COVID-19 stressor reduces risk taking: the role of trait interoception

Daily life situations require decision-making processes. Despite the influence of the rational choice theory (Becker 1968; Bernoulli 1954; Deming, von Neuman and Morgenstern, 1945)—which argues that individuals select the best behaviour in order to maximize their utility regardless of the context—the Prospect Theory proposed by Kahneman and Tversky (1979) demonstrated that changes in the context could lead to a disruption in the risky choice consistency. This disruption can be induced by the Framing Effect (FE), a widely studied cognitive bias that suggest that the way different alternatives are displayed affects people’s inclination to make a specific choice. For example, several studies have shown a reversed risk preference when presenting the same option in different frames (positive vs. negative frames). In addition, it has been reported a risk seeking when the choice involves losses and a risk aversion in choices involving gains (Detweiler et al. 1999; Kühberger et al. 1999; Pinon & Gambara 2005; Tversky & Kahneman 1981). This frame-dependent inconsistency has been observed in hypothetical scenarios involving human lives (the “Asian Disease Problem”; see Tversky & Kahneman 1981) and also in economical risky choice tasks (de Martino et al. 2006).

Together with the FE, stress seems to be an important modulator of decision-making strategies. In fact, many real-life decisions are made under stress, and the decision itself could be a source of pressure. Previous research has shown a link between stress and a performance decrease in decision-making tasks (Starcke & Brand 2012; Wemm & Wulfert 2017). Furthermore, classical research found that higher arousal level leads to magnify the risk seeking on loss-domain choices and risk aversion on gain-domain choices (Porcelli & Delgado 2009; Ring 2015; Yamakawa et al. 2016). In other words, higher arousal increases FE. This could be due to the release of glucocorticoids and noradrenaline, which reduce intracellular signalling along prefrontal-amygdala circuits (Arnsten 2009; Yamakawa et al. 2016). In fact, under psychological stress conditions the amygdala activates hypothalamic and brainstem structures leading to an enhanced noradrenaline and dopamine inputs and impairing the prefrontal function. Since prefrontal cortex structures are involved in higher order decision-making and inhibition of impulsive behaviours, acute stress could lead individuals to perform a biased and non-reflexive strategies, especially in loss frames (Arnsten 2009). This line of evidence is in accordance with the salience-of-losses hypothesis (Metz et al. 2020), which argues that the activation of the salience network in stress conditions should increase the psychological impact of losses. Nonetheless, it is important to emphasise that the effects of acute stress on decision-making can be moderated by other emotional and psychophysiological variables, such as the trait interoception and alexithymia. Interoception—the ability to process and integrate inner body signals (Chen et al. 2021)—has recently become a new topic of investigation even though the Somatic Marker Hypothesis (Damasio 1996) was formulated few decades ago. Some evidence shows that interoceptive awareness could predict decision-making performance under risk (Salvato et al. 2019). In the case of alexithymia, it seems to reduce the FE by blocking the incorporation of emotional feedback into decision-making (Shah et al. 2016a, b). It is known that alexithymia and interoception have opposite directions in both emotional awareness and FE susceptibility (Laloyaux et al. 2015; Shah et al. 2016a, b; Sütterling et al. 2013). Moreover, a recent study by Manzoor et al. (2021) showed that interoception and alexithymia could interact and predict low rates of FE sensitivity.

It is important to notice that most of the research makes use of artificial elements or paradigms to elicit emotional responses rather than real-life-related ones. This can lead to a bias ignoring cognitive variables such as threat perception (Lazarus 1966; Lazarus & Cohen 1977; Lazarus & Folkman 1984). For example, Henze et al. (2017) found that the Trier Social Stress Test (TSST)-induced similar neuroendocrine responses to that of a real-life oral exam, but instead the naturalistic stimuli had a more pronounced psychological impact in the sample. In this context, COVID-19 pandemic has become a powerful stressor since its inception (Lee et al. 2020). Concretely, in Spain, almost 20% of the population suffered high levels of anxiety and severe depressive symptoms during lockdown, being young adults the most affected demographic cluster (Becerra-García 2020; Universidad Complutense de Madrid, 2020). In addition, governments, citizens and health officials had to make decisions where the lives of others were at stake, making the pandemic an opportunity to study the effects of real stressors on risky choice tendency.

Based on all the above, the present study aims to investigate how negative COVID-19-related emotions can modulate decision-making processes and how emotional and psychophysiological variables such as interoception and alexithymia can moderate this relationship. Using the FE paradigm and following the salience-of-losses paradigm, it is expected that the exposure to a COVID-19 emotional stressor would increase FE in a risky choice task. According to previous evidence, we also expected an interaction between interoception and alexithymia mediating the influence of perceived stress in FE scoring.

Table 1 shows descriptive statistics. Participants who exceeded ± 3 standard deviations were filtered out as outliers. Kolmogorov–Smirnov normality test was performed and variables that did not follow a normal distribution were converted using a log (y + 1) transform.

Control and experimental groups were equally distributed in age, gender and socio-economical variables. No differences were also found in their experience regarding COVID-19 pandemic.

Going further on their COVID-19 experience, most of the participants reported not having received a diagnosis nor experiencing compatible symptoms (87.7%), while none of those who received a diagnosis or experienced symptoms needed to be hospitalised. 61.3% of our sample reported having a close friend or familiar with COVID-19 diagnosis of which 32.3% needed health care assistance. Finally, 9.4% of our sample informed having a close friend or familiar who died because of the infection.

The results obtained in the present research shown that a COVID-19-related emotional stressor can modulate decision-making on risky choices. Classical theories about the effect of stress on FE argues that higher levels of stress could increase risk aversion in gain frames and risk tendency in loss frames. Our data do not support this. Instead, our results reject the salience-of-losses hypothesis and suggest a decreased risk seeking in both frames after being exposed to a stressor making individuals more conservative in any type of risky choices. In other words, stress and higher arousal may reduce loss aversion.

Two different neural systems are reported to be related with risk tendency and risk aversion—closely linked to loss aversion (Kahneman et al. 1991). The first one, the appetitive system, involves striatal structures and frontal regions such as the ventromedial, ventrolateral or dorsolateral prefrontal cortex (Sokol-Hessner et al. 2013; Tom et al. 2007). These structures are associated with reward seeking and could explain how individuals tend to make more conservative choices on gain frames. The second neural circuit, the aversive system, involves structures such as the amygdala or the posterior insular cortex (LeDoux 2012). In this sense, the aversive system has become a key feature providing the neural basis of loss aversion and explaining how classical research on FE found an increased risk tendency in loss frames under stress conditions. However, the influence of both systems does not seem to be as simple. It is known that the amygdala can also be activated by potential gains and that higher rewards or incomes can be processed as a potential loses (Molins & Serrano 2019). Moreover, we cannot define the role of these neural circuits only in terms of “appetitive” or “aversive” stimuli. Decision-making results from computing processes where a particular alternative is valuated regarding its cost–benefit subjective balance (Croxson et al. 2009). Thus, it should not be contradictory that different sources of stress affect decision-making differently. In fact, our results are in-line with recent previous evidence and provides further evidence on how emotional stress reduces risk seeking and loss aversion, being in accordance with a new body of evidence (Molins et al. 2021; Margittai et al. 2018). It is also important to notice that we used a real-life-related source of stress in our experiment. That is, our audio-visual stimulus is not a real-life stressor per se but it can potentially prime participant’s personal experiences regarding the pandemic. In this sense, in a COVID-19-related context a particular amount of loss may be processed as a potential gain compared to the risk of lose a higher amount of resources. Furthermore, the reduction in loss aversion supports the reward-alignment hypothesis (Metz et al. 2020), whereby the reduction in loss aversion could be induced by an activation of the reward system. The reward-alignment hypothesis has an empirical evidence since Mather and Lightball (2012) found an enhanced dopamine activity under stress conditions in reward-related central structures. Decision-making, ultimately, would be determined by the balance of the activation of these different neural circuits and the posterior cost–benefit computation in the prefrontal cortex. Both the nature of the stressor and the alignment hypothesis provide a coherent explanation for the results obtained in our study.

In addition, the observed relationship between interoception and decision-making is in accordance with the above. Linear regression showed that interoception is a significant predictor of bet acceptance in loss frames after being exposed to the stressor. The more interoception rate, the more bet acceptance in loss frames. Insular and cingulate cortices—also involved in the aversive system—are a central hubs for processing and integrating inner body signals (Chen et al. 2021). Consequently, we hypothesise that higher interoceptive awareness could break the loss aversion reduction in balancing the individual cost–benefit valuation towards the aversive system. Based on previous evidence (Manzoor et al. 2021), we expected higher rates of alexithymia to be related to less framing susceptibility. We have not found this effect, but subsequent studies should inquire in this concern.

Nevertheless, we cannot forget to mention a few limitations of our study. Firstly, our sample was small, considering test performed, and mainly composed by women (N = 85). There is evidence that risk taking may differ between men and women (Mather & Lightball 2012) and between different hormonal responses during acute stress (Alacreu-Crespo et al. 2019). In this regard, it might be interesting to replicate our results with a sex-balanced sample and controlling endocrine responses. Secondly, our control group did not behave as a real control group. Although our control group stimulus was previously used in prior research and did not produce changes in electrodermal activity and emotional stress (Molins et al. 2021), our sample reported a significant reduction in perceived stress levels with respect to the pretest. This should be considered as a limitation, but it is importance to notice that their change in bet acceptance had the opposite directionality to that observed in the experimental group. Furthermore, it is possible that some cognitive biases influenced the control group. Despite the limitation, the results observed are consistent with our purposes. Finally, we must mention some methodological issues. Due to COVID-19 restrictions and the high cumulative incidence, we had to opt for an online methodology to carry out our experiment. As described in Methods, we provided a step-by-step instructions to our participants trying to standardise as much as possible the procedure. We encouraged our sample to control their environment (e.g. controlling the lighting and reducing the presence of potential stressors) and choose a moment of the day when they felt most relaxed. However, we cannot not assure all procedural instructions were followed as we wanted. Thus, it would have been preferable to obtain complementary psychophysiological measures such as heart rate or electrodermal activity to strengthen our data, but it was not possible and would probably have been a source of stress given the sanitary situation. Further research should solve this.

Even so, our study provides substantial evidence of how emotional stress could affect decision-making under risk and puts the focus on how different sources of stressors may trigger different risky decisions. These results could be also relevant in the field of public health and clinical intervention. COVID-19 has become a drastic source of stress in general population over a long period of time. For its part, interoception seems to be an important modulating variable. The practical applications derived from this line of research can help general population as well as patients with impaired interoceptive processing—i.e. autism spectrum disorders—to make better decisions under emotional stress and uncertainty conditions.