Intimate partner violence (IPV), a pattern of abusive behaviors, including physical violence, sexual violence, stalking, and psychological aggression (which entails coercive tactics) by a current or former intimate partner, is a global public health crisis and human rights issue associated with significant health, economic, and social costs (Breiding et al., 2014). Between 20 and 50% of college students will experience a form of IPV during their college years (Scherer et al., 2016); these rates of victimization are highly concerning, especially given that a history of relationship abuse is a prominent risk factor for future victimization, highlighting the cyclical nature of this crisis.

A question that plagues clinicians and abuse survivors alike is why people who have experienced IPV and other forms of relationship adversity gravitate toward people who end up revictimizing them (Capaldi et al., 2012), often against their best intentions (Valdez et al., 2013). Although the answer to this question is unknown, one partial explanation is that people who have experienced IPV can develop psychopathological symptoms at full-blown or subclinical levels that can result in them processing social information in future relationship encounters, including dating interactions, differently (e.g., Zamir et al., 2018). These differences in information processing may lead them to make decisions in dating situations that influences their choice in partners. However, it is currently unknown how soon in interpersonal interactions these dynamics come into play. Are people able to accurately detect risk for aggression (i.e., hostile attitudes directed at another person; Buss & Perry, 1992) or IPV perpetration risk (i.e., risk of harm from an intimate partner), upon their first encounter with a potential dating partner? And are people equally good at detecting propensity for aggression and IPV perpetration risk, or does this vary as a function of their relationship experiences, such as their IPV victimization history and their experiences in attachment relationships? These are the questions we sought to answer with the current investigation.

To assess these important questions, we adopt a framework from personality psychology (the person perception methodology; Gosling et al., 2002) that has revealed important insights in terms of the information people glean from first impressions, as well as the degree to which individual differences in perceivers impact the accuracy of these impressions. Here we capitalize on this paradigm to examine whether college women, who are at heightened risk for IPV victimization and sexual assault during their college years (Moylan et al., 2019), can reliably detect college men’s propensity for aggression or IPV perpetration risk in romantic relationships after viewing online dating profiles. We also examine whether the women’s characteristics (IPV victimization history, attachment orientation) play a role in shaping their observations of these unknown men. Although men are also at risk for IPV during the college years and at later stages of development (Breiding et al., 2014), as well as LGBTQIA + individuals, in this initial study on this topic we focus exclusively on heterosexual women as they represent the group most at risk for victimization during this timeframe. Findings have the potential to reveal important insight and implications for prevention, as well as allows us to standardize and restrict the stimulus set of potential targets, thereby reducing statistical noise.

Person Perception Methodology

With impressive accuracy, the person perception methodology (Gosling et al., 2002) has illustrated that naïve observers (individuals who have received no previous training in person perception), can detect critical information about others’ internal states and characteristics when presented with limited information about them. These naïve observers can form reliable impressions of others’ personality traits based on the appearances of an office or bedroom (Gosling et al., 2002), the content of social media posts (Linkov et al., 2014), or narratives about one’s relationship (Borelli et al., 2018, 2019). A Brunswikian lens (1956) emphasizes the connection between objective behaviors exhibited by one individual to the emotions and cognitions perceived by others; this paradigm posits that naïve observers are able to detect individual characteristics in others by examining the behavioral residue of these qualities (e.g., a messy bedroom could indicate a lack of conscientiousness, and an observer who perceives this bedroom could then develop an impression of the person whose bedroom it is – that the person is not conscientious)—a detection process which is theorized to operate outside of one’s conscious awareness yet influence consciously accessible perceptions of others (Gosling et al., 2002).

The person perception methodology has particular relevance to romantic relationships (Borelli et al., 2018), as individuals often rely on the instantaneous judgments we render about others’ state of mind, values, and intentions to determine their level of interest in another person. Past research on social judgements, for instance, has revealed that people tend to seek out partners who appear demographically similar to them and who have similar attitudes, values, and personalities (Miller, 2012), while avoiding partners who appear to be dissimilar and those who deviate from expected relationship norms (Miller, 2012). Few studies have utilized the person perception methodology when examining romantic relationships, but those that have reveal that naïve observers can accurately rate participants’ strength of attachment (Borelli et al., 2018), the emotional ties people form throughout their lives to specific individuals from whom one seeks support and safety from (Bowlby, 1973), and accuracy in these ratings is not associated with the observer’s own attachment (Borelli et al., 2019). However, greater attachment avoidance is associated with less confidence in one’s ratings, providing a tantalizing hint that relationship history may influence the way a person may perceive or interpret behavioral residue. Additional research, however, is needed to expand upon this finding. For instance, how may attachment history impact accuracy of person perception when the situation is more personal or immediate to the individual and involves more attachment-relevant content? For instance, what if the observer is asked to evaluate a person’s dating profile as though they are in a potential dating situation? What if they are asked to evaluate them on characteristics with relevance to threat, such as potential for aggression or IPV perpetration risk? Both of these shifts render the situation more likely to activate the attachment system (Mikulincer & Shaver, 2007), whereas in prior studies, participants evaluated other people’ own relationship narratives of people (Borelli et al., 2018, 2019). Answering these questions has important implications for our understanding of how individuals interact with new people, as well as how their initial assessments may influence these interactions.

Accuracy in Person Perception

A major component of person perception research examines the degree of accuracy displayed in naïve observers’ views of others' traits, as well as the factors that enhance and limit accuracy. A common method of operationalizing observer accuracy in person perception is self-other agreement (Bernieri et al., 1994; Borelli et al., 2019), which refers to the degree of matching between naïve observer and observees’ perceptions of the observees’ traits (e.g., naïve observers’ perceptions of a man’s aggression history compared to that man’s report of his history) and then computed through the use of discrepancy scores, which uses the difference between the two ratings to determine whether the observer overestimated or underestimated their rating of the unknown person (Bernieri et al., 1994; Cohen et al., 2013). The accuracy of naïve observers’ ratings may be influenced by contextual factors, including the quality and quantity of the information presented to them, visibility of traits, and characteristics of the individual being observed (Biesanz et al., 2007; Funder, 2018). From a risk management perspective, overestimating traits in potential partners is more beneficial compared to the potential risks from underestimating them.

Participant Characteristics as Predictors of Person Perception

Intimate Partner Violence Victimization

Individual differences can also influence accuracy, as observers’ perceptions of others’ attributes can be influenced by their own characteristics (Hehman et al., 2017). One characteristic that may influence accuracy in naïve observers’ perceptions of an individual’s propensity for aggression and IPV perpetration risk is the observer’s own relationship abuse history. IPV victimization is associated with increased risk for revictimization by a new partner (Kuijpers et al., 2011) and a history of IPV victimization is associated with more distorted and harmful relationship views (Cravens et al., 2015). Taken together, this suggests that history of IPV victimization may affect one’s ability to accurately detect propensity for aggression or IPV perpetration risk in initial encounters with a potential partner, though this has yet to be examined within the person perception research tradition. IPV revictimization may occur as a result of various factors, including proximity to violence, sociodemographic factors (Kuijpers et al., 2011), and a desensitization to inappropriate relationship signals and scripts that may otherwise appear problematic for someone with no history of relationship abuse (Valdez et al., 2013). Exposure to IPV, for instance, increases one’s risk of posttraumatic stress symptomatology, which can result in symptoms that can block someone from making sound decisions regarding IPV risk. For instance, some people with posttraumatic stress disorder exhibit experiential avoidance, or the tendency to avoid distressing stimuli as a way to cope and protect oneself from the difficulties the stressor presents (Zamir et al., 2018), which can lead to an inability, whether consciously or unconsciously, to recognize early signs of danger. Another pathway by which IPV exposure can impact risk detection is that people exposed to IPV who develop posttraumatic stress symptoms can exhibit symptoms of hypervigilance and re-experiencing (Hebenstreit et al., 2015), which could make victims hyper-attuned to potential warning signs and threats of danger – people in this situation may be overly-sensitive to potential risk and unable to feel physically or emotionally safe. Ultimately, both of these pathways may lead to a failure to discriminate true risk, resulting in lower accuracy (in both directions, under- and over-estimating) in naive observers’ perceptions of unknown others’ aggression and IPV risk perpetration potential. Examining these associations and how IPV victimization history influences an individual's perception of potential romantic partners is paramount to our understanding of IPV and the cycle of violence.

Attachment Orientation

Other forms of relationship adversity can influence how observers perceive others (Fletcher & Kerr, 2010; Overall et al., 2015) – for instance attachment theory purports that in close relationships, individuals construct a model of the other person as either a source of comfort or instability based on their responsiveness in times of need. These expectations then guide the emotions and cognitions a person utilizes when navigating all close relationships (Hazan & Shaver, 1987; Mikulincer & Shaver, 2007), referred to as an individual’s attachment orientation or internal working model (Bowlby, 1973; Waters et al., 2009). Individual differences in attachment orientation, ranging from low to high anxiety and low to high avoidance (with low scores on both anxiety and avoidance denoting security), are reflective of beliefs and expectations people have about themselves and others, as well as their attachment histories (Hazan & Shaver, 1987). People with more secure orientations tend to have confidence and trust in personal relationships (Campbell & Stanton, 2019), while people higher in attachment anxiety are more likely to need reassurance and have a fear of abandonment (Borelli et al., 2018; Campbell & Stanton, 2019). Individuals higher in attachment avoidance, in comparison, display fears of interpersonal dependence and prefer space and isolation in relationships (Campbell & Stanton, 2019).

Attachment orientations not only inform how an individual interacts with others, but also act as the lens through which people interpret all attachment-relevant information (Borelli et al., 2019), including the perceptions people have when they interact with new people, highlighting their relevance for person perception. Like the lens of a camera, individuals may filter their interpersonal experiences through their attachment orientation, including when they perceive unfamiliar people (Mikulincer & Shaver, 2013). People who are more insecure interpret others as similarly insecure, as a way to make sense of their behaviors. Additionally, adults with higher attachment avoidance may ignore stimuli that evoke attachment-related thoughts and feelings (Tucker & Anders, 1999), such as overly romantic language or desires for closeness—themes which are often championed by individuals seeking a romantic partner (Frost & Forrester, 2013). Past research on the influence of attachment orientation on perceptions of unknown individuals, however, is sparse, with only one study to our knowledge examining this scenario (Borelli et al., 2019) and not finding significant associations between attachment and perceptions of unknown others. Importantly, these findings differ from person perception research of known individuals, which has consistently found associations between attachment orientation and observer accuracy. For instance, adults higher in attachment avoidance are less accurate in perceiving the emotions of others (Borelli et al., 2016; Feeney et al., 1994; Overall et al., 2015). People higher in attachment anxiety are have greater accuracy in perceiving a partner’s emotional states when the relationship is in distress (Fletcher & Kerr, 2010), but less accuracy in decoding the emotional tone of a partner’s behavior (Feeney et al., 1994) or their partner’s feelings about the relationship (Tucker & Anders, 1999). Therefore, more research is needed to understand whether attachment orientation is associated with observer accuracy of rating unknown individuals—information that has far-reaching implications for romantic relationship research and could provide vital insight into how initial perceptions of others influence interpersonal interactions.

The lack of research on the topic renders it difficult to make directional predictions regarding the nature of the associations between attachment and perceptions of others’ aggression and IPV risk perceptions. On the one hand, higher avoidance and higher anxiety might lead to underestimation of aggression and IPV risk perceptions, though for different reasons. Higher avoidance could lead to decreased accuracy of a potential romantic partner’s aggression and IPV perpetration risk due to the purposeful ignoring of information that is relevant for developing an accurate assessment of these target traits. For individuals higher in attachment anxiety, their fear of rejection and abandonment may make them idealize potential partners and overlook critical cues (e.g., controlling language) that may otherwise alert them to the risk of violence (Sandberg et al., 2019). Thus, they may be more inclined to underestimate risk in potential romantic partners due to their strong desire to establish close relationships (Sandberg et al., 2019). On the other hand, it is not difficult to wage an argument for the exact opposite case, wherein higher avoidance and anxiety lead to overestimation of risk – avoidance could be related to a paranoid type of perspective where others are viewed as threats to the self (Overall et al., 2015). Anxiety could lead to a stance where people misperceive actions as being threatening due to a hypersensitivity to anger and strong emotions (Overall et al., 2015). In sum, given the exploratory nature of this work, we do not hold a priori directional predictions regarding the links between attachment and perceptions of risk.

Current Study

Building on prior person perception research, this study aims to examine the degree to which female participants (i.e., naive observers), who have received no previous training in person perception, form accurate impressions of men’s propensity for aggression or IPV perpetration in romantic relationships based on viewing men’s de-identified dating profiles. Additionally, we aim to examine the degree to which individual differences in participants' own IPV victimization history and attachment orientation predict accuracy in these perceptions. We pursue the following aims within a sample of college females, as they are at heightened risk for IPV victimization during their college years (Moylan et al., 2019). We first predict that female participants will accurately rate the profiles of more aggressive men and men with more history of IPV perpetration as significantly higher in aggression and IPV perpetration risk compared to the profiles of less aggressive men (Hypothesis 1a; H1a) and men with less/no IPV perpetration history (H1b). In other words, we predict that their ratings of the men’s aggression/IPV perpetration risk will be sorted similarly based on the level of aggression/IPV perpetration history self-reported by these men. Next, we examine the associations between female participants’ personal characteristics (i.e., IPV victimization history, attachment orientation) and their rating styles. For IPV victimization history, we acknowledge that posttraumatic symptoms, such as experiential avoidance and hypervigilance, factors not measured in this study, may nonetheless differentially affect participants’ interpretation of social information and thus their ratings. As a result, we examine these associations in an exploratory way without an a priori hypothesis, evaluating associations between IPV victimization history with perceptions of aggression (Research Question 1a; RQ1a) and IPV perpetration risk (RQ1b). Given the mixed findings in the literature between attachment orientation and perceptions of others, we similarly examine these associations in an exploratory manner, evaluating associations between attachment avoidance and anxiety with perceptions of aggression (RQ2a) and IPV perpetration risk (RQ2b). Finally, we explore whether participant characteristics (IPV victimization history, attachment orientation) are associated with participants’ accuracy in rating of aggression (calculated using the men’s own ratings of their aggression as a reference point) and IPV perpetration risk (calculated using the men’s own abuse perpetration history as the reference point). As above, we examine these associations in an exploratory way without an a priori hypothesis as little evidence is available on how these characteristics may impact participant accuracy in estimating men’s propensity for aggression (RQ3a, 4a) and IPV perpetration risk (RQ3b, 4b) in unknown partners.

Method

This study was approved by the Institutional Review Board (HS#2020–6304). To facilitate the study design, procedures were completed in two-parts: Part 1 involved having men create de-identified dating profiles and report on their aggression and IPV perpetration history, while Part 2 involved female participants rating the dating profiles in terms of aggression and IPV perpetration risk, as well as reporting on their own attachment orientation and IPV victimization history. Each part of the study is detailed below.

Part 1: Developing De-Identified Dating Profiles for Use with Female Participants

Eligibility for Part 1 included being over 18 years of age, identifying as male, identifying as heterosexual, currently being in a romantic relationship or previously being in a romantic relationship, and being able to read and write in English. Past relationship history was required for men due to this study examining IPV perpetration risk based on history of IPV perpetration. Men who met eligibility criteria were recruited from the online social science subject pool at a 4-year research university and asked to complete an online survey and activity. The survey included questions regarding their own experiences (e.g., aggression, trauma history) and their romantic relationship history. The activity centered on creating a de-identified (i.e., fake name, non-identifiable meme/photo as their profile picture) dating profile that was formatted to mirror the questions and style of popular dating sites (e.g., Tinder, Hinge). Specifically, they were asked to create their own responses to the following pre-set prompts: About me (e.g., “I am reserved at first but warm up quickly #dating #naughtyandnice”); Dating me is like… (e.g., “the best thing that's ever happened to you”); Relationship pet peeves (e.g., “neediness”); Perfect first date? (e.g., “all the first ones are bad ones, just wait”); Typical Saturday (e.g., “chill at home, video games, grocery shopping”); I’ll brag about you to my friend if… (e.g., “you have a great body”); I’ll fall for you if (e.g., “you keep looking at me when I speak to someone else”); I geek out over (e.g., “movies/tv shows”); Life goal (e.g., “to provide financial security to my parents”); Friends would describe me as (e.g., “funny, intelligent, risk taker”). Men were also asked to select a picture from a list of pre-selected images from popular culture (e.g., a meme, a character from TV/Film, sports team logo, nature photo, etc.) to serve as their profile image. These procedures were selected as a way for the men to personalize the profiles and choose images that would reflect the type of person they were trying to present (e.g., a friendly and likable character to demonstrate that they are a friendly and likable guy, a tough or macho character to demonstrate that they are tough and macho), similar to how dating profiles work in real life. This image/icon selection procedure stood in the place of men uploading their own images, which would have resulted in identifiable profiles when the goal was to keep the profiles de-identifiable. Men were told that the dating profiles would be utilized to test different profile layout formats and were encouraged to answer as honestly as possible to all dating profile prompts to ensure accuracy. (See Figure 1 in the Appendix for a sample dating profile). A total of N = 41 men completed the survey and activity. Men’s aggression was assessed using the Buss-Perry Aggression Questionnaire (BPAQ; Buss & Perry, 1992; α = 0.84), a 29-item measure that assesses various forms of aggression, including physical aggression, verbal aggression, anger, and hostility (e.g., “if I have to resort to violence to protect my rights I will”), on a 4-point scale from 0 (extremely uncharacteristic) to 4 (extremely characteristic). The scale is scored in this study as the average of responses on each item across the entire scale, with higher scores reflecting strong self-reported aggression. This measure has high validity and reliability in past studies with adults (Harris, 1997).

Similarly, IPV perpetration was assessed using the perpetration scale from the Conflict Tactics Scale 2 Short Form (CTS2S; Straus & Douglas, 2004). This self-report scale assesses history of physical abuse perpetration (e.g., “I punched or kicked or beat-up my partner”), psychological abuse perpetration (e.g., “I insulted or swore or shouted or yelled at my partner”), sexual abuse perpetration (e.g., “I insisted on sex when my partner did not want to or insisted on sex without a condom (but did not use physical force)”), and injury to a partner (e.g., “my partner went to see a doctor (M.D.) or needed to see a doctor because of a fight with me”). We followed the recommended scoring method for this scale, which is to dichotomize the data as either present (coded as 1) or not present (coded as 0) for each form of abuse and then sum the total perpetration score ranging from 0 to 4 (Straus & Douglas, 2004).

On average, the men in the profiles reported being 21.98 years of age (SD = 4.11), with an overall mean aggression score of 1.35 (SD = 0.55), and a mean IPV perpetration score of 0.93 (SD = 1.25). In an effort to identify various levels of aggression and IPV perpetration for female participants to rate, we identified relevant profiles (based on their reported perpetration history and aggression scores) and placed them in three relevant categories on the measures of interest: low, moderate, or high aggression/IPV perpetration. N = 3 profiles met the criteria for low aggression and low IPV perpetration (operationalized as no history of abuse perpetration and low aggression scores, Mlow = 0.63, SD = 0.09), n = 3 profiles met the criteria for moderate aggression and moderate IPV perpetration (operationalized as some abuse perpetration history [2 out of 4 forms] and moderate aggression scores, Mmod = 1.58, SD = 0.19), and n = 3 profiles met the criteria for high aggression and high IPV perpetration (operationalized as history of all four forms of abuse perpetration–physical, emotional, sexual, and injury–and high aggression scores, Mhigh = 2.82, SD = 0.23), resulting in a sample of 9 dating profiles. On average, the 9 men in the selected profiles reported a mean age of 23.40 years old (SD = 2.45), an overall mean aggression score of 1.44 (SD = 0.79), a mean IPV perpetration score of 2.00 (SD = 1.73), and reported having been in 3.44 relationships (SD = 1.24). Most men self-described as Asian (33%) or Latino (33%) (with 11% African American, 11% white, 11% multiracial) and all but one man reported a preference for dating one person at a time compared to multiple people at once. Supplemental Table 1 in the appendix provides descriptive statistics on all of the men who created profiles (N = 41), as well as the men whose profiles were selected for the study (n = 9). T-tests revealed there were no significant differences between the group of men whose profiles were selected for the study and the larger sample of men from which the profiles were drawn on the number of past relationships, age, or dating style.

Part 2: Female Participant Study Protocol

Following the creation of the dating profiles, female participants were then recruited from the online social science subject pool at a 4-year research university to participate in a study about romantic relationships. Eligibility for Part 2 included being 18 years of age or older, identifying as female, identifying as heterosexual, currently being in a romantic relationship or previously being in a romantic relationship, and being able to read and write in English. Past relationship history was required for participants as this study examines IPV victimization history as a predictor of participant accuracy in ratings of men’s aggression and IPV perpetration risk. The resultant sample consisted of N = 443 females. On average, participants had a mean age of 21.87 (SD = 2.77), reported having been in 3.04 relationships (SD = 1.09), and most (97.7%) participants reported a preference for dating one person at a time. Nearly half (49.4%) of participants self-described as Asian, with 29% Latina, 12% white, 6% African American, 2% Middle Eastern, 1% Pacific Islander or Native to Hawaii making up the rest of the sample’s demographics.

Procedures

Those who met eligibility criteria completed an online activity, followed by a survey about their experiences (e.g., IPV victimization history, attachment orientation, demographics). As part of the activity, each participant was assigned to view all 9 de-identified dating profiles (order counterbalanced to ensure there were no order effects) and asked to rate each person's profile on a scale from 1 (not at all) to 7 (very) on various traits related to romantic relationships, including aggression and dangerousness (our term for IPV perpetration risk). Participants were instructed to form an impression of the individual based on the dating profile and then rate them accordingly on the following traits: financially responsible, romantic, lazy, funny, aggressive, extraverted, successful, attractive, and dangerous. Including additional traits for evaluation by participants was employed due to our concerns that participants would be able to identify the aims of the study if they exclusively completed measures on aggression and IPV perpetration after viewing each dating profile; thus, we chose to have them evaluate each dating profile on a series of traits. As a result, the participants’ aggression/IPV perpetration ratings of the men were on a different scale than the men used to report their own ratings and were standardized for analysis and interpretation (e.g., participants rated men’s aggression on a scale of 1 to 7, while men’s aggression was assessed via the Buss-Perry Aggression Scale, which ranges from 0 to 4). Participants were naive to all information about the profile creators. On average, participants spent 37.12 min (SD = 69.29) to complete the ratings. To conduct analyses, we first computed standardized mean scores for each rater’s evaluation of aggression/IPV perpetration risk for each profile, resulting in a total participant score for each profile’s aggression or IPV perpetration level.

Measures

IPV Victimization History

Participants’ IPV victimization history was assessed using the victimization scale from the CTS-2S (Straus & Douglas, 2004). This self-report scale assesses history of physical abuse victimization (e.g., “my partner pushed, shoved, or slapped me”), psychological abuse victimization (e.g., “my partner insulted or swore or shouted or yelled at me”), sexual abuse victimization (e.g., “my partner used force (like hitting, holding down, or using a weapon) to make me have sex”), and injury by a partner (e.g., “I had a sprain, bruise, or small cut, or felt pain the next day because of a fight with my partner”). The scoring method for this scale is to dichotomize the data as either present (coded as 1) or not present (coded as 0) and then scored continuously from 0 to 4 for each form of abuse (Straus & Douglas, 2004).

Attachment Orientation

Participants completed the Experiences in Close Relationships—Relationship Structure Scale (ECR-RS; Fraley et al., 2011), a reliable and valid 9-item measure of adult attachment orientation as it pertains to romantic relationships (Fraley et al., 2011). Participants were asked to focus on their current/most recent committed romantic relationship. The measure uses a 7-point Likert-type scale (strongly disagree [1] to strongly agree [7]) to measure attachment avoidance (e.g., ‘‘I prefer not to be too close to my romantic partners;” α = 0.77) and anxiety (e.g., ‘‘I worry a lot about my relationships;” α = 0.80).

Data Preparation

Levels of Aggression/IPV Perpetration Risk

To assess whether participants accurately rated more aggressive profiles and profiles of men with higher IPV perpetration as significantly more aggressive/dangerous compared to less aggressive profiles and profiles of men with less IPV perpetration history (Hypothesis 1), we created mean ratings for aggression and IPV perpetration by profile level of aggression/IPV perpetration (low, moderate, high) to compare against one another using analyses of variance tests.

Discrepancy/Accuracy Score Computation

In order to evaluate participant accuracy in ratings of men’s aggression and IPV perpetration risk (Hypothesis 3), we created directional discrepancy scores by taking the difference between each participant’s standardized rating of the man’s aggression/IPV perpetration risk level and the man’s standardized self-rating of aggression/IPV perpetration (Dimler et al., 2017). We completed this procedure for the 9 profiles rated by all participants, and then we computed a mean score of all these differences. Negative values signify that, on average, the participant overestimated the profile creator’s aggression/IPV perpetration risk (relative to the man’s own rating), while positive values indicate that the participants underestimated the profile creator’s aggression/IPV perpetration risk (relative to the man’s own rating).

Data Analyses

To test study hypotheses and research questions, we conducted analyses of variance and hierarchical linear regressions via SPSS Statistics for Macs, Version 26.0, controlling for participant age, number of relationships, and race (using dichotomized race variables for Asian, Latina, white, and African American individuals), as past research has illustrated that these factors are associated with IPV victimization and may influence how participants perceive the dating profiles (Cho, 2011; Halpern et al., 2009). After removing all participants who failed to complete the first activity (N = 11), analyses included N = 453 participants with no missing data. Associations between key study variables and demographic variables were also assessed using independent samples t-tests, analyses of variance, and correlations. When examining participant characteristics as predictors of their rating tendency and accuracy, all characteristics (IPV victimization history, attachment anxiety, attachment avoidance) were entered in the second step of the regressions. Participants’ mean aggression/IPV perpetration risk rating of the dating profiles was the dependent variable for regressions related to Research Question 2, while participants’ mean discrepancy rating was the dependent variable for regressions related to Research Question 3.

Results

Over half (62.3%) of the participants in our sample reported IPV victimization history (with 54.3% of participants reporting a single form of victimization, 45% of participants reporting two forms of abuse, 8% of participants reporting three forms of abuse, and 14% of participants reporting all four forms of abuse). The most endorsed form of IPV was psychological abuse (54.9%), followed by sexual abuse (23.3%), then physical abuse (16.3%) and finally, experiencing an injury as a result of abuse (12%). Participants who self-identified as Latina had the highest percentage of victimization (65.9%), followed by Asian (64.1%), white (61.5%), and African American (54.5%) individuals. See Table 1 for descriptive statistics and bivariate correlations among key study variables.

Table 1 Descriptive statistics and bivariate correlations for key study variables among female participants
Full size table

Hypothesis Testing

  • Hypothesis 1a: Does the sample as a whole rate highly aggressive profiles as more aggressive than moderate or low aggressive male profiles? Analyses of variance revealed that, on average, there was a significant difference between how female participants rated low, medium, and high aggression profiles [F(2,1326) = 56.99, p = 0.001]. A Tukey post hoc test revealed that the judges rated the high aggression profiles as significantly more aggressive than the moderate aggression profiles (p < 0.001), and moderate aggression profiles as significantly more aggressive than low aggression profiles (p = 0.02). Participants were able to accurately differentiate between men’s propensity for aggression based on viewing their dating profile.

  • Hypothesis 1b: Does the sample as a whole rate male profiles with high IPV perpetration history as more dangerous than male profiles with moderate and low IPV perpetration history? Analyses of variance revealed that, on average, there was a significant difference between how female participants rated profiles with low, medium, and high IPV perpetration history [F(2,1326) = 10.41, p = 0.01]. A Tukey post hoc test showed that the judges rated the high IPV perpetration profiles as significantly more dangerous than the moderate IPV perpetration profiles (p = 0.001); the ratings between moderate and low IPV perpetration profiles were not significantly different (p = 0.52). Participants were able to accurately differentiate between high IPV perpetration risk profiles but were not able to distinguish between moderate and low IPV perpetration risk profiles.

Research Questions

Research Questions 1a, 2a

Are female participant characteristics (IPV victimization history, attachment orientation) predictive of their ratings of men’s aggression? After controlling for participant age, race, and number of relationships, R2 = 0.03, p = 0.05, the step containing female participants characteristics (IPV victimization history, attachment anxiety, attachment avoidance) was significantly associated with female participants’ ratings of men’s aggression, ΔR2 = 0.02, p = 0.02. However, only attachment anxiety (b = 0.09, SE = 0.04, p = 0.02) and avoidance (b = -0.05, SE = 0.02, p = 0.02) were significantly associated with participants’ ratings of men’s aggression, with those higher in attachment anxiety rating men’s as more aggressive and those higher in attachment avoidance rating men as less aggressive. Participants’ IPV victimization history was not associated with participants’ ratings of men’s aggression (b = -0.04, SE = 0.03, p = 0.19). See Table 2 for regressions examining participant characteristics in predicting men’s aggression based on their dating profiles.

Table 2 Regression examining female participant characteristics (attachment orientation and relationship abuse history) in predicting men’s aggression and intimate partner violence perpetration risk
Full size table

Research Questions 1b, 2b

Are female participant characteristics (IPV victimization history, attachment orientation) predictive of female participants’ ratings of men’s IPV perpetration risk? After controlling for participant age, race, and number of relationships, R2 = 0.02, p = 0.11, the step containing participant characteristics (IPV victimization history, attachment anxiety, attachment avoidance) was not significantly associated with participants’ ratings of men’s IPV perpetration risk, ΔR2 = 0.01, p = 0.42, with neither relationship abuse history (b = 0.02, SE = 0.03, p = 0.46) or attachment orientation (anxiety: b = 0.05, SE = 0.04, p = 0.26; avoidance: b = -0.04, SE = 0.02, p = 0.15) being significantly associated with ratings of men’s IPV perpetration risk level. However, participant race (African American; b = 0.47, SE = 0.24, p = 0.05), was positively associated with ratings of men’s IPV perpetration risk level. See Table 2 for regressions examining participant characteristics in predicting men’s IPV perpetration risk level based on viewing their dating profile.

Research Questions 3a, 4a

Does accuracy in female participants' ratings of male aggression vary as a function of female participants' characteristics? Controlling for participant age, race, and number of relationships, R2 = 0.03, p = 0.03, this analysis revealed that the step containing participant characteristics was significantly associated with standardized discrepancy scores in ratings of men’s aggression, ΔR2 = 0.02, p = 0.03, with both attachment anxiety, b = -0.07, SE = 0.03, p = 0.01, and attachment avoidance, b = 0.04, SE = 0.02, p = 0.01, but not IPV victimization history, b = 0.02, SE = 0.02, p = 0.43, explaining a significant proportion of the variance in the model. Thus, participant attachment anxiety was significantly associated with overestimating men’s aggression, while participant attachment avoidance was significantly associated with underestimating men’s aggression. See Table 3 for the regression examining participant characteristics in predicting accuracy in ratings of aggression.

Table 3 Regression examining female participants’ characteristics (attachment orientation and relationship abuse history) in predicting accuracy in participants’ ratings of men’s aggression and IPV perpetration risk
Full size table

Research Questions 3b, 4b

Does accuracy in female participants' ratings of male IPV perpetration risk vary as a function of female participants' characteristics? Controlling for participant age, race, and number of relationships, R2 = 0.03, p = 0.09, this analysis revealed that the step containing participant characteristics was not significantly associated with standardized discrepancy scores in ratings of men’s IPV perpetration risk level, ΔR2 = 0.01, p = 0.35, with neither IPV victimization history (b = -0.3, SE = 0.03, p = 0.32) nor attachment orientation (anxiety: b = -0.04, SE = 0.03, p = 0.26; avoidance: b = 0.03, SE = 0.02, p = 0.15) explaining a significant proportion of the variance in the model. However, participant race (African American; b = -0.39, SE = 0.19, p = 0.04), was negatively associated with accuracy of men’s IPV perpetration history, illustrating that African American participants overestimated IPV perpetration risk. See Table 3 for the regression examining participant characteristics in predicting accuracy in ratings of IPV perpetration risk.

Discussion

The reasons why people select partners who may harm them continues to perplex clinicians and survivors alike. This study attempted to shed light on this issue through the use of a highly controlled laboratory paradigm, finding that, on average, college females could distinguish between high, medium, and low aggression males and between males who were high and low in IPV perpetration risk. These novel findings suggest that propensity for aggression may be fairly detectable even from online dating profiles. This large-sample study provides important information regarding college females’ reliability and accuracy in detecting both aggression and IPV perpetration risk at higher levels.

In contrast, propensity for IPV perpetration may be more difficult to discriminate at lower levels. In this situation, face-to-face contact or more in-depth exposure to the person may be necessary to parse out subtler dangerousness tendencies. Alternatively, it may generally be more challenging to detect IPV perpetration risk in another person without knowing them well enough to have witnessed their disposition across various scenarios, highlighting the barrier that online spaces may present for risk detection (Chin et al., 2019). However, once face-to-face contact has been made, it may be more difficult for people to extract themselves from the situation, even if they do detect risk of harm. Once someone has already made an in-road into one’s life, it may be harder for someone to set a limit with them (foot-in-the-door technique; Crano & Sivacek, 1982). This may be particularly true for people who have a prior history of IPV; standing up for themselves in risky situations may be even more challenging (Cravens et al., 2015). Additional research is needed to confirm and replicate these findings, as well as build on our understanding of how college students perceive aggression and risk for harm when online dating.

We also examined whether participant characteristics (IPV victimization history, attachment orientation) were predictive of the overall magnitude of their ratings of men’s aggression/IPV perpetration risk level. IPV victimization history was not associated with participants’ ratings of men’s aggression; women with a history of IPV victimization did not perceive men as significantly more or less aggressive compared to women without a history of IPV victimization. Perhaps some survivors have blunted sensitivity to threatening or abusive characteristics in others while others have heightened sensitivity to the same cues, resulting in bifurcating paths and the absence of a linear association between relationship abuse history and rating patterns. In future studies, it will be important to measure the psychological symptoms that can follow IPV, such as symptoms of post-traumatic stress disorder (PTSD) for its association with perceptions of risk. Further, it will be crucial to examine potential moderators of the link between PTSD symptoms and perception of risk. Further, future work may wish to examine the interaction between IPV history and attachment, testing the question of whether people who have experienced IPV and have an avoidant way of orienting to attachment relationships perceive risk differently than people who have experienced IPV who have high attachment anxiety. It is possible that our lack of effects represent true non-effects, and that experiencing violence may not impact people’s perceptions of others’ characteristics (e.g., their potential for aggression or their IPV perpetration), at least not when exposed to people in this dose. If experiencing violence is not associated with more sensitivity to aggression/IPV perpetration risk (e.g., higher ratings), this begs the question as to whether college women who have experienced IPV are more likely to enter into dating situations where they have fewer interpersonal tools to manage (i.e., women with histories of IPV may be less equipped to set limits or boundaries when aggressive men become controlling, for instance).

In contrast with the pattern we observed with respect to IPV victimization history, higher attachment avoidance was significantly associated with rating men as less aggressive, while higher attachment anxiety was significantly associated with rating men as more aggressive. This aligns with theory, as high avoidance people are thought to have less trust in others (Mikulincer & Shaver, 2013), and may not be able to recognize higher levels of aggression if everyone seems equally aggressive. Additionally, avoidant individuals may pay less attention to important character traits that signify aggression due to their disinterest in others. In terms of romantic relationships, such findings suggest that avoidantly attached individuals may be at risk for IPV victimization due in part to their inability to accurately pick up on cues related to aggression. This may be heightened in online dating settings where avoidant individuals tend to view everyone as equally untrustworthy (Chin et al., 2019), and thus, may fail to parse out differences in aggression levels. Importantly, viewing men as less aggressive could result in females high in avoidance getting into dating situations unprepared and being more likely to be exposed to IPV, which could result in higher levels of suspiciousness and distrust about potential partners in the future. One prior study of unknown partners found that college students high in avoidance reported lower levels of confidence regarding their perceptions of others (Borelli et al., 2019) – this finding leads us to wonder whether high avoidance participants would show some malleability in their perceptions. Perhaps providing females high in avoidance with feedback about the aggressiveness of potential dating partners could be useful, while working to ensure that this feedback does not increase avoidant behavior of all partners, but rather serves to improve accuracy in perceptions.

By comparison, people high in attachment anxiety may perceive others as more threatening and/or have a difficult time trusting their perceptions of men as safe or having good will (Mikulincer & Shaver, 2013). This finding is consistent with the notion that attachment anxiety involves perceiving threat more readily in the environment; in this study, females higher in attachment anxiety perceived higher levels of threat (in the form of aggression) within the profiles. While this may be viewed as a positive in terms of risk assessment, one important caveat is that a hallmark of anxious attachment is a desire for closeness and connection. Thus, in romantic relationships this tendency to see others as threatening must be balanced by the individual’s desire for intimacy. In the case of high anxiety females, this may result in individuals ignoring warning signs in favor of the romantic commitment (Mikulincer & Shaver, 2013). In other words, just because high anxiety females more readily perceive aggression (as we demonstrated using the standardized stimuli), it does not mean this would prevent them from pursuing a relationship. The drive they experience for connection may be so strong that they may overlook the warning signs they perceive in favor of pursuing a relationship. We did not measure desire to pursue a relationship with the potential dating partners in this study; this is an important question for future research.

In terms of individual characteristics influencing participants’ rating of men’s IPV perpetration risk, our results revealed that neither IPV victimization history, nor attachment orientation, significantly predicted participants’ rating of men’s IPV perpetration risk level or the variability in their ratings of men’s IPV perpetration risk level. In other words, participants’ ratings of men’s IPV perpetration risk varied along with men’s self-ratings but not as a function of their own IPV victimization history or attachment orientation. In fact, only self-identifying as African American (n = 23) was significantly associated with participants’ IPV perpetration risk ratings, as they tended to rate men as more dangerous. We interpret this finding cautiously given the small sample, the low base rates of IPV perpetration overall, and the fact that this finding was unanticipated. However, if it is a true effect, it may allude to the broader contexts in which African American individuals are exposed to violence in the United States, specifically the increased rates of IPV and community violence exposure (Cho, 2011), which may influence the ways in which they perceive others as being dangerous or threatening. Future research is needed to replicate these findings and determine whether these individual differences influence the perception of IPV perpetration risk in unknown individuals. Such findings have the potential to inform prevention efforts and extend our understanding of trends within IPV victimization on college campuses.

Finally, we theorized that participant characteristics (IPV victimization history, attachment orientation) would be associated with participant accuracy in rating men’s aggression/IPV perpetration risk. This prediction was somewhat supported. Specifically, for aggression, higher attachment anxiety was associated with overestimating men’s aggression, while higher attachment avoidance was associated with underestimating men’s aggression. This trend, while unexpected, makes theoretical sense when considering that attachment anxiety often materializes in a person through fears about a partner’s commitment and devotion (Campbell & Stanton, 2019). This fixation on one’s partner and the fear of them leaving may lead individuals high in attachment anxiety to be more perceptive of cues that signify tension in another person, such as a tendency for aggression or a feeling of uncertainty, therefore increasing perception. People higher in attachment avoidance, on the other hand, limit their access to attachment-relevant information, such as desires for closeness or intimacy (Campbell & Stanton, 2019; Tucker & Anders, 1999), and seek out opportunities for isolation in romantic relationships, all of which may impede their ability to accurately assess the level of aggression in a potential partner. Past studies demonstrate that attachment avoidance is associated with less confidence in naive observers’ rating of unknown individuals (Borelli et al., 2019), which, when combined with our finding of less accuracy, may signify a link between attachment avoidance and deficiencies in an individual's ability to detect other individual’s characteristics. When it comes to identifying men’s aggression, this may be particularly true as avoidant individuals may be less skilled in discerning between a man’s emotional states and personality dispositions that may be indicative of aggressive actions or behaviors (Mikulincer & Shaver, 2007, 2013). Such findings highlight the importance of educating college students on unhealthy relationships and warning signs, as well as how to best respond to such behaviors.

In terms of participant characteristics predicting accuracy in perceiving IPV perpetration risk, findings revealed that neither IPV victimization history nor attachment orientation was associated with participant accuracy of their perception of men’s IPV perpetration risk level. As with the absolute magnitude of IPV perpetration risk ratings, when predicting the accuracy of these ratings, only participant race/ethnicity was significantly associated with accuracy of men’s IPV perpetration risk: self-identifying African American judges were revealed to overestimate men’s IPV perpetration risk. This is particularly interesting in our sample, as African American judges reported the least amount of IPV victimization yet they demonstrated the most caution when evaluating others. This tendency to overestimate threat may be a byproduct of lived experiences; in the United States African American women are frequently exposed to racism, sexism, and discrimination (Cho, 2011), which may result in an expectation that people, particularly unknown men, will be untrustworthy or unkind. This finding provides additional evidence that individual characteristics can influence accuracy in person perception research and calls for additional empirical investigations to examine how individual differences may inform the ways in which naive observers perceive IPV perpetration risk in unknown individuals, information which holds important implications for abuse research and may illuminate processes which increase or decrease the likelihood of certain individuals’ being able to identify a potentially abusive partner. These findings provide important insight into female college students’ views of potential dating partners and highlights the importance of educating individuals on warning signs of abuse.

Strengths and Limitations

To our knowledge, this is the first study examining whether college women can accurately identify a target's propensity for aggression or IPV perpetration based on viewing a dating profile, as well as whether accuracy in these observations vary as a function of participant IPV victimization history or attachment orientation. These novel findings have relevance for our understanding of romantic partner selection among college students, particularly our understanding of the selection of abusive partners. Our female participant sample was large and ethnically diverse. Limitations include the fact that our analyses were cross-sectional, largely exploratory, and the study was not pre-registered. Though guided by theory, the findings should be considered tentative until replicated. Second, our participants were entirely female and heterosexual, thus the extent to which these findings are generalizable to men/non-binary individuals/and members of the LGBTQIA + community remains unclear. Our college student population did not experience as much adversity as women residing in a domestic violence shelter, so it is unknown how these findings would apply to people in more acute contexts. Third, in order to mask the purpose of the investigation (and reduce participant reactivity), aggression and IPV perpetration risk were assessed via different measures which required standardization of scores for comparison rather than the utilization of raw scores. Further, we measured accuracy using self-other discrepancies, which is useful in terms of interpretability and is the accepted metric in this field, but carries some limitations due to self-ratings often carrying bias and potentially leading to an underestimation of the true effects.

Additionally, although we recruited college males to create dating profiles because of our interest in studying college females and the people they are most likely to date (Kuperberg & Padgett, 2016), the males we recruited did not report high levels of aggression or IPV perpetration history, a trend in the literature (Kolivas & Gross, 2007), reducing the range in the sample. In future work it may be worth increasing the sample of males in the hopes of increasing the variability in aggression/IPV perpetration scores or recruiting a non-college student sample to increase the range, although this may not be as generalizable. Further, we developed profiles in as true-to-life a manner as possible while still protecting the men’s privacy – this entailed using memes or icons that the men selected to represent their personalities and characteristics that they wanted represented rather than pictures. The inclusion of actual photos chosen by the men could have introduced additional sources of noise, such as information about race, physical attractiveness, and facial expressions, which may or may not be related to the constructs of interest in the study. Using standardized pictures in the profiles would remove the personalization of the picture because they would not be chosen by the person creating the profile. Yet, pictures represent an important part of a dating profile that reveal crucial information about the person in the profile. It will be important in future work to assess participants’ perceptions of the icons/memes and what they glean from them, as well as to continue to consider the best way of handling this aspect of the dating profile.

Finally, to reduce participant fatigue and ensure good quality data, we only asked participants to view a limited number of men’s dating profiles (N = 9); however, this reduced our sample of profiles, and could have limited the variability and the external validity of the findings, necessitating the need for replication. Future research should include a larger sample of profiles with more varied rates of aggression and IPV perpetration, as well as information on whether the female participant has a history of IPV perpetration or mental health diagnoses, to better elucidate the relationship between participant accuracy and participant characteristics in perceiving men’s aggression and IPV perpetration risk. In addition, future research should investigate participants’ perceptions of the information they use to develop judgments about the men, as well as the extent to which they feel their romantic relationship history influences this person perception process. Such findings have broad relevance for our understanding of romantic partner selection for college students and may inform intervention efforts aimed at educating college students on safe dating practices and warning signs of abuse, such as signs of aggression.

Conclusion

With a high degree of accuracy, female college students can discriminate between men's propensity for aggression after viewing an online dating profile. Participants higher in attachment anxiety were more likely to overestimate their assessment of men’s aggression, while participants higher in attachment avoidance were more likely to underestimate their assessments. On the other hand, IPV victimization history was unrelated to accuracy in assessment of aggression or IPV perpetration risk, perhaps illustrating a potential area of risk for revictimization. Replication and extension will elucidate the processes by which individuals perceive and assess aggression and dangerousness in potential romantic partners.