in adolescence and the neurobiological mechanisms of reward processing in the adolescent brain. In recent examples, researchers aim to disentangle the contributions of risk- and reward-sensitivity to adolescent risk-taking. Here, we discuss recent findings of adolescent's risk preferences and the associated neural mechanisms. We highlight potential frameworks that target individual differences in risk preferences in an effort to understand adolescent risk-taking, and with an ultimate goal of leveraging undesirable levels of risk taking.
Introduction
Adolescence is a transitional life phase, marking the change from childhood to adulthood. It is a period in life known for changes in motivated, goal-directed behavior, and is associated with a heightened sensitivity to rewards and a greater tendency for risk taking. For instance, adolescents and young adults take more daily-life risks than children and adults [1]. Adolescents also rate themselves higher than children and adults on reward-drive [2], sensation-seeking [3], and show a heightened neural response to rewards [2,4,5]. This heightened reward-sensitivity may serve as a differential susceptibility marker that makes some adolescents more sensitive to the environment than others [6]. For instance, research found that adolescents with high compared to low neural activation to rewards conformed more to observed peer norms in both positive and negative directions [7]. This emphasis on potential rewards (e.g., money, power, acceptance) may also lead to risky decision-making by steering adolescents towards the high-reward, high risk options. The strong focus on studying adolescent reward-sensitivity, however, may have limited a developmental perspective on how adolescents process risk. Although correlated (and sometimes even conflated) with reward, risk preference is not the same as reward-sensitivity. In this concise review, we explore to what extent adolescent decision-making is related to changes in risk preferences and what neural mechanisms underlie adolescent risk-processing.
Measuring risk preference in adolescent risky decision-making
Risk has been defined in various ways, ranging from psychological definitions such as a potential danger or loss to more economic definitions such as a higher variance in outcomes [8]. The extent to which a person responds to risk and/or chooses the riskier option in their environment is called an individual's risk preference, also referred to as “risk attitude,” “risk tolerance” or “sensitivity to risk” [9]. Individual's risk preferences have been related to real-world consequences, such as misbehavior in school and a lower likelihood to graduate [10], drug use, not wearing a seat-belt, or financial insecurity [11, 12, 13].
The measurement of risk preference knows two traditions. First, risk preferences have been revealed by using monetary lotteries in experimental studies examining how people make decisions under risk. Typically, people are asked to choose between options with explicitly stated risks and rewards (e.g., (A) do you prefer 45 Euro for sure, or (B) 100 dollar with a probability of .5?). A risk neutral person would follow the objectively calculated expected value when choosing between these options (hence choose B). A risk-averse person may overweigh the utility of the sure option over the riskier option (and hence choose A). Another way in which risk preferences have been measured is by using stated preferences based on self-reports. Numerous questionnaires exist for adult and adolescent samples that also assess domain-specificity of risk taking (e.g., social, financial, recreational, health-safety). Findings indicate that even a one-shot question may result in reliable risk preferences (e.g., “would you describe yourself as someone who tries to avoid risks (risk-averse) or as someone who is willing to take risks (risk-prone)?“) [12]. A recent study by Frey et al. [14] showed that self-reports outperformed behavioral risk-taking tasks in terms of reliability, retest-stability and validity. Moreover, in a latent-modeling approach a stable, “trait,” factor emerged from self-reports that qualified as an overarching, domain-general risk preference, as well as a series of factors capturing (domain) specific aspects of risk preference. These findings indicate the importance of measurement when studying risk preference. In addition, they indicate that individual's risk preference comprises both general and domain-specific dimensions.
Even if risk preference is a stable trait, it may still show mean-level changes across development. Empirical evidence from large world-wide longitudinal studies suggest an increase in stated risk-seeking preferences in adolescence and early adulthood [1,15], followed by a decline across the adult lifespan [16,17]. However, variations in developmental changes may occur across different domains of risk taking (i.e., health-safety, recreational, social, ethical, and/or financial domains) [15,16]. Behavioral tasks have shown mixed findings regarding age differences in revealed risk preferences, with a meta-analysis showing a decrease in risk-seeking preferences from adolescence to adulthood [18]. It has been suggested that age differences in risk preferences might be moderated by specific task characteristics, including whether measures tap into learning, or cognitive functions [18,19].Thus, although findings are mixed depending on the measure used, risk preferences seem subject to developmental fluctuations.
Adolescent risk preference in context
Risk preference may also show elements of state specificity. With states we refer to relatively changeable variations around a person's mean risk preferences that may be associated with situational factors. One particular situation of interest to adolescence is a social context. Adolescence is considered a key period for social development with a social reorientation towards peers, and a heightened sensitivity towards meeting social needs [20, 21, 22, 23]. Studies have included different social contexts to examine the impact on adolescent risky decision-making.
One particular context is social observation: in which a peer, friend, or other observes the behavior of the adolescent. Research has shown that adolescents risk preference increases when their choices are monitored by a peer [24∗, 25, 26, 27] (but see [28, 29, 30]), while the mere presence of a peer was not sufficient to influence risky choice [24]. The effect of social observation on risk preference may be particularly salient in late adolescence [25], although the findings of a recent study suggests this particular developmental pattern is found only in more affective decision-contexts [24].
A different context is when adolescents are confronted with someone else's behavior. Recent research on social influence showed that when participants saw a previous participant select the risky option, their risk preference increased. A similar pattern, with risk preference decreasing, was observed for the condition in which participants saw a safe choice of a previous participant [31, 32, 33, 34, 35, 36]. Some studies suggest that risk-promoting peers potentially have a larger influence [35], albeit less so in late adolescents [36]. Others state that safe-promoting peers weigh more heavily [33], and particularly in late adolescents [36]. Paradigm-specific differences may lead to some of these mixed findings. For instance, we may be more easily swayed by others if their risk preference is not too far away from our own risk preference. Clearly, the extent and conditions for “state” changes in risk preferences need to be further disentangled.
It has been suggested others influence our risk preference by adding to the subjective value of the presented options [33] and altering reward-related brain activation [27,37]. Alternatively, an interesting perspective is that adolescents may be more uncertain about their own preferences than adults, and hence depend more on what others do [38]. This resonates with studies showing that uncertainty about others' behavior increases rapidly in adolescence [39]. A longitudinal follow-up in this study [38], suggested that such a heightened uncertainty related to a more positive development of peer relations across adolescence. Taken together, these findings suggest adolescent's social susceptibility may be an opportunity for promoting positive behavior using peer influence and can be potentially adaptive in terms of strengthening friendships and relations [38,39]. These hypotheses will need to be addressed in future studies.
Risks in the adolescent brain
Developmental neuroimaging work has studied the functional neural correlates of rewards and risks, to gain a better understanding of the neural correlates underlying adolescents’ risky decision-making (see Figure 1). Typically, these studies link an individual-difference measure, such as self-reports or indices of daily-life risk taking, to neural activation during a risky decision-making task. A large body of work shows that reward anticipation and receipt increased neural activity in the ventral striatum and the medial prefrontal cortex (PFC), which are projection targets of midbrain dopamine neurons [40,41]. These findings raise the question of whether this same circuit might process risk preference [42].
In adults, past functional neuroimaging studies have identified multiple brain regions associated with making decisions involving risk [43,44]: The (posterior) parietal cortex is thought to code probability [45] and its gray matter density has been related to individual differences in risk preference [46,47]. The ventral striatum is thought to reflect the potential of rewards for a giving option and may promote an approach-tendency towards risk. The (ventral) medial PFC integrates the (subjective) magnitude and probability of rewards for a risky option and controls the tendency to avoid or approach risks. For instance, functional connectivity of ventral medial PFC and action-related brain regions has shown to be higher in individuals with higher risk preferences [48]. Finally, activity in the amygdala and anterior insula would reflect the degree of risk and may promote risk-avoidance [43,45,49]. A recent meta-analysis supports these findings and highlights the insula as a crucial component of risk anticipation and processing, involved in coding known and unknown risks [44].
This key role of the insula is also confirmed in developmental studies of risk processing [50, 51∗, 52]. A study tracking parametric changes in risk in a risky-choice paradigm showed that adolescents versus children and adults exhibited heightened anterior insular risk activity to options of greater risk [53]. A larger developmental study (N = 256) confirmed that anterior insula activation scaled with parametric changes in risk [54]. Thus, risk-related brain regions typically found in adult studies, play a key role in developmental populations as well. Several recent studies showed that risk-related insula activation interacts with PFC regions and shuns adolescents from health-risk behaviors [55, 56, 57]. These findings indicate that risk engages brain regions that arewell-connected to brain regions associated with reward processing and valuation.
Opportunities for an individual difference perspective in adolescent risk processing
Several recent perspectives call for the need to characterizing individual differences in neurobiological development and cognition [58]. Particularly, the mixed findings of age-related changes in adolescent risk preferences may indicate these preferences differ profoundly between individuals [58,59]. One way to further understand developmental and individual differences in risk preferences is to use a model-based approach that formalizes the decision process. Where utility-based models present an integrated framework of value-based decision making, one of the few models disentangling the influence of risk and returns on risky decision-making is a risk-return framework. Risk-return models describe risky decision-making as a function of three variables: (1) the perceived return of available choice options (i.e., subjective expected value), (2) the perceived riskiness of those options (i.e., subjective outcome variability), and (3) the decision maker's attitude toward perceived risk—that is, his or her willingness to trade perceived risk for possible return [8,60]. One interesting suggestion from this body of work is that differences in perceptions are more influential than individual's willingness to take risk [60, 61, 62]. For instance, research on the COVID-19 pandemic showed that adults' risk perception (i.e., how risks are perceived) and risk preference (how likely one is to take risks) correlate positively to mitigation behaviors concerning social distancing and hygiene [63]. Risk perceptions were, however, correlated with a larger number of mitigation practices. Since risk perceptions may also be more malleable than risk preferences, they are potentially an interesting target for adolescent policy and behavioral interventions.
An exciting approach to understand and quantify individual differences in risky decision-making is by examining heterogeneity within groups. Latent clustering approaches aim to find clusters of individuals that are described on the basis of a risk profile, i.e. a type of persons with a similar configuration of (multidimensional) risk preferences. A large-scale population study (N = 3123) used such a multidimensional cluster approach on self-reported risk data [64]. Findings showed four risk profiles that accounted for two thirds of participant (i.e., the “cautious,” the “recreational adventurers,” the “financial gamblers,” and the “daredevils”). These approaches were shown to be associated with socio-demographic indicators, including age and gender. These clustering techniques have the potential to be extended to neuroimaging data (e.g. the studies by Becht et al., van Duijvenvoorde et al. [65,66] and a developmental longitudinal (i.e., latent change) perspective [67]. Ultimately, such analysis technics may foster greater specificity for certain groups of adolescents regarding behavioral interventions.
Conclusion
Although reward-sensitivity is integral to adolescent risky decision-making, we argue that a focus on risk broadens the perspective on adolescent behavior (see Figure 2). Here, we discuss recent findings on the measurement of risk preferences, the (social) context-specific influences on adolescents' risk preferences, and the associated brain regions involved in risk processing. A promising avenue that emerges from the work in this review is the potential to engage the social sensitivity of adolescents to steer decision-making, and the potential to embed models that target individual differences in multidimensional risk profiles. These renewed directions suggest that a perspective on risk sensitivity, in addition to adolescent reward-sensitivity, provide key insights to leverage undesirable levels of risk taking in young people and improve adolescents’ health and wellbeing.