Does affective empathy capacity condition individual variation in support for military escalation? Evidence from a survey vignette

Introduction

Does an individual’s affective empathetic capacity systematically condition their willingness to support pre-emptive military action in a crisis? The answer to this puzzle is important for several reasons. On a practical level, modern American politicians routinely call for pre-emptive military actions against “rogue states” (e.g., Trump, 2017). Understanding individual variation in voter support for these calls to arms, then, has axiomatic value for understanding real-world American foreign policy. On a more erudite level, a great deal of scholarly research currently demonstrates cognitive/strategic empathy’s capacity to dampen the psychological processes that contribute to conflict escalation (Kertzer et al., 2019; Yorke, 2023). ¹ Yet most existing work on affective empathy ignores its potential to lessen conflict escalation, choosing instead to focus on its humanitarian effects (e.g., Bayram and Holmes, 2020). ² Therefore, this note is the first (to my knowledge) to systematically investigate whether affective empathetic capacity might dampen support for conflict. ³

To investigate that possibility, I here present 121 individuals with a survey vignette about a hypothetical crisis scenario in which a rogue state is on the verge of acquiring nuclear weapons. In this context, I find that survey respondents with greater affective empathetic capacity scores will be less likely to support a military strike. This result confirms that affective empathy systemically conditions support for pre-emptive military action and has significant implications for our understanding of emotions, empathy, and conflict.

Theory and hypothesis

Why should an individual’s affective empathetic capacity affect their decision-making about whether to support military escalation? First, affective empathy removes the emotional stumbling blocks to the cognitive activation required for perspective taking. The affective desire to share emotions such as pain and pleasure provides a powerful mental incentive to evaluate one’s actions through the eyes of others, as perspective taking facilitates the sharing of emotions (Head, 2012). Affective empathy thus reinforces cognitive empathy—and its well-known tendency to deescalate conflicts (Yorke, 2023).

Second, affective empathy provides an altruistic rationale for conflict de-escalation. The altruistic motivation stems from feeling other-oriented emotion. Specifically, when one’s emotions are elicited by and congruent with the perceived welfare of another person in need, those emotions provide a motivational incentive to try and increase the other person’s welfare in order to have the empathy-inducing need removed (Batson, 2010). Both Edele et al. (2013) and Winczewski et al. (2016) demonstrate that this altruistic motivation cannot be explained by cognitive empathy. These rationales lead to my hypothesis.

Hypothesis:

Individuals with greater affective empathetic capacity will be less predisposed to support military action during a crisis.

The hypothesis is subject to some clarifying caveat and scope conditions. First, I caveat my hypothesis by explicitly stating that its independent variable (IV) is affective empathetic capacity—not situational affective empathy. I make affective empathetic capacity the IV to ensure external validity since, in the real world, nothing guarantees that a person’s heightened predisposition to feel emotions in general will translate into situational affective empathy. It is common, however, in existing international relations (IR) studies of cognitive empathy to experimentally induce perspective taking (e.g., Kertzer et al., 2019). To maximize external validity, I thus sacrifice the ability to compare my study “apples to apples” with existing experimental studies of cognitive empathy. Second, the hypothesis’s scope does not extend to escalations that are justified with a humanitarian rationale. It is entirely possible—even probable—that the theory’s logic would lead more emotionally empathetic individuals to be more supportive of escalation if the rationale were humanitarian.

Survey design and data collection

To investigate whether affective capacity might be correlated with support for crisis escalation, I turn to the results of a survey designed and administered on Mechanical Turk (MTurk) during 2020. Notably, 2020 was after the beginning of MTurk’s quality crisis (Kennedy et al., 2020). Therefore, to ensure only high-quality responses, I require that my 121 respondents (1) qualify as MTurk Masters, (2) complete two attentiveness checks, ⁴ and (3) not share IP addresses.

The overall design of the survey instrument is shown in Figure 1. It began with a standard battery of demographic questions before introducing respondents to a survey vignette borrowed largely verbatim from Tomz and Weeks (2013). The vignette describes a scenario in which a rogue state will build nuclear weapons within six months. The vignette specifies, “the country’s motives remain unclear, but if it builds nuclear weapons, it will have the power to blackmail or destroy other countries.” Respondents were then told that the state has one-tenth the military power of the United States and that a pre-emptive military strike on the country’s nuclear facilities would be successful.OPEN IN VIEWER

The vignette next informs the survey respondents that the state in question is autocratic and led by a single leader. Providing this information standardizes respondents’ understandings of regime type. The survey vignette further specifies the political situation in the foreign country:

“The country’s leader does not have unchallenged authority. Powerful domestic actors have urged the leader to take a more hard-line approach to relations with the United States. The leader has repeatedly referred to the United States as a ‘threat’.”

The vignette also shows the face of the “leader” in question (Figure 2): an AI-generated white male with a neutral expression. By using an AI-generated face that mirrors the demographic plurality within the sample, I keep the situation general for scientific purposes, avoid making the scenario about any real country, and control for variation in how respondents might imagine the autocrat’s appearance.OPEN IN VIEWER

I then ask respondents a series of questions about how they would respond to the situation. The first is whether or not they would choose to launch a military strike on the rogue state. I choose to make this variable, the dependent variable (DV), binary to force respondents to take a position. I next ask respondents a series of questions about their perceptions of the material consequences of attacking. This allows me to control for cognitive and other non-affective factors that may have affected a respondent’s decision-making.

Finally, I administer the Toronto Empathy Questionnaire (TEQ) to gauge each respondent’s affective empathetic capacity. I choose to administer my measure of affective empathetic capacity last to avoid biasing the respondents. ⁵ The TEQ is a 16-item questionnaire in which each element is measured on a 5-point Likert scale (Spreng et al., 2009). Affective empathetic capacity (the IV) is calculated by summing individuals’ responses to each item. TEQ scores can range from 0 to 64 and correlate positively with behavioral measures of social decoding and self-reported measures of empathy and negatively with measures of autism symptomatology (Spreng et al., 2009).

Analysis and results

The observational results yielded by the survey instrument are illustrated in Figure 3. In illustrating each subject’s decision about whether to strike, Figure 3 strongly suggests that individuals with higher TEQ scores are less likely to support escalation. To quantify this plainly observable effect, I employ binary logistic regression analysis.OPEN IN VIEWER

In Table 1, I show the results of four model specifications using logistic regression. I begin with a naive model. Using affective empathetic capacity as the IV, I find that a one-point increase in TEQ score is non-trivially associated with a 9% decrease in the probability of an individual supporting a strike. I next examine the effect of TEQ score on willingness to strike when controlling for demographic variables. Once again, TEQ score is the strongest predictor of a person’s willingness to escalate. I also observe that self-reported Republicans are 227% more likely to support escalation than non-Republicans. In the third model, an interaction effect is introduced to control for the empirically demonstrated effect of gender on TEQ score. ⁶ The result is to substantially increase the coefficient of TEQ score without fundamentally altering the results from the previous two models.

OPEN IN VIEWER

Table 1.

Effect of TEQ scores on strike probability under different model specifications.

DV: Willingness to strike
	Naive	+Demographics	+Inter. eff	+ Material beliefs
TEQ score (0–64)	−0.094***	−0.091***	−0.150***	−0.147***
	p = .0001	p = .001	p = .0002	p = .001
Demographic controls
White (0/1)		0.467	0.370	0.680
Age range (1–5)		−0.301	−0.352	−0.332
Democrat (0/1)		0.346	0.472	0.492
Republican (0/1)		1.214*	1.525**	1.793**
Education (1–5)		−0.362	−0.362	−0.444
Female (0/1)		−0.093	−5.697**	−4.935*
Interaction Effect Control
I (TEQ Score* Female)			0.133**	0.117**
Dummies for Material Outcome Beliefs
Belief in High Casualties				0.530
Belief in Damaged Relations w Allies				−1.140**
Expectation of Counterattack				0.174
Belief Stops Proliferation Long-Term				2.110***
N	121	121	121	121
Log Likelihood	−71.224	−67.700	−64.468	−54.040
AIC	146.449	151.400	146.935	134.080

Finally, in the last model specification, I control for a series of material expectations that respondents reported after deciding whether or not to strike. Controlling for cognitive and rational material factors that may have influenced a respondent’s decision-making allows me to better measure the true effect of affect. As expected, I see that multiple material factors impacted respondents’ decisions. However, even controlling for these material calculations, affective empathetic capacity continues to be one of the strongest probabilistic predictors of support for military action.

Conclusion

In this note, I evidence that affective empathy substantially conditions support for pre-emptive military action. In particular, using evidence from a survey vignette, I demonstrate that when rogue states are on the verge of proliferating, people with higher TEQ scores are less likely to support a military attack. The strength of these findings is that they are statistically robust regardless of model specification.

However, the findings also suffer from several limitations. For one, they rely on a single study with a relatively small sample size ( N = 121 ). ⁷ I encourage future researchers to expand upon this work with larger samples. Second, the survey vignette, while general and not about any specific country, does provide numerous details. This controls for variations in how respondents might imagine details of the scenario described, but it also might reasonably raise concerns about the study’s scope. I thus encourage other researchers to expand upon this work by studying other potential scenarios.

My limited findings, though, do have significant implications. First, they demonstrate that empathy broadly—not just cognitive empathy—can palliate conflict. This finding opens the door to studying a new type of empathy in a security context—an exciting prospect given the amount of scholarly attention that cognitive empathy alone has attracted over the past five decades (Yorke, 2023). I, therefore, hope that this note will contribute to new research at the intersection of empathy, emotion, and security studies.

Supplemental Material