Facing Others’ Trauma: A Role-Taking Theory of Burnout

Burnout

“Burnout” was first introduced in the 1970s to describe the disaffection of social aid volunteers in New York (Freudenberger 1974) and has since extended to a broad range of occupational contexts (Demerouti et al. 2001; Koopman and Hakemulder 2015; Maslach and Leiter 2017). Burnout has also expanded beyond professional sectors and been applied to family care (Mikolajczak et al. 2019), news consumption (Boukes and Vliegenthart 2017; Mannell and Meese 2022), social media participation (Han 2018; Sharma et al. 2022), social justice activism (Cox 2011), and many other settings. Burnout is defined by feelings of cynicism, detachment, and emotional exhaustion, with effects such as depression, anxiety, and withdrawal from burnout–inducing situations (e.g., paring back family duties, quitting a job, limiting news consumption, breaking from social media, reducing involvement in activist organizations; Boukes and Vliegenthart 2017; Cox 2011; Koutsimani et al. 2019; Mannell and Meese 2022; Maslach and Leiter 2017; Mikolajczak et al. 2019).

There are many interrelated causes of burnout, including overwork, undersupport, lack of resources, repetitiveness, and prolonged exposure to distressing circumstances (Bakker and de Vries 2021; Demerouti et al. 2001; Han 2020). It is this last factor, exposure to distressing circumstances, that we directly attend to in the present study. We do so because this factor weaves throughout multiple domains in which burnout occurs, with broad relevance across professional and personal life. For example, burnout is disproportionately high among those who work in social and humanitarian services, where practitioners are likely to interact with traumatized clients and encounter textual, verbal, and graphic depictions of traumatic events (Beck 2011; Bemiller and Williams 2011; Kim and Stoner 2008; Kolb 2014; Pearlman and Saakvitne 1995; Schaufeli, Leiter, and Maslach 2009; Singer et al. 2020; Wasco and Campbell 2002). More recently, research shows similar patterns in nonwork settings. For example, studies find news–related burnout connected to COVID-19 coverage (Gottfried 2020; Mannell and Meese 2022), a range of burnout indicators associated with doom–scrolling through negative content on social media (Han 2018; Sharma et al. 2022), and activist burnout during the Black Lives Matter and #MeToo movements, which concentrate attention on racial and gender violence (Gorski 2019; Richardson 2020; Strauss Swanson and Szymanski 2020). In each of these cases, trauma is present and prominent, with the potential for burnout to arise (Cohen and Collens 2013; Silver, Portnoy, and Peters 2015).

Although trauma exposure opens the door to burnout, this outcome is neither inevitable nor uniform across subjects and circumstances. The literature points to a range of variables that can mitigate or amplify the likelihood that burnout will ensue, such as strength of social support networks, institutional cultures and policies, and individual personality traits (Galek et al. 2011; Rupert, Miller and Dorociak 2015; Strauss Swanson and Szymanski 2020; Swider and Zimmerman 2010). What remains understudied is how interactional, social–psychological factors figure in. This leaves open a key piece of the puzzle, given that social–psychological processes link the micro to the macro, the self to the other, and the structural to the personal. We thus build on existing work by investigating a fundamental social–psychological construct—role-taking—in the burnout pathway, gaining insight into how burnout eventuates.

Role-Taking

Mead (1934) theorized role–taking as a hallmark of human development and community social life. Role-taking is the process and practice of putting the self in the proverbial shoes of another—understanding the world from the other’s perspective and behaviorally responding in turn. Role-taking serves as a fundamental building block of the social–psychology paradigm, underpinning structurally situated social interactions in which individuals present, interpret, and respond to meaningful social cues (Cast 2004; Davis and Love 2017; Love and Davis 2014, 2021; Mead 1934; Turner 1962).

As a distinctly structural concept, role–taking models social interaction between occupants of interrelated social roles. These roles inform who each person is in an interaction situation, accompanied by normative expectations about the thoughts, feelings, and behaviors of each role occupant (Love and Davis 2014, 2021; Mead 1934; Schwalbe 1988; Shott 1979; Turner 1962). From this structural base, role–taking holds two psychological dimensions—perspective-taking and empathy—accounting for cognition and affect, respectively. Perspective-taking is the cognitive process of discerning what others think and feel, and empathy refers to shared feeling and being moved by others’ emotions (Davis and Love 2017:167). Variation in role–taking is shaped by a number of factors, including social status hierarchies (Love and Davis 2014, 2021), similarity (or difference) between interactants (Cast 2004; Love and Davis 2014), and adherence to stereotypic beliefs (Avenanti, Sirigu, and Aglioti 2010; Johnson et al. 2009; Parlan 2015).

Active role–taking is often linked to prosocial outcomes such as social cohesion, helping behavior, and favorable attitudes toward outgroups (Bailenson 2018; Hooper et al. 2015; Penn et al. 2010; Rosenberg et al. 2013; Shott 1979; Stryker 1957). Yet, role–taking with traumatized persons may hold mental and emotional burdens, revealing a vulnerability in role–taking practice. Such potentials remain unexplored in the existing theoretical program, which has trained focus on role-taking’s interpersonal advantages. Burnout thus represents a productive entry point for building a holistic theory of role–taking, one that accounts for its multidimensional properties.

Hypotheses

Stimulus

The stimulus for this study is a narrative from a court order in a real civil protection order case. These court orders can restrain an abuser from contacting, threatening, or harassing a victim (Groggel 2021; Lynch et al. 2022; McFarlane et al. 2004). Victims typically apply for a protection order by filing a court petition that includes a written first–person account of abuse. We selected one of these accounts from an online justice database to which the authors have access and research permissions.

The case involves a woman filing against her ex–boyfriend. It details incidents of physical, sexual, and psychological abuse, including tactics of isolation and humiliation. The account is 297 words long and presented in handwritten text. To protect anonymity, a researcher rewrote the text verbatim in her own handwriting (rather than retaining the narrator’s original handwriting), and we assigned pseudonyms to the victim and perpetrator in the case. ³ We obtained approval from the Institutional Review Board and provided participants with warnings about the nature and content of the case they would encounter. Participants could withdraw from the study at any time and still receive full payment.

The victim narrative served as a theoretically relevant stimulus from which participant responses were derived. This approach circumvents the abstract self–reports that are common in survey research while allowing for targeted theory testing. However, it is limited in its recreation of our phenomenon of interest—burnout—which develops over time through prolonged trauma exposure. This is a necessary limitation due to ethical and practical concerns that preclude ongoing trauma exposure and related effects for purposes of research. To balance the dual priorities of theory testing and ethical–pragmatic constraints, we thus used a single, acute case of trauma exposure (the first–person victim narrative) to surface and observe relevant theoretical indicators. For this reason, our design represents a conservative test, with effects that would likely amplify in real–world settings where trauma exposure endures.

Sample

We recruited participants from Prolific Academic (http://www.prolific.co), an online crowdsource research platform. Crowdsource platforms are suitable for theory testing (Manago, Mize, and Doan 2021), as we do in the present work. Although these platforms often produce younger and more educated samples than U.S. workers in general (Hitlin 2016), they provide satisfactory data quality, with greater diversity than traditional university student samples (Hays, Liu, and Kapteyn 2015; Pe’er et al. 2021). Prolific also compares favorably against other crowdsource platforms, showing low levels of dishonesty among participants and high levels of naivety to common experimental tasks (Pe’er et al. 2017).

We collected an initial sample of 750 responses and dropped cases from the data set if a participant missed attention check questions or did not answer questions on focal independent or dependent variables or the response raised other quality concerns such as nonsensical or irrelevant textual inputs (3.6 percent). ⁴ We removed these observations through listwise deletion, creating a final sample of 723 participants (Allison 2001). Table 1 describes our sample: 51 percent are women, 73 percent of the sample is White, and 39 percent of the sample is married. We also asked about experiences of intimate partner violence due to its relevance for our stimulus. About 25 percent of the sample reports having been in an abusive relationship themselves, while 47 percent report having a close friend or family member that is or has been in an abusive relationship. This aligns with reports that more than half of Americans know a victim of either domestic violence or sexual assault (Avon 2013). Approximately 37 percent of the sample has a bachelor’s degree, which aligns with national trends, but our sample does have a higher number of advanced degrees compared to the general population. ⁵

OPEN IN VIEWER

Table 1.

Survey Respondent Demographics

	N	Percent
Gender
Men	355	49.1%
Women	368	50.9%
Race
White	532	73.6%
Black or African American	104	14.4%
Asian	109	9.4%
American Indian, Alaska Native, Native Hawaiian Pacific Island, other	67	2.6%
Hispanic	32	4.4%
Age
18–29 years old	186	25.7%
30–44 years old	208	28.8%
45–59 years old	162	22.4%
≥60 years old	167	23.1%
Political affiliation
Republican	144	19.9%
Democrat	357	49.4%
Independent	162	22.4%
Other	24	3.3%
No preference	36	5.0%
Marital status
Single, never married	259	35.8%
Living with partner	73	10.1%
Married	282	39%
Widowed	19	2.6%
Divorced or separated	90	12.5%
Education
High school degree or less	72	10%
Some college	211	29.2%
Bachelor’s degree	272	37.6%
Graduate degree	168	23.2%
Household Income
Less than $15,000	73	10.1%
$15,000–$39,999	184	25.5%
$40,000–$59,999	148	20.5%
$60,000–$79,999	121	16.7%
$80,000 or more	197	27.2%
Close friend or family in abusive relationship (0 = no, 1 = yes)	380	47.4%
Respondent has ever been in abusive relationship (0 = no, 1 = yes)	177	24.5%

Note: Observations = 723.

Analysis

To test our hypothesized relationships, we used a linear multivariate path analysis in Stata 15.1 with standard errors estimated using Satorra-Bentler adjustment. The model also includes the covariance between the errors for cognitive and empathic role–taking and covariance between our two measures of continued engagement. ⁶ In Stata, we used the command “estat teffects” to report the direct, indirect, and total effects for each path of the hypothesized model (Sobel 1987). We note that the path between two of our key variables—role-taking and anticipated burnout—represent a seeming effect of the former upon the latter. However, it is possible that there is a bidirectional relationship such that participants who know they are sensitive to burnout may limit how actively they engage with the victim narrative. This is particularly the case with cognitive role–taking whereby participants may choose to read the narrative less closely as a self–protective measure. Results should be considered with this caveat in mind.

Burnout and engagement

Anticipated burnout

We conceptualize anticipated burnout as a sense of emotional fatigue, operationalized through an adapted version of Maslach and Jackson’s (1981) measure for experienced burnout. Our adapted version of the scale contains three items in which participants indicate how emotionally fatigued they are from reading the case report, how frustrated they would be if they had to read cases like this every day, and how they anticipate they would feel if their job entailed reading similar cases, such that exposure to trauma was ongoing. The scale items are shown in Table 2 (Cronbach’s α = .84). Responses are scored on a 7–point Likert scale, ranging from 1 = strongly disagree to 7 = strongly agree. The theoretical range of this scale is 3 to 21, where higher scores reflect greater anticipated burnout.

OPEN IN VIEWER

Table 2.

Anticipated Burnout Scale

I feel emotionally fatigued after reading the case description.

I would feel burned out if my job involved reading cases like this.

I would feel frustrated by my job if I had to read cases like this every day.

Role-Taking

Recent developments in role–taking theory convey the process as an interrelation of cognitive and affective elements (Davis and Love 2017). This is operationalized in our study through cognitive and empathic role–taking measures.

Cognitive role-taking

Cognitive role–taking is conceptualized as thinking with the other—entering the other’s perspective (Davis and Love 2017). We operationalize this with an adapted version of Green and Brock’s (2000) Narrative Transportation Scale applied to the victim narrative, measuring the extent to which the case occupied the mind of the reader. The specific scale items are shown in Table 3. We use a total of 10 items measured on a 7–point scale, anchored by 1 = not at all and 7 = very much. Our adapted scale has a Cronbach’s α of .71. The theoretical range of the Narrative Transportation Scale is 10 to 70, with higher scores indicating higher levels of cognitive role–taking.

OPEN IN VIEWER

Table 3.

Narrative Transportation Scale

While I was reading the case, I could easily picture the events in it taking place.

I was mentally involved in the case while reading it.

I could picture myself in the scene of the events described in the case.

While I was reading the narrative, activity going on in the room around me was on my mind (reverse-coded).

After the case description ended, I found it easy to put it out of my mind (reverse-coded).

I wanted to learn how the case ended.

I found myself thinking of ways the case could have turned out differently.

I found my mind wandering while reading the case (reverse-coded).

The events in the case are relevant to my everyday life.

I had a vivid mental image of Mary.

Role-Taking Antecedents

To understand variation in initial role–taking responses, we predict cognitive and empathic role–taking with two theoretically and empirically relevant antecedents: similar experiences to the victim and stereotype beliefs about intimate partner violence.

Stereotype beliefs

To capture participants’ attitudes toward intimate partner violence, we use the Domestic Violence Myth Acceptance Scale, an 18–item scale that asks a series of questions across four dimensions: character blame of victims, behavior blame of victims, minimization of the seriousness and extent of abuse, and exoneration of the perpetrator (Peters 2003, 2008). Participants selected from a 7–point Likert scale ranging from 1 = strongly disagree to 7 = strongly agree. Higher scores indicate greater endorsement of intimate partner violence myths that denigrate victims and reflect supportive attitudes toward perpetrators of abuse (Cronbach’s α = .92). To optimize relevance for this particular study, we added four additional scale items pertaining to stereotypes about intimate partner violence victims in the justice system (e.g., “Women file restraining orders to punish their husbands or boyfriends”) and on the belief that court filings are used by women as a tool of manipulation (e.g., “Women lie to judges about abuse as a manipulation tool in divorce or custody cases”). The Domestic Violence Myth Acceptance Scale and our added measures were highly correlated, and hence combined, based on the mean of standardized scores. Possible scores on the prestandardized, combined summative scale range from 22 to 134. Higher scores represent stronger adherence to myths about intimate partner violence. Table 4 displays the full set of 22 items (Cronbach’s α = .94).

OPEN IN VIEWER

Table 4.

Myth Acceptance Scale

Domestic violence does not affect many people.

If a woman continues living with a man who beats her, then it is her own fault if she is beaten again.

Making a man jealous is asking for it.

When a man is violent, it is because he lost control of his temper.

Some women unconsciously want their partners to control them.

A lot of domestic violence occurs because women keep on arguing about things with their partners.

If a woman doesn’t like it, she can leave.

Most domestic violence involves mutual violence between the partners.

Abusive men lose control so much that they don’t know what they’re doing.

I hate to say it, but if a woman stays with the man who abused her, she basically deserves what she gets.

Domestic violence rarely happens in my neighborhood.

Women can avoid physical abuse if they give in occasionally.

Many women have an unconscious wish to be dominated by their partners.

Women who flirt are asking for it.

I don’t have much sympathy for a battered woman who ketif going back to the abuser.

Women instigate most family violence.

Domestic violence results from a momentary loss of temper.

If a woman goes back to the abuser, it is due to something in her character.

Women lie to judges about abuse as a manipulation tool in divorce or custody cases.

Women often make domestic violence claims out of revenge.

Women file restraining orders to punish their husbands or boyfriends.

If Mary gets back together with her boyfriend, then she should not be given another restraining order.

Note: The items in italics are the four additional scale items pertaining to stereotypes about intimate partner violence victims in the justice system.

Control Variables

We account for a number of social demographic characteristics that have been associated with burnout and/or role–taking. These include age, gender, race, and level of education. ⁹ Age and gender have been identified as significant indicators of burnout (Cheng et al. 2013; Iglesias, de Bengoa Vallejo, and Fuentes 2010; Lim et al. 2010; Marchand et al. 2015; Rosenberg and Pace 2006), while gender and race have been shown to affect role–taking via status processes (Love and Davis 2014, 2021). Dummy variables were created for each racial category: Black or African American, Asian, Hispanic, and an aggregate category that consists of Native American or American Indian, Pacific Islander, or another racial category. For each race dummy variable, the reference category is White. Level of education often has diffuse effects and is thus included as a binary variable indicating less than a college degree (0) or bachelor’s degree or higher (1).

Role-Taking Antecedents

Hypotheses 1 and 2 predict antecedents to role–taking, suggesting that shared experiences of abuse will increase cognitive and empathic role–taking scores (Hypotheses 1a and 1b) and that stereotypic beliefs about intimate partner violence will decrease cognitive and empathic role–taking scores (Hypotheses 2a and 2b).

Role-Taking and Anticipated Burnout

Hypotheses 3a and 3b predict that role–taking increases burnout such that higher scores on both cognitive and empathic role–taking measures will predict higher scores on the Anticipated Burnout Scale.

Anticipated Burnout and Continued Engagement

Hypotheses 4a and 4b predict a negative relationship between Anticipated Burnout Scale scores and indicators of willingness to continue engaging with trauma–related material. As outlined earlier, we have two measures of continued engagement. One measure indicates participants’ willingness to read through another case report of abuse as part of the present study. The second measure represents the number of cases participants believe they could read before becoming emotionally fatigued.

Role-Taking and Continued Engagement

Thus far, we have examined predictors of cognitive and empathic role–taking, the relationships between both measures of role–taking and anticipated burnout, and effects of anticipated burnout on participants’ willingness to continue engaging with trauma–related material. Results indicate that in general, increased cognitive and empathic role–taking correspond with higher levels of anticipated burnout, which in turn reduces willingness for continued engagement across short–term (subsequent case) and long–term (threshold) measures. Here we examine more closely how role–taking operates on this pathway, testing direct, indirect, and total effects of cognitive and empathic role–taking on the outcome measures for continued engagement (Hypotheses 5a, 5b, 6a, and 6b).

Limitations and Future Directions

As with all studies, the present work has several limitations. We address these here to contextualize the findings and establish new lines of inquiry. These limitations and future directions span burnout studies, role–taking studies, and the interrelation of the two.

One limitation is our use of a cross–sectional study design. We did this for reasons of ethics and practicality—it would be ethically and logistically untenable to instantiate persistent trauma exposure for purposes of research. However, in practice, burnout generally forms over prolonged periods with effects that build and accrue (Demerouti et al. 2001; Han 2020; Maslach et al. 2001; Maslach and Leiter 2017; Petersen 2021). Future work will benefit from longitudinal studies that capture actual thresholds after which burnout occurs, most likely conducted through multimethod fieldwork in settings with endemic trauma exposure. This may include expanded and behavioral measures of continued engagement, building on the two self–report indicators we use in the present work.

A second limitation is that our stimulus exposes participants to trauma of a particular sort (intimate partner violence), and our predictors of role–taking align directly with this kind of trauma, addressing prior experiences of abuse and stereotype beliefs. However, trauma takes a diversity of forms, each of which has its own unique character. The dynamics at play in the present work may not perfectly translate to other examples of trauma exposure, forecasting the need for theoretical and methodological adjustments. An important direction for future work will be to trace the role–taking to burnout pathway through multiple domains, understanding how different kinds of trauma exposure operate across subjects and circumstances. This includes studies of specific samples for whom the trauma exposure under study is directly relevant (e.g., burnout among social workers, social media content moderators, heavy news consumers, or activist leaders). These domain–specific studies will also be instrumental for creating effective interventions that alleviate burnout because any intervention must be fitted to context.

A third limitation is that we could not rule out all possible pathways between our variables. Most notably, survey participants who know they are sensitive to burnout may seek to prevent this outcome by disengaging from the narrative, thus reducing cognitive role–taking. Results should thus be read with this caveat in mind, while future work may investigate issues of bidirectionality.

Finally, our treatment of role–taking, as a construct, was limited by several interrelated factors. These include our use of a simple emotional change measure for empathic role–taking when in fact, empathy is relational, interpretive, and multifaceted (Batson et al. 1991; Clark 1987, 1990, 2007; Ruiz-Junco 2017; Zaki 2019; Zaki and Cikara 2015); leaving the relationship between cognitive and empathic role–taking unaddressed; and focusing on participants’ general sense of connection with the victim rather than the accuracy with which participants’ understood the victim’s experiences. There are elements of our results that suggest these limitations should be further explored. For example, shared experience of intimate partner violence increased cognitive role–taking scores, as predicted, but did not significantly predict empathic role–taking. However, in analyses not reported, we did see a significant pathway from cognitive role–taking to empathic role–taking for participants with personal and relational experiences of abuse. Exploring the nature of these relationships—between cognitive role–taking, empathic role–taking, and personally relevant trauma—will have explanatory value for the conditions under which trauma exposure results in burnout. On a basic theoretical level, it will also shed light on the interrelation of role-taking’s component parts. These efforts will be further enhanced by including multidimensional measures of empathic role–taking beyond a simple indicator of emotional movement, such as that which we employ in the present work.

Results should be interpreted within the scope of limitations detailed herein. Taking account of these limitations and moving forward from them, this work sets the stage for novel developments in both theory and method while holding practical implications for the study and treatment of burnout in contemporary societies.