Abstract
Despite recognition that emotions are present and salient during a crisis, traditional views of crisis decision making, such as crisis decision theory and naturalistic decision making, emphasize mainly the role of cognitive processes. Several recent crises illustrate individuals face complex, dynamic, and significant situations requiring decisions with which they are unfamiliar and/or lack experience. Moreover, dangerous and life-threatening situations activate negative emotions such as anger, regret, guilt, fear, disappointment, and shame, which may uniquely affect recursive associations with the immediate cognitive schema elicited after a crisis. Also consider individuals do not experience crises in a vacuum. Rather, they perceive, interpret, and assess information via interactions with others, thus creating collective crisis decision making as a substantive level of analysis. As such, we present a multilevel theoretical model examining the interactive role cognitions and emotions play in crisis decision making, and offer implications regarding individual and collective decisions during crises.
On March 11, 2011, the Great East Japan Earthquake devastated the Fukushima Daiichi Nuclear Power Plant with seismic tremors that resulted in total loss of off-site electricity. A back-up power line to Reactor 1 immediately failed due to mismatched sockets, and the ensuing tsunami just an hour later flooded and destroyed emergency generators, seawater cooling pumps, the wiring systems, and power supplies to Reactors 1, 2, 3, and 4. The tsunami also washed away critical vehicles and heavy machinery and destroyed buildings. Over the next tense few days, Reactors 1, 2, and 3 sustained reactor core exposure and damage, and hydrogen explosions occurred in Reactors 1, 2, and 4. On March 15, 2011 a mass discharge of radioactive material occurred from Reactor 2 (The Fukushima Nuclear Accident Independent Investigation Commission [NAIIC], 2012).
Results from the Japanese government’s investigative commission report (NAIIC, 2012) were compiled from 900 hours of hearings, survey responses from employees of the power plant (2,415), and local residents forced to evacuate (10,633), as well as responses from over 1,167 interviews of national officials, regional officials, and officials of the Tokyo Electric Power Company (TEPCO), and local plant employees and subcontracted operators. Additionally, to assure a maximum degree of information disclosure and investigatory reach, social media was used to obtain more than 170,000 comments. The information contained in the report highlights how the crisis moved from individual employees within the plant seeking answers and managing their own emotions, to employees acting as a collective to make decisions in the absence of leadership, to the public being impacted by the unfolding disaster within the organization. The report concluded that while the tsunami was a natural disaster, the unfolding events following the tsunami were a “made in Japan” disaster through a failure of leadership.
Unfortunately, the Fukushima Daiichi nuclear accident is not the only crisis in recent memory that involved individual and collective decision making, from both victims and leaders, in the face of significant emotions. The Haiti earthquake, the H1N1 epidemic, the 2008 financial collapse, Hurricanes Katrina and Sandy, and the 2013 Oklahoma tornados illustrate that individuals and leaders face complex, dynamic, and significant situations requiring decisions that may depart from traditional rationality-based decision models. Growing evidence in diverse disciplines such as psychology (e.g., Lerner & Keltner, 2000), neurobiology (e.g., Damasio, 1994; Sanfey, Rilling, Aronson, Nystrom, & Cohen, 2003), and behavioral finance and economics (Lo, Repin, & Steenbarger, 2005; Loewenstein, 2000; Loewenstein, Weber, Hsee, & Welch, 2001) indicates emotions play a critical role in decision making. These emotion effects are more pronounced in crises wherein a high level of ambiguity and uncertainty renders traditional cognitive decision making heuristics less useful (Duhackek, 2005).
Rather than a purely cognition-focused approach, recent research suggests cognitive information-processing models and emotion information-processing models work in tandem to create decision outcomes (Forgas, 2008; Kőszegi, 2006; Nofsinger, 2005; Sayegh, Anthony, & Perrewé, 2004). Advances in neuropsychology regarding the pervasiveness of emotions in human behavior and interaction (Dolan, 2000; LeDoux, 2000) also support this assertion, indicating that the complex interplay between cognitions and emotions can no longer be ignored in understanding information processing and decision making during a crisis. This point is particularly salient as predominant models in crisis decision theory—both for those experiencing crisis (crisis decision theory; see Sweeny, 2008) and those leading during crisis (naturalistic decision making; see G. A. Klein, 1993; G. A. Klein, Orasanu, Calderwood, & Zsambok, 1993; Lipshitz, Klein, & Carroll, 2006; Lipshitz, Klein, Orasanu, & Salas, 2001)—do not include a specific link to affect and/or emotions, nor do they account for multilevel issues in decision making as a crisis increases in scope. As evidenced from the aforementioned, both emotions and multilevel issues are relevant and important considerations in understanding the complexity involved in decision making during crisis, and therefore crisis decision models should better reflect the potential influences and implications of affect and/or emotions on decisions at multiple levels of analysis during an unfolding and escalating crisis.
The present work provides two key advancements to traditional crisis decision making literatures and models. First, deriving from a risk-as-feelings framework, the role of emotions, particularly negative emotions, in individual-level crisis decision making will be explored within the context of two key predominant cognitive crisis decision models: Sweeny’s (2008) crisis decision theory model that applies to individuals experiencing crisis, and naturalistic decision making (G. A. Klein, 1993; G. A. Klein et al., 1993; Lipshitz et al., 2006; Lipshitz et al., 2001), which specifically addresses expert decision making within crisis situations. Sweeny’s model derived from a comprehensive review of existing crisis decision making research, and follows a rational-based cognitive approach to decision making for an individual experiencing crisis. Naturalistic decision making models focus on the unique challenges experts and/or leaders face in crisis decision making; and while also relying on a cognition framework, a heuristics-based approach to decision making is used to explain how experts produce decisions based on personal experience and expertise. And, while neither model includes emotions within the decision process, neither model excludes the potential of emotions within the process. Because people experiencing a crisis, including leaders, experience emotions (Burns, Peters, & Slovic, 2012), the role of emotions within these cognitive decision processes needs to be examined.
The second contribution of the current work is to look beyond the individual affective and cognitive experience during crisis to consider how crisis decision making may evolve to include groups and collectives as the crisis escalates. Current crisis decision models are generally focused on individual-level experiences, and yet, as illustrated by the Fukushima Daiichi disaster, crises have the potential to involve hundreds, thousands, and potentially millions of people. Thus, it is important and timely to understand the multilevel implications of complicated affective and cognitive decision processes on both victims and leaders.
We adapt an organizational view of crisis (Pearson & Clair, 1998; Sayegh et al., 2004) where crisis is defined as an unusual event, which may have unknown causes and effects, with high ambiguity and low probability of occurrence. In addition, a crisis poses a serious threat to individual and organizational viability and as such, requires rapid response. Pearson and Mitroff (1993) outlined five phases of a crisis life cycle: signal detection, preparation/prevention, containment/damage control, business recovery, and learning. Arguably, while decisions are made within each phase, decisions made within the containment/damage control phase represent some of the most chaotic, time-pressured, and critical decisions of the entire crisis life cycle. Thus, our model predominantly focuses on the containment/damage control phase due to the relevance of emotions and cognitions and the potential multilevel interactions, but we acknowledge that emotions and cognitions across multiple levels can, and likely do, play a role in all phases of a crisis.
Individual-level crisis decision making
Crisis decision theory
Sweeny’s (2008) review and organization of crisis decision theory (CDT) highlights a predominantly rational approach that combines coping theory to highlight potential response types from people experiencing crisis with decision making theory to allow for predictions related to specific choices made by a person in a specific crisis situation. Sweeny (2008) illustrates a three-stage process that a person experiencing crisis moves through on the way to a decision.
Stage 1: Assessment of severity/information about causes
Sweeny (2008) asserts a person uses three categories of information in the process of assessing the severity of the negative event: information about causes, comparative information, and information about consequences. An individual’s perception regarding who or what caused the crisis is a critical factor in assessing severity of the event (Lazarus, 1991; Roesch & Weiner, 2001). These individual perceptions are critical as they directly affect both severity assessments and response choices.
Stage 2: Determining response options
Once the severity of a crisis event is assessed by an individual, two factors limit the response options generated and/or considered in this phase: controllability of outcomes and feasibility of responses (Sweeny, 2008). An individual’s perception of control of negative events determines the availability of certain responses and options, while resources factor into the feasibility of responses. Active responses are more likely under controllability perceptions, with options that are feasible related to available resources. This stage within CDT draws from some aspects of secondary appraisals within traditional coping theories (Sweeny, 2008), where an individual evaluates present resources to determine options for coping (Folkman, Lazarus, Dunkel-Schetter, DeLongis, & Gruen, 1986; Lazarus & Folkman, 1984).
Stage 3: Evaluating response options
While CDT acknowledges that determining response options (Stage 2) can be coupled with evaluation of response options, Sweeny (2008) highlights that the phase also can stand alone as a separate stage. As such, considerations that are taken into account when evaluating response options include resources necessary to execute a response, and both the direct and indirect consequences of a particular response. For clarification, feasibility (from Stage 2) considers the possibility of a response, whereas information about required resources considers the desirability of a response; and, direct consequences consider the depth of a response’s impact on the crisis, while indirect consequences consider the breadth of a response’s impact. Direct consequences consider efficacy of response in regard to improving the crisis, with consideration of the magnitude of the improvement as well as the reversibility of consequences. Indirect considerations include potential emotional impact, public image of self-interested individual action, as well as consequences for other areas of a person’s life and consequences for others. However, given emotions and cognitions may have reciprocal influence, it may be relevant to consider how emotions influence the determining of response options in CDT.
Naturalistic decision making
Understanding how negative emotions may arise in crisis decision making may offer a means for better managing the decision process to successfully navigate the dangers involved in specific decisions. This management process may fall directly on experts and/or leaders who are attempting to mitigate damage from crisis, be it loss of life and/or resources. Interestingly, crisis decision theory specifically regarding experts is somewhat limited, with the exception of naturalistic decision making theory (NDM). NDM (Lipshitz et al., 2001) is a cognitive-only decision theory developed to explain how experts make more effective decisions than those made by nonexperienced individuals involved in a crisis. Unlike CDT, however, NDM does not suggest experts predominantly rely on rational decision models, but rather heuristic decision models designed to tap their prior experience and honed instincts.
NDM highlights how experts (i.e., authorities or leaders) move through a decision process in real-world contexts that are familiar and meaningful to the decision maker, and often with ill-defined goals, time pressure, high personal stakes, and other complexities (Lipshitz et al., 2001). NDM does not rely on normative choice models as a starting point but rather incorporates the essential characteristics of proficient decision makers, situation–action matching decision rules, context-bound information modeling, process orientation, and empirical-based prescription (Lipshitz et al., 2001).
Process orientation
The process orientation of NDM contrasts with a more classic input–output orientation in that NDM attempts to describe a cognitive process followed by proficient decision makers rather than predict a decision option. The cognitive process highlights what information experts/proficient decision makers seek during a decision process, and how interpretation and decision rules are applied (Lipshitz et al., 2001). This process orientation reveals a situation–action matching decision rule feature.
Situation–action matching decision rules
Within the situation–action matching process, options are selected or rejected based on compatibility with the specific situation (Endsley, 1997; G. A. Klein, 1998; Pennington & Hastie, 1993) and the values of the decision maker (Beach, 1990) rather than on the merits relative to the options (Lipshitz et al., 2001). Moreover, while there is an analytic component, NDM often relies on pattern matching and informal reasoning (Cohen, Freeman, & Wolf, 1996; G. A. Klein, 1998; Lipshitz, 1993). Given that emotions and cognitions move in tandem, emotions may play a role in the informal reasoning.
Because proficient decision making is driven by expert- or experience-based knowledge, NDM depicts a situation–action matching response process resulting in the selection of a response due to the appropriateness of that response for the specific situation. Key to the situation–acting matching process is that options are evaluated sequentially, one at a time, in comparison to a standard the expert decision maker holds. Amidst several options, experts quickly screen many of them and compare them to a standard, rather than one another. This narrows the option field to only one or two options which can then be compared (Beach, 1990; Montgomery, 1988). This matching process differs from a concurrent choice model more typical of classical decision theory, where a response selected after an evaluation of all alternatives indicates the response is superior to any other alternative.
Context-bound informal modeling
Experience-tied knowledge is key for expert decision makers, and NDM depicts to what information experts actually attend and use in the process. Specifically, expert knowledge is domain- and content-specific (Ericsson & Lehmann, 1996; Smith, 1997) and semantic and syntactic content are important to expert decision makers (Searle, 1995; Wagenaar, Keren, & Lichtenstein, 1988). This modeling drives the process toward empirical-based prescriptions.
Empirical-based prescription
The goal of a decision prescription is focused on improving feasible modes of making decisions, with an emphasis on “feasible.” For example, solutions that highlight an optimized proof that is not likely to be implemented should not be considered optimal. Rather, the descriptive validity of a solution must focus on what is likely to be implemented and performed by a recipient during a decision process, not what “should” be implemented as a result of an optimization process outcome. This places the emphasis squarely on the demonstration of feasible expert performance, which in concert with the three other key NDM characteristics should provide the expert with sound heuristics on which to build decisions.
However, absent in the NDM process is the incorporation of emotions into the decision process, despite convincing evidence that emotions are critical within decision making under risky conditions (Slovic & Peters, 2006). Similar to conclusions regarding CDT, because of the reciprocal influence of emotions and cognitions, understanding how emotions may factor into the NDM process could provide a more complete picture of how experts arrive at decisions, and assist in enhancing the efficacy of the prescriptive decision process.
The risk-as-feelings hypothesis provides a pathway for assimilating emotions, specifically negative emotions, into cognitive-only-based decision processes. Prior to Loewenstein and colleagues’ risk-as-feelings hypothesis (Loewenstein et al., 2001), theories of choice under risk were predominantly cognitive and consequentialist. Anticipatory emotions (i.e., immediate and visceral emotional reactions to risks) supplemented the consequentialist perspective where cognitions predicted what future emotions may be experienced, which accordingly reflected in how potential outcomes were weighed, and in doing so, introduced the notion that feelings such as dread and anxiety may directly and reciprocally impact the cognitive evaluation process leading to behavior, as well as directly impact behavior itself (Loewenstein et al., 2001).
Risk-as-feelings
Developed as the risk-as-feelings hypothesis (Loewenstein et al., 2001), risk-as-feelings refers to an instinctive and intuitive reaction to danger (Slovic & Peters, 2006). Slovic and colleagues (Slovic, Finucane, Peters, & MacGregor, 2004; Slovic & Peters, 2006) characterize reliance on risk-as-feelings as “the affect heuristic,” where affect refers to the hedonic tone of a stimulus as either pleasurable or painful and experienced as a feeling state (with or without active cognitive awareness). As noted by Schwarz and Clore (1983), experienced affect is used as information in the decision making and judgment process regarding risks and benefits, and therefore is particularly relevant to both risky decisions and decisions in situations such as a crisis that may involve considerable risk. In other words, individuals judge risk on how they feel about the risk, in addition to what they think about the risk.
The affect heuristic may be particularly important during times of crisis, as reliance on affect is generally quicker, easier, and more efficient in addressing complexity and uncertainty (Slovic & Peters, 2006), and complexity and uncertainty are certainly key elements of crises. For example, Finucane, Alhakami, Slovic, and Johnson (2000) found the prototypical inverse relationship between risks and benefits greatly increased under time pressure, as analytic, deliberative opportunities were reduced. Given the significant time constraints found in crises such as earthquakes, tsunamis, tornados, and floods, it is critical to understand the affective component that pushes individuals towards a quick, easy, and efficient decision process. And, while positive affect is part of affect-as-information theory, prior research indicating the likelihood of negative affect and perceived risk to spike during crisis onset (Burns et al., 2012) helps narrow the focus of the current work to the more common response of negative affect and emotion, as we specifically focus on this in the initial onset of a crisis.
While early views of crisis response from publics predicted panic, research tends to indicate that while panic can happen, this type of extreme response is rare during disasters (Aguirre, 2005; Mawson, 2005). Rather, in situations of perceived severe physical danger (i.e., fear), a person may act in ways counterproductive to survival: seek others before taking action during a fire or evacuation, deny the crisis and delay critical exit/evacuation options, and/or move toward danger irrationally such as going outside during an earthquake (Aguirre, 2005; Mawson, 2005; Reifels et al., 2013). Experiencing anger during crisis may lead to aggressive behaviors toward trained first responders and refusals of assistance (Cohn, Carley, Harrald, & Wallace, 2000). While fear is generally associated with a flight-or-fight protective response (Aguirre, 2005; Mawson, 2005), a variety of negative emotions may be related to dysfunctional or nonhelpful outcomes and therefore warrant attention related to crises.
Negative emotions within crises
Because of the more immediate threat involved in the containment/damage control phase of a crisis, we chose to focus on specific negative emotions that have been linked to prior crisis-based (Lazarus, 1991) and/or decision research (de Hooge, Zeelenberg, & Breugelmans, 2007; Zeelenberg, van Dijk, Manstead, & van der Pligt, 2000): fear, shame, disappointment, anger, regret, and guilt. However, recent research has indicated that solely viewing these emotions within a generalized negative valence may not be appropriate (Angie, Connelly, Waples, & Kligyte, 2011), as they may contribute to differing responses, especially in regard to decision options. For example, anger may produce risk-seeking choices due to assessments of high certainty and control, whereas fear may produce risk-avoidant choices due to assessments of uncertainty and lack of control (Lerner, Gonzalez, Small, & Fischhoff, 2003; Lerner & Keltner, 2000, 2001). Similarly, regret and disappointment have been shown to have differing behavioral consequences, where regret often results in helping action but disappointment may result in withdrawal (Zeelenberg et al., 2000). This appears similar to differences between shame and guilt as well, where the action tendency for guilt is to make up for wrongdoing, but inactivity and powerlessness may be more likely when experiencing shame (de Hooge et al., 2007).
Emotions can be differentiated in terms of action tendencies and action, among other factors such as feelings, thoughts, and appraisals (Frijda, 1986). Within this approach, research has noted associations between emotions and specific patterns of appraisals (Kuppens, van Mechelen, Smits, & De Boeck, 2003), and made the case for action tendencies to be considered as central to emotional components as these too help characterize emotions (Frijda, 1986; Lazarus, 1991). As such, we take a componential approach to examining these six negative emotions related to crisis, specifically through exploring a control appraisal–action tendency pattern among these specific emotions. Previous research has employed an appraisal of control along with an examination of action tendency for exploration of the appraisal basis of anger (Kuppens et al., 2003). We employ a similar approach by examining the control appraisal–action tendency similarities of the negative emotions of anger, regret, guilt, fear, shame, and disappointment and explore how those similarities may impact cognitive decision models.
Control appraisal
Among the six negative emotions considered, anger, regret, and guilt indicate some similarity in regard to control. Specifically, the appraisal of control has been linked to anger in prior research (Lerner & Keltner, 2001; Scherer, 1993), indicating a sense of power over what is happening. And while research on guilt may not discuss “control” per se, the actor acknowledging that his/her actions have caused problems (Lewis, 1971) indicates that control and/or feelings of responsibility may center with the actor. Similarly for regret, intense feelings that the actor should have known better (Zeelenberg et al., 2000) may be related to research showing regret has high scores on control potential (van Dijk, van der Pligt, & Zeelenberg, 1998), indicating control and/or feelings of responsibility are somewhat centered within the actor.
Conversely, fear and guilt do not seem to share an appraisal indicating some regard for control, but rather the opposite, where appraisals of powerlessness may prevail. Fear may produce appraisals of uncertainty and lack of control (Lerner et al., 2003; Lerner & Keltner, 2000, 2001), and disappointment may be similar to fear in that feelings of uncertainty are associated with this emotion (Zeelenberg et al., 2000). Similarly, shame corresponds to a global and painful judgment of both specific behaviors and general self in a negative light, resulting in feelings of powerlessness (Tangney, Miller, Flicker, & Barlow, 1996). Additionally, due to shame being on others’ opinions of the self, helplessness and dependency on others’ reactions have been reported (Ferguson, Stegge, & Damhuis, 1991), which also may hamper an individual’s control of the situation. Table 1 summarizes the control appraisals of the six negative emotions.
Comparison of control appraisals and action tendencies among crisis-based negative emotions.
Activation level and action tendency
Among the six negative emotions considered, anger, regret, and guilt indicate some similarity in regard to action tendency as well as control. As previously noted, in some form anger, regret, and guilt result in action tendencies that are more active in nature. Lerner and colleagues (Lerner et al., 2003; Lerner & Keltner, 2000, 2001) note anger may produce risk-seeking behavior as a tendency, which can be considered an active response. Research on guilt indicates individuals will have a tendency to undertake action to make up for wrongdoing (de Hooge et al., 2007; Tangney et al., 1996), which represents an action response. Finally, research related to regret notes a tendency to correct one’s mistake, undo the event, and get a second chance, which were noted as active attempts (Zeelenberg et al., 2000); although recent research noted regret may be less connected to aversive arousal than guilt due to its intrapersonal focus (Imhoff, Bilewicz, & Erb, 2012).
Conversely, fear—which differed in control appraisal from anger, regret, and guilt (i.e., control)—may have an action tendency more similar to these emotions in that fear produces urgent action tendencies. However, while anger, regret, and guilt produced a more engaged form of action tendency such as risk-seeking, corrective, and/or second-chance tendencies, fear produces an urgent action tendency. Interestingly, research on flight during intense fear generally does not reveal productive avoidance (i.e., flight to safety), but rather behavior seeking proximity of familiar people and places, even if that means remaining in danger (Mawson, 2005). As such, while still active, the nature of the engagement differs for fear in that behavior may not be constructive, but rather affiliative- and approach-oriented towards loved ones.
Similarly, disappointment may have an action tendency that favors getting away from a situation (Zeelenberg, van Dijk, Manstead, & der Pligt, 1998), which is another withdrawal-type action. Additionally, shame is similar in that there also may be a tendency to do nothing (Zeelenberg et al., 2000), where inertia may result of thinking taking any action will make no difference (Seligman, 1975). Table 1 also summarizes the activation level and action tendencies of the six negative emotions.
While this particular grouping of negative emotions is unique to crisis decision making, the notion of a componential approach to relations among emotions is not unique in emotions research (Frijda, Kuipers, & ter Schure, 1989). Within a componential approach, emotions are structures of both appraisals and action tendencies. Moreover, activation level and action tendencies are needed to fully account for the structure of emotional behavior (Frijda et al., 1989). In research regarding emotional names applied to emotional experiences, Frijda et al. (1989) indicated that cues within components may overlap, but also add some independent contribution to prediction. For example, when control appraisal was of primary interest, guilt and shame were more likely to cluster together in appraisal profiles. However, in action readiness profiles, guilt and shame did not cluster together (Frijda et al., 1989).
Prior research (Lerner & Keltner, 2001; Scherer, 1993; Zeelenberg et al., 2000) on the specific negative emotions of interest and their likely control appraisals guided classification as either related to “control” or “no control.” Likewise, prior research (de Hooge et al., 2007; Lerner & Keltner, 2001; Mawson, 2005; Scherer, 1993; Tangney et al., 1996; Zeelenberg et al., 2000) on activation level and action tendencies for the six negative emotions of interest guided classification as either “high” or “low” activation, and either “engagement,” “affiliation,” or “withdrawal” tendencies. Control appraisals, activation levels, and action tendencies indicate anger, regret, and guilt may share more characteristics than fear, disappointment, and shame. And, while fear has a high activation level, the lack of control appraisal and questionable productive action tendency due to affiliation actions seem to position this negative emotion somewhere between anger, regret, and guilt on one hand, and disappointment and shame on the other.
Predominantly, the control appraisal and subsequent action tendencies produce three differing categories of negative emotions: (a) control-based and active consisting of anger, regret, and guilt; (b) lack of control and affiliation consisting of fear; and (c) lack of control and avoidant consisting of disappointment and shame. Using all three components likely provides a richer descriptive of how and why negative emotions may differentially relate to a crisis decision process. Because negative emotions are prevalent in crises, do these three groupings of negative emotions interact differently with cognitive-based crisis decision models? We explore the implications of this question on two key individual-level crisis decision models that do not currently include emotions as a model component.
Crisis decision theory and negative emotion
The risk-as-feelings framework has postulated that decision making responses to risky situations are in part influenced by cognitive evaluations that have affective consequences, where feeling states also exert reciprocal influence on the cognitive evaluations (Loewenstein et al., 2001). However, prior research including the risk-as-feelings hypothesis also asserts that feeling states respond to factors that do not involve cognitive evaluation of risks, suggesting that responses to risky situations and decisions result from direct emotional influences (Clore, 1992; Clore, Schwarz, & Conway, 1994; Zajonc, 1980). As such, determining response options within CDT should be modified to reflect that an individual experiencing crisis has the potential to determine a response to the risky crisis situation based on the way the person feels about the crisis. Moreover, because the determinants of emotions differ from cognitive determinants, emotional reactions to risk may diverge from cognitive reactions to the same risk (Loewenstein et al., 2001).
While certain emotions may be critical for producing cautious and risk-averse decisions, the degree of emotional response to the stressors of crisis needs to be appropriately moderated to produce effective decisions. Since perceived risk and negative emotions often escalate at the beginning of a crisis (Burns et al., 2012), the strong visceral feelings generated that circumvent cognitive processing are likely related to negative affect. As suggested earlier, negative emotions may represent three different types of response, depending on the control appraisal and action tendency associated with the emotion. For example, anger, regret, and guilt reflect an appraisal of control and an action tendency and, as such, the response option selected may reflect more of an engagement response. One example of this type of response to the Japan disaster was quoted in the NAIIC (2012, p. 58) report: “The attitude and responses of TEPCO and the government, who seem to think so little of us, make me angry rather than sad. I demand quicker and more sincere response.” Clearly, anger on the part of a local Japanese resident leads to demands on the government, which is a form of active engagement on the part of this victim. Moreover, the NAIIC (2012) report classified 8,047 free comments from victims/residents into 63 common topics, and nearly one third of the topics included a “demand” made by residents to the government for issues such as compensation, further investigation, and decontamination.
Conversely, disappointment and shame reflect an appraisal of lack of control and withdrawal tendency and, as such, the response option selected may reflect more of an avoidant response. Finally, fear reflects an active appraisal but with no control, which may result in affiliative responses that may include waiting to take action with a primary social group. Examples of feeling no control from the Japanese official report also are evident in comments made by hundreds of victims, for example: “I have no idea how one is supposed to live like this” (NAIIC, 2012, p. 60) and “feeling of being bogged down day-to-day, no joy, no hope” (NAIIC, 2012, p. 61). While CDT has outlined how an individual moves through a cognitive-only decision process, we assert negative affect may play an important role within CDT processes that influence decisions/responses. As such we propose:
Naturalistic decision making and negative emotion
Negative emotions in context-bound information processing
The importance of having active and engaged leaders during a crisis cannot be understated. Because proficient decision making is driven by expert- or experience-based knowledge, NDM depicts what information decision makers attend to and which arguments are employed (Cohen & Freeman, 1997; Crandall & Getchell-Reiter, 1993; Lipshitz, 1993). To the extent that the expert can see patterns and relationships in a current crisis that may have some relevance to stored schema, the perception of the richness of context may be impacted by the expert’s ability to process meaningful information.
This processing of meaningful information makes the risk-as-information framework particularly relevant for NDM. First, the focus on decisions under risk in the risk-as-feelings hypothesis mirrors the types of decisions often highlighted in NDM or, in other words, complex and risk-laden decisions. Second, unlike classical decision models, the risk-as-feelings hypothesis focuses on emotions experienced during the decision process (i.e., anticipatory emotions), rather than the more traditional anticipated emotions related to particular outcomes and alternatives, which aligns with NDM’s focus on decision processes, rather than an outcome focus.
Due to limited time during an active crisis, NDM requires rapid engagement with regard to information processing. A control appraisal, as opposed to a lack of control appraisal, may be critical for leaders and experts in that feelings of control may facilitate engagement in a solution process more rapidly than when one experiences feelings of little control. This is particularly relevant for negative emotions in that not all emotions lead to feelings of control. The distinction between which negative emotions leaders experience may have important implications related to NDM as to how rapidly contextual information is attended to and assessed by leaders. As noted in interviews with experts from Japan (NAIIC, 2012), because negative emotions such as anger (e.g., “it is outrageous that TEPCO claims the radiation released from its power plant is…an ownerless object for which they cannot be held accountable”; p. 73); regret (e.g., “Chairman, Nuclear Safety Commission, admitted that the safety guidelines were defective and expressed his apology”; p. 73); guilt (e.g., “we realize fully that the belated or indefinite evacuation instructions…affected residents severely”; p. 77); and fear (e.g., “I felt endangered”; p. 68) elicit control appraisals, this may enable the necessary rapid contextual-bound information processing found within NDM. Conversely, leaders experiencing negative emotions such as disappointment with TEPCO (e.g., “I thought I had been abandoned”; p. 68) and shame (e.g., “I do not want the executives to be so dismissive of this accident”; p. 66) may have felt a lack of control surrounding the situation, therefore seriously limiting their ability to engage in complex information-processing functions.
Negative emotions in situation–action matching
This screening and narrowing process is an evaluation of a situation matched to an action, which, as suggested, may be influenced by emotions and, in particular, the action tendency of emotions. For example, negative emotions may represent at least three different types of response options in part related to an action tendency associated with the emotion. As noted previously, the action tendency of anger, regret, and guilt may link to an engagement response, while disappointment and shame may link to a withdrawal tendency. Finally, fear is associated with an active affiliative tendency, meaning an individual’s action may be predominantly focused on finding close others before escape (Mawson, 2005). Therefore, we assert:
Individual-level crisis decision making summary
Figure 1 highlights the unique relationship between individual cognitions and negative emotions, and their proposed outcome-related action tendency, activation level, and control appraisals. Cognitions, heuristics, and negative emotions interplay differentially with one another to result in differential decision response outcomes, for both victims and leaders/experts, at the individual level in a crisis.
Individual-level crisis decisions with negative emotions.
Leaders and experts use heuristics to quickly process information and narrow limited decision options quickly. To the extent a leader is experiencing anger, regret, or guilt, the information process may proceed rapidly, which can be critical during a crisis. However, a leader experiencing disappointment and/or shame may not be able to engage in rapid information processing, which could cost lives during a crisis. Likewise, anger, regret, and guilt may lead to action-oriented engaged behaviors on the part of a leader, while disappointment and shame may result in avoidant behavior. As noted in the official report of the Japanese disaster, “at the time of the accident, neither the Chairman nor the President of TEPCO were present or accessible, an inconceivable situation for an operator of nuclear power plants” (NAIIC, 2012, p. 33).
To the extent victims look to leaders for guidance during crises, withdrawal behaviors may leave victims with no guiding presence. A leader experiencing fear may seek to affiliate, which given the precious nature of time during a crisis may not be a productive behavior either. The point is that even leaders and experts can experience negative emotions during a crisis, and to the extent that behaviors can be anticipated based on those negative emotions, more contingencies can be understood and incorporated into the decision process. Understanding that a fearful leader may seek primary social contacts can help leaders train differently for crises and understand how to proceed given a variety of emotions.
Both CDT and NDM models of crisis decision making primarily focus on individual decision making (either victim or expert); and we assert that negative emotion plays a critical role in understanding potential actions taken by both victims and experts. However, since negative emotions may influence the type of responses, considering how negative emotions may shape an individual’s response is a critical component to predicting and managing crisis decisions. For individuals experiencing a crisis, it may be appropriate to consider that victims experiencing anger may not respond the same way as another victim in the same crisis experiencing shame. The victim experiencing anger may engage and take action, while the victim experiencing shame may withdraw and avoid taking any action. Likewise, some experts experiencing the same crisis may experience regret and move to rectify the situation, while other experts may experience shame and seek to withdraw.
Managing an effective crisis decision process means understanding how emotional response tendencies may push victims, and even experts, down a decision path that does not seem productive given the behavior necessary to promote safety and security. An awareness of the types of response behaviors to which negative emotions may relate—engagement, affiliation, and/or withdrawal—may help predict crisis decision behavior more appropriately. If behavior can be predicted more successfully, mitigation procedures for nonproductive behaviors may be implemented in a more targeted and effective way by leaders attempting to manage a crisis. These include, for example, more targeted messaging which encompasses emotion knowledge training—that is, what information the experience of anger, shame, or guilt provide—and training experts in emotional regulation procedures to direct more effective information processing and situation–action matching.
Individual to higher level crisis decision making
Understanding decision processes at an individual level of analysis provides a foundation to examine the implications of decision making as a crisis grows in scope and complexity. As evidenced by the opening discussion of the Fukushima Daiichi nuclear disaster, individuals and/or experts do not experience crisis in isolation and therefore, crisis decision models may be more robust when levels of analysis implications are considered beyond individual emotions and crisis decisions.
Stephens, Hamedani, Markus, Bergsieker, and Eloul’s (2009) research on “stayers” and “leavers” during Hurricane Katrina indicated that while some decision makers focus on choice and independence (i.e., “leavers”), other individuals expressed a more interdependent focus, where a desire to care for others and depend on others (i.e., “stayers”) took precedent. This interdependency notion introduces the idea that individuals may consider others during decision making, which is supported by prior crisis research (Drnevich, Ramanujam, Mehta, & Chaturvedi, 2009; Jin, 2010; Seeger, Sellnow, & Ulmer, 2003). In other words, an individual likely takes into account social factors such as how others may be perceiving, experiencing, and/or assessing the same crisis while making decisions. This allows for the notion that people experiencing crisis may be drawn toward or repelled from one another in different ways (Kahn, Barton, & Fellows, 2013).
A multilevel decision model offers the potential to examine this interdependent crisis decision making as individuals begin to include other individuals, groups, collectives, and experts/leaders into their crisis decision process. Table 2 highlights key relationships within the individual and collective crisis decision model.
Propositions for individual and collective crisis decision making.
Decision making in collective settings: Multilevel crisis decision model
Before presenting a multilevel decision model, a brief review of multiple levels of analyses may be helpful in defining terminology such as group and collective to better understand how these entities may interact within a crisis context. The levels-of-analysis theoretical framework is based on a traditional view within organizational research (Dansereau, Alutto, & Yammarino, 1984; Dansereau & Yammarino, 1998; Dansereau, Yammarino, & Kholes, 1999; K. J. Klein, Dansereau, & Hall, 1994; Rousseau, 1985; Yammarino & Dansereau, 2002, 2011) and recognizes four potential levels of analysis: person (i.e., individual), dyad, group, and collective. A person level of analysis refers to decision making where individuals are independent of one another. A dyad level of analysis reflects interdependency on a one-to-one basis between a pair of decision makers. A group or team level of analysis involves interdependency, not always face-to-face, across several individuals, although without echelons or hierarchy. And, finally, collectives are interdependent individuals, dyads, and/or groups that are held together due to hierarchical structuring and a set of shared or common expectations (e.g., “there is a crisis and it must be dealt with”). Moreover, as noted by Miller (1978), collectives may not involve a direct interaction between individuals within the collective (e.g., many individuals experiencing a crisis never see, talk to, nor know one another).
For the purposes of crisis decision making, while individuals may be acting solely or independent of others, what may be more likely is that an individual acts interdependently with at least one other person to reach a decision (i.e., dyad level of analysis). Another possibility may be that an individual interacts with multiple people experiencing the same crisis (i.e., group level of analysis), or possibly even neighbor groups, media, family outside the crisis, and/or government agencies (i.e., collective level of analysis; Drnevich et al., 2009; Jin, 2010; Seeger et al., 2003). Thus, these potential interdependencies at increasingly higher levels of analyses indicate a need to understand crisis decision making as multilevel phenomena.
Collective emotion and cognition within a crisis may be best described as a “strong” emergent state, rising from cognitions and emotions at the micro (i.e., individual) level, enduring for some period of time, and capable of exerting downward causal forces on the underlying parts (Lichtenstein, 2014). McGrath, Arrow, and Berdahl (2000) note that higher level (i.e., collective) variables both emerge from and subsequently constrain and shape lower level dynamics. For example, collective fear may exert a downward causal force on subsequent lower level group or dyadic interactions by increasing the probability of personal expressions of fear among dyads and group members.
By contrast, a weak emergent state would have no downward influence and, therefore, fear would have not causal force on subsequent interactions. Examining mechanisms of emergence such as social contagion and emotions cycles with collective cognitions and emotions may help illustrate how the process is more “strong” than “weak” in nature. How those cognitions and emotions emerge at the collective level, and how these emergent emotions can exert downward forces on underlying individual cognitions and emotions at the height of a crisis is not well understood. Therefore, the next section examines potential mechanisms at work within the emergence process, exploring how individuals experiencing crisis may form a collective emergent state with regard to cognitions and emotions.
Emergence of individual-to-collective information processing
Convergence, a bottom-up process, represents a cognitive process whereby an information-processing function shifts from an individual to a group (and therefore, shifts the level of analysis as well) to guide collective information processing (e.g., team mental model formation; Dionne, Sayama, Hao, & Bush, 2010; Kozlowski & Klein, 2000; McComb, 2007). Within this emergent process, McComb (2007) highlights three key phases that typify this cognitive processing shift from individual to team/group level of analysis: orientation, differentiation, and integration. Although these phases were established for explaining the emergent process from individual cognitions to group cognitions, recent research (Hareli & Rafaeli, 2008) suggests emotions of an individual influence the emotions of others, and therefore their emergence may adhere to similar emergent phases.
Additionally, while convergence is considered in mental model formation of groups, Hareli and Rafaeli (2008) note how emotions can affect unintended third parties, which refers to situations in which members of collectives may not be interacting but could be building a shared understanding of emotions. Thus, emergent, shared emotions and cognitions may not be limited to the group level of analysis as presented in prior convergence research, but likely extend emergence to a collective as well. To explore this notion, we discuss the mechanisms that may underlie the emergence of individual cognitions and emotions to collective-level cognitions and emotions. Research on social contagion and emotion cycles can explain how cognitions and emotions emerge at higher levels of analysis to be used recursively in information processing at the collective level.
Collective risk-as-feelings
Whereas the risk-as-feelings hypothesis has strong roots in individual risky decision models, given that emotions and cognitions have been linked to levels of analysis higher than the individual (Ashkanasy, 2003), risk-as-feelings likely applies to risky decisions at the collective level of analysis as well. Social contagion research may aid our understanding of the evolution of individual cognitions and emotions to a higher collective level of analysis. Classic research regarding social contagion and adoption of innovation (Burt, 1987) links the process to risk and uncertainty, which individuals manage by “drawing on others to help define a socially acceptable interpretation of the risk. Thus, social contagion arises from people proximate in a social structure using one another to manage uncertainty” (p. 1288). In social contagion, people turn to each other for cues to manage a situation (Burt, 1987), and to the extent that a person’s emotions are part of the cue, a normative understanding is created. As a result of the shared understanding, people may adopt the emotions of others as they share the same evaluation of risk and uncertainty (Ashkanasy, 2003; Barsade, 2002).
In particular, as crisis involves risk and uncertainty, to the extent that emotions are a part of the individual cognitive decision process, these emotions will be cued by individuals to others looking for assistance in defining the interpretation of the situation and its risks. As such, social contagion as a means for building a collective emotional position could reveal how emotions influence cognitive decisions made by a collective during a crisis, as individuals make dyadic connections, which then connect with other dyads, which then connect with groups, which then connect with other groups until a larger, like-emoted collective is established.
Similar to the recursive relationship discussed in the individual-level model, collective cognitions and collective emotions also are likely to have a recursive relationship. Research on mass panic provides evidence of highly intense fear (i.e., emotion) among crowds leading to judgments (i.e., cognitions) regarding rational courses of action such as collective flight or affiliative bunching (Durodié & Wessely, 2002; Mawson, 2005). Conversely, the emotional response to danger may be diminished by affiliating with familiar persons and surroundings (i.e., judgments to emphasize social ties), which may account for why sometimes panic is attenuated in crises (Aguirre, Wenger, & Vigo, 1998; Mawson, 2005).
Thus, collective emotions likely impact collective cognitions, and collective cognitions likely impact collective emotions. Collective risk-as-feelings likely provides collective information to use in the crisis decision and judgment process regarding risks and benefits. Similar to the individual level, experienced collective emotions and feelings are used as information in the collective decision making and judgment process regarding risks and benefits. Thus, as it was at the individual level of analysis, cognitive social cues help manage uncertainty while cognitive emotions help collectives decide how they think and feel about risk. As such, the following proposition is offered:
Given that crisis decision models have not addressed a collective risk-as-feelings hypothesis, examining the underlying emergence process involved in collective emotions and collective cognitions may improve the applicability of crisis decision models such as NDM and CDT as the crisis increases in scope. In particular, as NDM focuses particularly on expert decision making, understanding how emotions emerge to the collective level may benefit an expert’s decision performance during the most critical phases of an unfolding crisis. Research on emotion cycles (Hareli & Rafaeli, 2008) provides a compelling framework for experts to understand how one person’s emotions may induce comparable or different emotions in others through processes such as mimicking or interpretation. To the extent people look toward experts during a crisis, an expert should understand the important implications of his/her own emotions during crisis.
Emotion cycles
More recent research on emotion cycles also provides evidence related to emotions expressed at an individual level of analysis influencing and/or being adopted by other individuals, resulting (potentially) in collective-level emotions. Hareli and Rafaeli (2008) note that emotions expressed by an individual have the potential to influence the emotions of others, and via mimicking, the social presence of a specific emotion can be extended. Additionally, people can respond to others’ emotions via interpretation and assignment of meaning, which also extends the social presence of a specific emotion (Hareli & Rafaeli, 2008). Key to this research is the conclusion that these emotion cycles can involve both intended targets and unintended third-party targets of the emotion. Thus, emotion cycles also explain how specific emotions can emerge from an individual level to a collective level.
Emotion cycles research also indicates that emotions of an individual can influence not only the emotions of others, but the thoughts and behaviors of others as well (Hareli & Rafaeli, 2008). Emotion cycle research highlights three likely initial effects indicating how the emotions of one person can influence emotions, cognitions, and behaviors of others: (a) mimicking of emotion, (b) emotion interpretation and reacting, and (c) drawing inferences regarding an emotive agent. For experts within a crisis attempting to mitigate it, all three have key implications for crisis decision making.
Mimicking
Based on the notion that moods and emotions tend to spread among individuals (Hatfield, Cacioppo, & Rapson, 1992), an expert’s emotions can impact the unfolding of interactions among the collective. While a crisis may trigger negative emotions initially, even for experts, George and colleagues (George, 2011; George & Zhou, 2007) maintain, negative emotions promote problem identification. As the expert gains a sense of control and clarity, negative emotions give way to more positive ones, which can free a cognitive process focused on solutions for the expert as well as promote his/her sense-making (Maitlis, Vogus, & Lawrence, 2013). This improved processing of meaningful information supports the risk-as-information framework, suggesting that positive emotions enable active engagement and rapid context-specific information processing.
Specifically for NDM, mimicking suggests that as experts move to positive emotion and improved active/rapid contextual-bound information processing, the collective experiencing the crisis would also have negative emotions initially, but as the expert gains control and shifts to more positive emotions, the collective would follow as well and begin to shift to more positive collective emotions. In a sense, the mimicking creates an emotional contagion and, to the extent the contagion reflects positive emotions, the collective will mimic them.
However, there are two other processes within emotion cycles that may impact experts and NDM as well, and the emotion developed at the collective level may not mimic the expert’s emotion, but create uniquely different emotions.
Emotion interpretation
Given the processes that comprise emotion cycles, experts may not be able to project positive emotions and ensure the collective emotes the same way. Rather, the collective experiencing the emotion may interpret the emotion and react with a complementary or situationally appropriate emotion of their own. Moreover, emotional signaling to others is not a one-way directional process owned by the expert, meaning an expert’s emotions can be influenced by the collective’s response to the expert. Complicating this factor is that the expert’s emotions may inspire emotions in others who are not the direct target or even direct witnesses of the expert’s emotion (Hareli & Rafaeli, 2008).
Within NDM, a key aspect of the heuristic-based decision process is appropriately matching the selection of a response to the context. This situation–action matching process makes risk-as-information particularly difficult to implement, especially if the collective emotion differs from the expert emotion. In addition to managing their own emotions, the expert within a crisis must be an active participant within emotion cycles to at least empathize with the collective to understand and predict the impact of a collective’s emotions on the situation–action matching process. This allows the collective to map their own emotional reality, which arguably provides one of the ultimate challenges to managing the expert response to crisis given that a collective may not experience the emotion the expert believes best promotes effective crisis decision making. This discrepancy in negative emotion may slow the pace of contextual-bound information processing in that the expert may engage in more extensive information processing and be open and attentive to new information as he/she attempts to draw patterns and relate to stored schema.
Finally, collectives may draw inferences from emotion and extend the meaning of the emotion (Hareli & Rafaeli, 2008). For experts, this could mean knowledge of the expert’s emotion can lead the collective to presume knowledge of other things about the expert.
Inference drawing
This process may be particularly important within NDM, as the collective may draw inferences on the expert’s power, competence, and credibility. Importantly, initial emotions expressed by experts can lead to assumptions within the collective regarding the status and power distribution within a collective experiencing crisis. Interestingly, research shows that emergent leaders that show anger are judged as more competent than leaders who show sadness (Tiedens, 2001), and too much or too little emotion seems to lower credibility (Golding, Fryman, Marsil, & Yozwiak, 2003). Moreover, the authenticity and appropriateness of the expert’s emotion can be judged by the group. In other words, an expert initially expressing anger during a man-made crisis may be an entirely appropriate response, even from a leader, in that the collective views this emotion as consistent with their own emotion, and detection of emotional inauthenticity by a collective may lead to poor perceptions of the expert (Ashkanasy, 2003).
To the extent the expert is seen as a person in power that is credible and competent, expert negative emotions may be foundational to the development of collective emotion cycles. Moreover, the expert within NDM being viewed as a person in power with competence and credibility likely allows him/her to exert downward causal forces on the collective’s emotions. As such, experts may create strong emergent collective emotional states within NDM, in part based on inference drawing by the collective. This extension of the meaning of expert emotion is likely critical in sustaining strong collective emotion cycles. As such, the following proposition is offered:
Emotions cycles and NDM
Thus, emotion cycles play an important role in expert decision making within NDM, from the initial emotion displayed by the expert, to the emotional interpretation in which a collective may engage, as well as the inferences inspired by emotions that consider expert power, competence, and credibility. Ultimately, the expert likely needs to provide active situation–acting matching responses to a crisis, especially if the collective needs to be removed from harm’s way. For example, anger is a negative emotion related to solutions and therefore motivates engagement (Folger, 1987). Within NDM, while emotions may not always match between followers and experts, negative emotions on the part of experts may be appropriate as long as these emotions are consistent with active action and engaged responses to the situation experienced by the collective (i.e., anger, regret, and guilt). Negative emotions that promote withdrawal or avoidance of action (i.e., disappointment and shame), or promote affiliation (i.e., fear) above action may not motivate the collective to action within a critical timeframe of a relatively short crisis decision window. The expert needs to initiate an appropriate emotion to generate an appropriate risk-as-feelings in followers.
Collective risk-as-feelings may be shaped in part by expert emotion, and therefore NDM plays a central role in collective response and coping. However, despite prior research that indicates followers may mimic emotions of those in power positions, such as experts within crisis, a collective must have available mental resources to process information about an expert’s emotion. Unfortunately, time pressure on a collective limits the mental resources available (Jett & George, 2003), meaning that although an expert may be in a power position, the time limitations of a crisis may impede the formation of emotion cycles. A collective risk-as-feelings may not be easy to pinpoint within NDM, and therefore some understanding of various collective negative emotions and their potential corresponding collective response and coping style may help experts predict collective decision response within a crisis.
Collective cognitive response and coping
Cognitive appraisal
While cognitive appraisal is traditionally viewed as an individual-level appraisal theory, groups and/or collectives assess favorability of events as well. For example, groups make assessments of collective disadvantage, which renders intergroup comparisons and group identity salient (Tajfel & Turner, 1979; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987). Once group identity is salient, groups appraise events they encounter as a group rather than as individuals (Smith, 1997). Following the same trajectory as individuals, groups and collectives likely assess whether a crisis event disrupts their valued goals (i.e., threat assessment) or presents obstacles that can be overcome (i.e., challenge assessment). A threat assessment by the collective would be linked to negative collective affect, whereas a challenge assessment by a collective would lead to positive collective affect. Because collective emotions are likely to emerge from the experienced individual-level emotions, the immediate onset of a crisis is likely to prompt the elicitation of negative individual emotions, which through social contagion and emotion cycles are likely to evolve as collective negative emotions. And, as research indicates, emotions can be a collective property (Choi, Sung, Lee, & Cho, 2011; Kelly & Barsade, 2001).
Thus, because collective cognitions and collective emotions act recursively, we may expect similar information processing and appraisals as seen in the individual-level model. Recall anger, regret, and guilt produce a more engaged form of action tendency such as risk-seeking, corrective, and/or second-chance tendencies. The collective response may not differ from the individual-level response.
Fear, which differed in control appraisal from anger, regret, and guilt (i.e., control) may have an action tendency more similar to these emotions in that fear produces an urgent avoidance tendency. However, Mawson (2005) noted those taking flight during fear may prioritize seeking proximity of close others and familiar places. Thus, the engagement may differ for fear in that the action and/or behavior may not be constructive.
Similar to individual responses, collective disappointment may have an action tendency that favors getting away from a situation (Zeelenberg et al., 1998), which is another withdrawal-type action. However, disappointment may lead to inactivity (Zeelenberg et al., 2000), predominantly related to a belief that taking any action will be futile (Seligman, 1975). Shame is similar in that withdrawal or hiding tendencies are associated with this emotion (Tangney et al., 1996) and again, while active, the nature of the engagement is avoidant. Therefore, we propose (also see Table 1):
Coping
While coping is an important part of the individual appraisal process, recent research also illustrates how collective appraisals and collective emotions lead to collective coping (van Zomeren, Spears, Fischer, & Leach, 2004). Similar to individual-level models of coping, collective coping approaches also include problem-focused coping based on high certainty and control and emotion-focused coping based on low certainty and control. Because anger, regret, and guilt are based on high-certainty and control appraisals (Lerner et al., 2003), collectives experiencing these negative emotions may be more likely to use problem-focused coping such as taking action and searching for instructions. Conversely, because fear, disappointment, and shame are based on low-certainty and control appraisals (Lerner et al., 2003), collectives experiencing these negative emotions may be more likely to use emotion-focused coping such as venting and searching for emotional support. As such, we propose (also see Table 1):
Summary of multilevel crisis decision making
Thus, we have presented a multilevel crisis decision model that incorporates emotions into individual and collective crisis decision models, illustrates how cognitions and emotions at the individual and collective levels share a recursive relationship that influences appraisals and coping at those levels, and presents experts/leaders as a critical risk-as-feelings bridge within the individual and collective crisis decision process. The multilevel model, summarized in Figure 2, provides an illustration of the decision process that may occur during the containment/damage control phase of a crisis. This multilevel model, developed in the discussion and propositions asserted before, highlights the relationships among individual and collective cognitions and negative emotions, and their proposed outcome-related action tendency, activation level, and control appraisals—the interplay of cognitions and negative emotions at both individual and collective levels during a crisis.
Multilevel crisis decision making.
Example of multilevel crisis decision making
We now revisit the Fukushima Daiichi nuclear power crisis to highlight the interaction of both cognitions and emotions at the individual and collective decision levels, as well as highlight the criticality of the sense-making role of the experts/leaders during the crisis.
Individual level of analysis
Regarding the individual-level information processing, the report quotes individual employee responses such as, “because workers were desperately needed, I didn’t have time to confirm the well-being of my family, which bothered me so much that I could not concentrate on my duty…I feared for my life” (NAIIC, 2012, p. 68). Another employee reported, No information whatsoever about the station blackout was delivered to the end-workers. I had to learn about the emergency evacuation orders for residents within 20 km of the plant from TV…I asked to evacuate, but they declined my request. (2012, p. 66)
Collective level of analysis
As the nuclear disaster unfolded, employees in the plant, hampered by extreme conditions, lack of emergency instructions, and inadequate emergency and equipment manuals, attempted to mitigate the nuclear crisis. Results from the report revealed most (80%) TEPCO workers did not evacuate; however, only approximately 47% indicated they received an updated explanation of risk following the earthquake. Of the primary and secondary subcontracted workers that remained on-site, more than 95% were not updated regarding the hazardous conditions in the reactors. The report concluded that a “made in Japan” disaster followed the natural disaster when employees remaining on-site in the power plant reported a growing mistrust of the leadership (e.g., “they couldn’t trust anyone but themselves”; NAIIC, 2012, p. 66) and a significant lack of information related to hazards, risks, and ensuing evacuations (e.g., “for workers, there were almost no evacuation instructions”; p. 68). One employee noted “on the news it was reported that the plant workers who were dealing with the accident were prepared to die…there is no way we were ready to die”; p. 70).
In addition to employees beginning to understand they were working as a large collective for an even larger collective (e.g., “can we even say that our work was for the country?”; NAIIC, 2012, p. 70), the residents living around the Fukushima Daiichi power plant, dealing with their own personal crises related to the natural disaster, also became part of the collective as evacuation orders due to the nuclear disaster affected more than 145,000 residents. Survey responses from affected residents indicate a collective information-processing method (e.g., “young people were emailing ‘evacuate’ to each other, almost like chain mail…after hearing a neighbor who has a policeman in his family say ‘I’m going to evacuate because it seems dangerous,’ I decided to evacuate”; p. 54) and a sense that the evacuation information was downplayed to avoid a collective, high-intensity response (e.g., “because they might panic…because people in areas with more danger would not be able to evacuate”; p. 55).
Experts/leaders
As the nuclear crisis at the power plant began to escalate, the report concluded that TEPCO’s chairman and president were not available to communicate, and further did not understand the emergency response structure. Moreover, when the general manager of the Fukushima plant asked for technical advice from TEPCO’s head office, he did not receive a response. By then the prime minister of Japan was communicating directly with the plant’s general manager, and TEPCO headquarters approved instructions from a different government agency leader that were contrary to decisions being made on-site, all of which represented disruption and general confusion regarding the chain of command. The report concluded that a management mindset of “obedience to authority” (NAIIC, 2012, p. 33) flawed the crisis response and caused critical delays and confusion.
The Fukushima Daiichi disaster illustrates how the lack of a credible expert in a sense-making role contributed to a disaster “made in Japan” and produced ineffective and incorrect timing and instruction of evacuation orders. Critical in this expert role would be recognition that individuals both inside and outside the plant were experiencing high-intensity emotions and attempting to understand an evolving disaster. An effective expert should have messaged the collective working inside the plant to contain the disaster, as well as the various collectives outside the plant responding to the unfolding disaster regarding responses and coping behaviors. The confusion and lack of a timely expert/leader emergence potentially exposed tens of thousands of workers and residents to elevated radiation levels.
The investigative report revealed that a natural crisis evolved to an organizational crisis, which had individual employees experiencing individual cognitions and emotions while attempting to mitigate the organizational disaster. As the crisis escalated within the nuclear power plant, employees tried to make decisions in the absence of any clear expert emerging to solve problems. While the organizational collective performed heroically, they were not prepared or equipped to communicate to the larger collective (i.e., residents, and even the nation as a whole), causing critical delays in evacuation of the general public. As government agencies began mismanaging the residential evacuations, a collective of outraged Japanese citizens, who were not in the evacuated areas but were witnessing the crisis via the media, emerged as well. This example illustrates the individual and collective cognitions and emotions (expressed via interviews, surveys, and social media) experienced during an escalating crisis and the critical importance of experts’/leaders’ roles, both negative and positive, in sense-making.
Boundary condition: Emotion ambivalence
A key assumption of the model in the interest of parsimony and simplicity is that each emotion occurs in isolation (e.g., fear or anger). In reality, however, emotions occur together and could even co-occur along with positive emotions such as hope, pride, and compassion. Our approach to theorizing emotions’ effects follows a well-supported cognitive appraisals and discrete specific emotions perspective (Lerner & Tiedens, 2006). When two or more emotions co-occur, however, we expect that our propositions will involve slight caveats. For example, an individual experiencing fear and anger simultaneously might make more risky decisions if they perceive complete control over the future courses of action (driven by the anger appraisal), and adopt less risky, flight-based response strategies when they perceive little control over future courses of action (driven by fear). However, each scenario is also specific to a particular crisis event, and therefore likely evokes a different course of action. Given the establishment of key parameters of the crisis (e.g., factors such as control and favorability), we assert the core of our theory could then be extended and interpreted as parameters become clearer.
Practical implications for crisis management and expert decision makers
While managing a single individual or group through a crisis process may be difficult enough, consider there may be several, dozens, or even hundreds of groups affected. While within-level issues at the group level are complicated (due to differences among group members), within-level issues at the collective level are more complicated (due to differences among multiple players who may be individuals, groups, etc.). The potential for unique collectives with few within-collective differences but strong differences between one another interacting within the same problem space during a crisis seems particularly challenging. Similarly, the potential for many collectives of different individuals with unique perspectives within collectives exists as well, with an equally challenging task of reaching agreement and decision among the diversity within the collective. Moreover, these collectives are likely interacting with at least one expert and possibly more, as many organizations that manage and respond to crises are within the same decision space, represented by an expert or a team of experts. An important issue related to these conditions potentially affecting the rate and quality of the decision process centers around leadership.
Leadership literature regarding emergent leadership indicates followers may not necessarily believe the manager/leader in an authoritative position is the best person for solving problems or following in a crisis (Bass, 2008), which seems particularly relevant to inference drawing in emotion cycles. Network mapping may be able to indicate an emergent leader or more central “node” within a group or collective that followers view as more capable than the assigned leader (Akaishi et al., 2010; Dionne et al., 2012). This “central node” view indicates that the best person for addressing a crisis may not be a formal leader holding the highest position of authority, but rather a widely acknowledged expert in a specific type of crisis response or crisis management that has established critical relationships with several key constituents. This issue can be illustrated in the Fukushima Daiichi report detailing how the prime minister’s direct involvement with plant officials was detrimental to the crisis response (NAIIC, 2012).
Finally, while the current model is concerned with the containment/damage control phase, a longer term perspective could plausibly consider how this collective emergent state affects, particularly in later phases, organizational phenomena such as climate and culture. The challenge for organizations within a crisis may be responding appropriately and timely to it, as to more purposefully manage potential changes to climate and culture. Therefore, a clear understanding of the emergence process leading to a state of collective cognitions and emotions may provide greater insight into crisis decision making. Improved and informed decision making during a crisis may have important implications for organizational phenomena such as climate and culture following the crisis.
Future research
As with any conceptual or model-building work, the next step should be model testing. The current work is no exception. The challenge for the multilevel model presented here is that it contains and interplays both cognitions/heuristics and (negative) emotions and both the individual and collective levels of analysis simultaneously. To date, we do not know of any published empirical research that involves the complexity of all these elements to use as a testing template. (Admittedly, there are some empirical studies involving individuals and collectives and/or cognitions or emotions separately, but none combining all the elements.) Nevertheless, beyond the notions discussed that will impact testing procedures, there are some other elements to consider.
In particular, when to expect a social contagion phenomenon may rest on debate within network theory regarding “who” is involved in the contagion (Burt, 1987), and a similar notion may apply to emotion cycles. In the management arena, collective affect has been examined as simple compositional models (Bartel & Saavedra, 2000) or simple digressions from a group mean (e.g., Barsade, Ward, Turner, & Sonnenfeld, 2000). However, drawing upon extensive multilevel modeling literature (e.g., Kozlowski & Klein, 2000) and the powerful role of one-time disruptive events that could defy these simple linear collective affective cycles, we expect emotional contagion within social networks to follow an exponential curve with a single influential originating source causing an outward spiral. For example, one powerful financial decision maker could express fear, setting off fear in the collective investor pool.
Application of cohesive models (socialization-based) and/or structural equivalence models (competition-based; Burt, 1987) to crisis decision making may better inform who will be involved in emotion-based decision making. Although structural equivalence models are not usually considered at their extreme definition (i.e., people fighting/competing for survival), within a crisis model when survival may be an issue, this contagion process may convey new meaning to shared emotions within a decision process.
Ultimately, research within crisis decision making has a goal of understanding crisis processes to save lives and property whenever possible. The importance of understanding a multilevel crisis decision process is to enable crisis responders and managers, as well as governments, to best manage the crisis with the ultimate goal being protecting lives and properties of stakeholders as effectively and efficiently as possible. Thus, the implications of the proposed multilevel model are significant for expert decision makers tasked with managing a crisis. In light of the potential impact of emotions in crisis decision response, a better understanding of the complex and dynamic multilevel decision process offers a significant potential contribution to social science, and perhaps more importantly, may better inform future crisis decision making on potential crises such as environmental/climate change scenarios, terrorist attacks, disease epidemics, and/or natural disasters.
Footnotes
Acknowledgements
The authors wish to thank Neal Ashkanasy for his helpful comments.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.