Partisan Animosity as Blame: A Unifying and Generative Framework for Understanding and Transforming Affective Polarization in the Political Sphere

Blame for Mental Contents is a Typical Form of Blame

Some might find it peculiar to think of partisan animosity as blame because it seems that blame is usually triggered by actions (see, e.g., all the work cited below in the section The Psychology of Blame) whereas partisan animosity seems to be triggered most often by perceptions of outparty attitudes or beliefs or intentions. This is less peculiar than it might seem at first. This is so for several reasons. First, many instances of partisan animosity are about actions—such as voting for the outparty in elections, circulating a petition in support of curtailing abortion rights, or making harsh or dishonest statements about the inparty on social media. Second, to take things to a deeper theoretical level, even when blame appears to be triggered by an action, it is actually triggered by the (assumed or inferred) underlying attitudes, beliefs, and intentions behind those actions. This is because blame is deeply rooted in social cognition and theory of mind (Malle et al., 2014), as we infer the mental states that drive actions and blame to varying degrees based on those mental state inferences. For instance, if a man interrupts a woman during a conversation, we might form a strong attitude of blame toward him because we assume he believes that women are not worthy of respect. However, if we later learn that he is deaf in one ear and did not realize she was speaking, our attitude of blame is tempered. Thus, the blame is not determined by his act (which is the same before and after we learn that he is deaf in one ear), but rather by the mental contents that seem to be reflected in that action.

Third, and perhaps most importantly, blaming others for their attitudes, beliefs, and intentions—even when those mental contents do not lead to action—is commonplace in social life. Although experimental social psychologists such as ourselves have tended to focus on blame for specific actions, moral philosophers (see Smith, 2013) have argued that we routinely blame others for their mental states in everyday life: How could you want such a thing?! How could you believe such a thing?! How could you feel that way in this situation?! How could you feel nothing in this situation?! On this view, the focus on attitudes, intentions, and beliefs in partisan blame is quite normal: How could you want a world full of guns?! How could you want a world where the unborn are murdered?! How could you believe that poor people are just lazy?! How could you believe that opening the doors to millions of immigrants would be good for society?!.

Why Does Blame for Mental Contents Play Such a Dominant Role in Partisan Blame?

Although blame for mental contents is a normal manifestation of blame, it is nevertheless arguably true that blame for mental contents is more central to partisan blame than to blame in other contexts. That is, partisan blame seems to be overwhelmingly directed toward the presumed mental contents of outparty members. Why would this be?

One reason has to do with the nature of political life today. That is, the outparty contains tens of millions of individuals spread across vast distances, and we observe the concrete actions of very few of them. Indeed, given the increasing geographic sorting of partisans into politically homogeneous communities (Brown & Enos, 2021), citizens today are less likely than ever to encounter outparty members and to directly observe their behavior. What we do observe, instead, is evidence of their attitudes, beliefs, and intentions through social media posts, polling data, online videos, and so on. Politics is, above all else, a game of words: arguments, tweets, podcasts, and opinion pieces that continuously convey information about what partisans think, believe, and feel. Naturally, because these mental contents are most visible to us, they become the primary target of our attitude of blame in the political arena.

A second reason is that blame often serves to protect what one perceives as a morally good community or society. When individuals believe that the outparty’s agenda would erode the moral fabric of their community, they blame in order to reduce the likelihood that this agenda will prevail. Before such feared outcomes materialize, blame tends to focus on the outparty’s “worldview”—its attitudes, beliefs, and intentions—because these mental contents are seen as the engines that will eventually produce concretely harmful actions and policies.

Offense Severity

One powerful contributor to blame intensity is offense severity, which refers to the degree of perceived wrongness that is detected. For example, causing a victim to feel profoundly humiliated is more wrong than causing him to feel a twinge of embarrassment. Offense severity is tightly linked to the degree of harm experienced by victims. This is in line with the dyadic theory of morality, which posits that actions are viewed as wrong in proportion to their perceived harm (Gray et al., 2012; Schein & Gray, 2018). Indeed, a number of studies have found that the more harm a victim experiences, the more blameworthy the perpetrator is perceived to be (Boon & Sulky, 1997; Gray et al., 2012; Gray & Wegner, 2011; Lerner & Miller, 1978; Miller & Vidmar, 1981; Schein et al., 2016; Schein & Gray, 2018; Vidmar & Crinklaw, 1974; Wortman, 1976). Similar to Malle et al. (2014), we assume that perceptions of wrongness are generally rapid, automatic, or “intuitive” (Haidt, 2012). Our model also emphasizes, however, that thoughtful analysis or reappraisal of severity is possible (Was it really that bad? Did it cause as much harm as I initially thought?) and that this can change one’s attitude of blame.

The importance of offense severity can also be seen at the group-level that interests us. For example, Longo et al. (2014) found that Palestinians who perceived Israeli checkpoints in the West Bank as humiliating (severe offense) showed a stronger attitude of blame toward Israel than did Palestinians who perceived the checkpoints as creating uncertainty about travel times (moderate offense).

Intentionality and Reasons

Perceived intentionality and perceived reasons, which are integrally related, are also major contributors to blame intensity (Malle et al., 2014). Perceived intentionality refers to whether the actor engaged in the action knowing what the consequences would be and desiring those consequences (Malle & Knobe, 1997). Across cultures, intentional violations elicit more blame than unintentional violations (McNamara et al., 2019), even in scenarios where unintentional actions yield severe negative consequences (Cushman, 2008). This propensity for stronger blame responses toward intentional violations holds true across various moral domains (e.g., harm and purity; Kupfer et al., 2020). A closely related concept to intentionality is perceived foreseeability (Fincham & Jaspars, 1980), because people cannot have intended an outcome that they did not foresee. Yet, importantly, they can be blamed for unforeseen harm if a perceiver believes that they could have or should have foreseen it (Lagnado & Channon, 2008; Malle et al., 2014).

Once intentionality has been established, individuals may move forward with evaluating the reasons that motivated the transgressor’s intentional bad action (Malle et al., 2014). Perceived reasons are what a perceiver infers to be the particular intentions behind an action. Considering a perpetrator’s reasons is integrally tied to assessments of blameworthiness because reasons determine the meaning of actions (Binder, 2000; Scanlon, 2008), thereby revealing aspects of the violator’s character, such as their motives, beliefs, and attitudes (Malle, 2004; Stueber, 2009). Whether or not a given reason aggravates or mitigates blame depends on social norms (Alexander, 2009; Shaver, 1985), and characteristics pertaining to both the blamer and the blamed (Polman, Pettit, & Wiesenfeld, 2013; Riordan et al., 1983; Tetlock et al., 2007). Generally speaking, blame-aggravating reasons establish that the action was committed based on immoral motives (e.g., selfishness, callousness, racism), whereas blame-mitigating reasons establish that the action was committed based on moral motives (e.g., protecting others, creating fairness). As it so happens, even children demonstrate the capacity for considering reasons. For example, Darley et al. (1978) found that children as young as 5 years old were inclined to recommend less punishment for a child who had harmed another child if the act was perceived to have proceeded from an intention to protect others.

Although we generally think of intentions and reasons as existing in the minds of individuals, these concepts are sometimes relevant to group perception, too. This is particularly likely when a group is perceived as entitative (Newheiser et al., 2012; Sacchi et al., 2009) or when people simply assume that reasons of a few group members reflect the reasons of all or most group members (Allison & Messick, 1985)

Freedom of Action

Another major contributor to blame intensity is perceptions of the transgressor’s freedom of action, called controllability in Weiner’s classic work (Weiner, 2006). We change the name based on Gill and Cerce’s (2017) finding that lay perceivers use multiple concepts of controllability when assigning blame. Here, violators evoke a stronger attitude of blame to the extent they are perceived as doing wrong while having an unconstrained (free) ability to do right. In a classic study, Weiner and Kukla (1970) demonstrated the impact of perceived freedom of action in evaluations of student performance. They discovered that teachers were more critical of students who did not expend effort (although they could have) than of students with equally low exam scores but who had low IQs (who cannot choose to be smarter).

Freedom of action has been manipulated in a variety of ways, such as through narratives involving biological impairment and determinist arguments about free will. For example, Gill and Cerce (2017) exposed participants to (fictitious) biological information about double-murderer Robert Harris. The information explained that Harris had suffered brain damage from a traumatic birth. Relative to a control group, participants exposed to this information showed tempered blame reactions to Harris, and this effect was driven by the perception that Harris had limited freedom to act otherwise. Similarly, Shariff et al. (2014, Studies 2–3) presented people with essays or neuroscientific articles arguing against free will. Then, they had participants read a story about a violent offender and were asked to judge what his prison sentence should be after he underwent a nearly 100% effective rehabilitation program (this latter detail was introduced to isolate participants’ desire for retribution from their desire for offender transformation/deterrence). They found that anti-free will arguments reduced support for retributive punishment against the offender.

Whereas work in the preceding paragraph reduced perceived freedom of action by reducing the sense that the actor is making any choice at all (i.e., via determinism information), other work focuses on how perceived freedom of action can be diminished when the actor clearly is making a choice but some of the options come with such dire side effects that no reasonable person would select those options. In the extreme case, an actor has diminished freedom of action when he “chooses” to say something because there is a gun to his head. As a more realistic, everyday example, Vonasch et al. (2018) found that a CEO is rated as having greater freedom of action than is a subordinate worker. This is because the subordinate’s role in the organization requires that he “do what he is told” (which might not be identical to what he wants to do), whereas the CEO’s role empowers her to “do what she wants.” Indeed, a subordinate could face termination for failing to do what he is told, whereas such external constraints on action are less of a concern for the CEO. This means that the subordinate has less freedom of action than the CEO. Thus, here, perceived freedom of action depends on perceptions of whether the actor is “forced” to make certain choices because alternative choices would have dire side effects (e.g., unemployment).

People also perceive the freedom of action of groups, which, in turn, impacts their group-directed attitude of blame. For example, Lickel and colleagues (2003) found that the extent to which people perceived the parents and peer groups of the Columbine High School shooters as either freely encouraging or failing to prevent the shootings, the more they were blamed for the massacre. Furthermore, those who view criminal offenders in general as having ample freedom of action are significantly more likely to support punitive criminal justice policies (Gill et al., 2021; Study 5).

Control of Self-Formation

Control of self-formation, the perceived extent to which immoral actors are believed to have created their own immoral character, is another major contributor to blame intensity. Gill and Cerce (2017) have shown that diminished perceptions of control of self-formation decrease blame, and that this happens independent of perceptions of freedom of action. Thus, control of self-formation and freedom of action are two distinct notions of agent control. In their experiments, Gill and Cerce reduced control of self-formation perceptions through historicist narratives, story-like accounts of a person’s character development centered on the impact of formative life experiences. In an experiment involving an office bully named James, participants were presented with historicist narratives (or not) about James. One narrative described how James had been bullied by his father during his upbringing, an experience that contributed to James becoming a bully himself. In another experiment, the narrative detailed James’s over-indulgent parents, who played a role in his developing arrogance and a lack of respect for others. Across both narratives (i.e., abuse narrative, spoiled narrative), the historicist narratives reduced blame of James for his bullying behavior, and this effect was driven by diminished perceptions of control of self-formation. Interestingly, the historicist narratives did not reduce perceived freedom of action: Participants in the narrative and no narrative conditions strongly and equally believed that James could choose to refrain from bullying.

Gill et al. (2021) showed how historicist narratives and perceived control of self-formation also function at the group-level. Participants completed a measure of historicist beliefs regarding criminals as a collective, that is, the extent to which they believed criminals, in general, develop criminogenic traits based on unfortunate formative experiences. A test of mediation indicated that those who endorsed historicist beliefs regarding criminals as a collective also reported reduced perceptions of criminals’ control of self-formation, which, in turn, was associated with less support for harsh criminal justice policies.

Negative Dispositional Attributions

Another critical contributor to blame intensity is negative dispositional attributions. Dispositional attributions involve an inference that an actor’s perceived immorality is reflective of a broader bad character: She mistreats her co-worker because she is a callous person (as opposed to, say, because her co-worker betrayed her in a vicious way). This idea harkens back to the classic person versus situation dichotomy of attribution theory (Heider, 1958; Jones & Harris, 1967; Kelley, 1973; see Moskowitz & Gill, 2013, for a review).

One illustration of the role of dispositional attributions in blame comes from Witte et al. (2006), who examined blame of a husband who physically assaulted his wife. The researchers varied whether the victim had been verbally aggressive prior to the assault, and whether the perpetrator had a history of violence. Results revealed that perpetrator blame was heightened when the victim had not engaged in verbal aggression prior to the assault. Presumably, this stemmed from a dispositional attribution: A husband who attacks his peaceful wife is a callous monster. Additional results suggested that victim blaming was heightened when the victim had been verbally aggressive prior to the assault and the perpetrator had never harmed anyone before. Again, this seems to reflect the operation of a dispositional attribution: A wife whose verbal aggression elicits aggressive behavior from a never-before-aggressive husband must be very cruel.

Halperin’s (2008) research on the Israeli–Palestinian conflict has demonstrated the consequences of dispositional attributions at the group level. He found that the extent to which individuals perceived acts of terrorism as reflecting an evil outgroup character predicted hatred of the outgroup. These feelings of hatred, in turn, predicted support for political and social exclusion, and even violence, against the outgroup.

Counterfactual Thinking

When a negative outcome occurs, observers are likely to engage in counterfactual thinking: How might this negative outcome have been avoided? (Kahneman & Miller, 1986; Roese, 1997). In the context of blame, counterfactual thinking involves mental simulation of alternative worlds that undo a harm that was done in the real world. The counterfactual thinker attends to which aspects of the world, when varied, seem most likely to prevent the harm. Research suggests that such counterfactual thinking tends to focus on undoing controllable actions by human agents (If only he had not done X) or on undoing abnormal conditions (If only she had done it the way she normally does; Alicke et al., 2015). An attitude of blame toward an actor is intensified to the extent that the counterfactual suggests that the actor could have easily done something differently and that this would have greatly reduced the probability of the harm (e.g., If only he had checked the door before going to bed. . .; Alicke et al., 2015).

Spellman and Gilbert (2014) discuss findings from Nadler and McDonnell (2011) through this counterfactual lens. Nadler and McDonnell reported a study in which participants learned about an explosion in a man’s shed that killed one of his neighbors. An important contributor to the explosion was the oxygen tanks stored in his shed. Results revealed that the man was blamed more for the explosion when he stored the oxygen tanks for business purposes than when he stored them because his daughter required oxygen for a chronic respiratory disease. Spellman and Gilbert argue that this pattern happened because it is easy to imagine a world in which the businessman stores his oxygen tanks elsewhere (e.g., in a warehouse connected to his business), whereas it is difficult to imagine a world in which the father who is caring for his daughter stores her desperately needed oxygen tanks elsewhere. Another study linking counterfactual thinking to blame was conducted by Branscombe et al. (2003). Those authors found that victims of sexual assault suffered more self-blame to the extent that they generated counterfactual worlds in which their own choices could have prevented their victimization (e.g., I should’ve stayed home that night).

Informational, Cognitive, and Motivational Biases

Although some prominent theorists assert that blamers succumb to a variety of blame-aggravating cognitive and motivational biases when assessing blameworthiness (Alicke, 2000, 2008; Clark et al., 2014), others argue that blamers are generally fair and rational agents who process blame-relevant information in a balanced, even-handed manner (Malle et al., 2014; Monroe & Malle, 2019; Monroe & Ysidron, 2021). Indeed, these latter theorists argue that there are alternative explanations for studies supposedly demonstrating bias and that the alternative explanations point to rational information processing under impoverished information conditions (Malle et al., 2014).

One body of research that purports to demonstrate bias is work on how blame-related inferences are (supposedly) distorted by spontaneously activated negative affective reactions to perceived harm or to those who produce such harm. As an example, Alicke (1992) conducted an experiment in which participants read one of two scenarios involving a man who sped home during a rainstorm, resulting in an accident that harmed others. In one scenario, participants were informed that the man was speeding home to hide cocaine from his parents (a character who should spontaneously activate negative affect), whereas in the other, they were told he was rushing home to hide an anniversary gift for his parents (a character who should not spontaneously activate negative affect). Even though participants received identical information regarding the accident in both conditions (e.g., he was speeding, there was a rainstorm), participants viewed the cocaine user as more responsible for the accident than the anniversary gift giver. Alicke argued that this reflects a bias in which exaggerated blame is pinned to those who activate negative affect (the drug dealer). Malle et al. argue that it’s unclear whether this work demonstrates such an affect-driven bias. Rather, they suggest that participants might reasonably assume that a person who accidentally leaves cocaine on his parent’s table drives more recklessly than a person who remembers to get his parents an anniversary gift. If so, then it is rational and defensible to view the cocaine user as more responsible for the accident.

Another type of (supposed) bias centers on how blame-related cognitive appraisals are affected by pre-existing negative perceptions of the transgressor’s character [Note: In the preceding paragraph, Malle et al. (2014) interpreted Alicke’s (1992) data as reflecting this phenomenon rather than the affect-driven bias posited by Alicke.] Here, rather than blameworthy acts leading to perceptions of bad character, pre-existing perceptions of bad character lead to perceiving an actor as more blameworthy (Nadler & McDonnell, 2011; Pizarro & Tannenbaum, 2012). That is, pre-existing dispositional attributions (I know from past experience that she is selfish) can bias perceptions of the actor’s intentionality, reasons, and so on, for an act that is perceived later (This morning, I’m sure she intentionally left the clogged toilet for someone else to deal with). As an example, Nadler and McDonnell (2011) had participants read about a case where a woman's dogs killed a child after escaping from their owner’s fenced yard. The dog owner was viewed as more responsible for the child’s death when she was described as having an unlikable, disreputable personality than when she was described as having a charming, prosocial personality. Malle et al. (2014) argue that it is not clear that this should be called a “bias.” Rather, they suggest, as in the preceding paragraph, that participants might simply be making the defensible assumption that a “bad person” probably exerted less effort to ensure that her dogs were contained than did a “good person” (see Nadler, 2012 for a rebuttal). Whether or not one wishes to use the label “bias,” we think this phenomenon likely plays an important role in partisan contexts: If one already believes that the outparty is full of “bad people,” then the intensity of one’s blame attitude toward them for new offenses will be stronger than if one had a more favorable pre-existing impression of them. This will happen because their “wrongness” will be more likely to be perceived as intentional, done for bad reasons, and so on (i.e., cognitive appraisals of their blameworthiness are influenced by the negative character impression).

Gill and Ungson (2018) provided evidence for motivated bias in blame. In one study, they had participants read about double-murderer Robert Harris. To manipulate motivation, some participants were instructed to act as prosecuting attorneys and others as defense attorneys. As an additional manipulation, some participants learned only about Harris’s murders whereas others also read a historicist narrative about Harris’s brutal childhood. After receiving information about Harris, the “attorneys” were asked to present arguments that supported either harshness (for prosecutors) or mercy (for defenders). After writing out their arguments, participants were told that the role play was over. Next, they were given some filler questionnaires and then they were asked for their personal opinions about Robert Harris. Results indicated that, in the condition where Harris’ brutal childhood was described, those who had previously been defenders blamed Harris less than did those who had previously been prosecutors. Coding of the arguments participants made while enacting their role as either defender or prosecutor suggested that motivated self-persuasion had occurred: Former defenders blamed Harris less than former prosecutors to the extent that, during the time they were motivated to make arguments for leniency, they framed Harris’ brutal history as severely limiting his freedom of action.

Beliefs About the Efficacy of Blame Expression

People are sometimes motivated to down-regulate blame expression based on beliefs they have about the perceived efficacy of blame expression. That is, some people believe that outward expressions of angry condemnation will produce no benefit or might even make a situation worse (If I yell at my wife, she will lose respect for me and become more oppositional). Beliefs about the efficacy of blame expression—Is full-throated expression of blame likely to facilitate my goals (e.g., to persuade this person to be kinder)?—are related to what emotion regulation theorists label instrumental motives for emotion regulation (Geddes & Lindebaum, 2020).

Gill et al. (2025; Study 2) provided evidence relevant to this idea. They examined the impact of a blame efficacy workshop on blame expression among first-year college roommates. Roommate pairs were randomly assigned to an in-person workshop that described the inefficacy of excessive blame expression or to a control (no workshop) condition. In the workshop, roommate pairs learned about, for example, Durtschi et al.s’ (2011) longitudinal research showing that angry blame expression during couple conversations was associated with diminished marital quality over time. Participants also learned tools to communicate more effectively when their roommates behaved badly. A follow-up survey later in the semester showed that the workshops successfully reduced belief in the efficacy of harsh blame (e.g., “Harshly blaming my roommate is likely to improve his/her behavior”) and these reductions in blame efficacy beliefs were associated with reductions in expressed hostility toward one’s roommate (Alam, 2023).

Beliefs About the Normative Appropriateness of Blame Expression

People will also down-regulate blame expression based on perceived norms regarding appropriateness of anger expression. That is, some people believe that outward expressions of angry blame are immoral. Here, the belief is not that blame expression is counterproductive, but rather that it is morally wrong. Because people want to be moral (Prentice et al., 2018), the belief that blame expression is counter-normative will motivate down-regulation of blame expression. Beliefs about the appropriateness of anger expression are not merely an individual matter, as there are often broad cultural beliefs about such matters in organizations (Geddes & Callister, 2007) and in nations (Matsumoto et al., 2008). Beliefs about the appropriateness of anger expression—Is full-throated expression of blame consistent with my notion of what it means to be a good person?—are related to what emotion regulation theorists label eudaimonic motives for emotion regulation (Geddes & Lindebaum, 2020)

Standing to Blame

Standing to blame is a concept that has received abundant attention from philosophers (Bell, 2013; King, 2024), but only limited attention from social psychologists (Ghezae et al., 2025). One’s standing to blame is challenged whenever one is asked: Who are you to judge? When people recognize that they lack standing to blame, they are more likely to keep their hostile criticisms to themselves (i.e., down-regulate their blame expression).

Whether one has standing to blame has been argued to depend on a variety of factors (King, 2024). These include whether one has engaged in similar wrongdoing in the past (a dishonest politician lacks standing to blame another politician for dishonesty, although politicians seem uniquely unlikely to recognize this!), whether one has a stake in the other’s behavior (e.g., other people’s private lives are none of our business), and whether one is part of the same moral community as the target of one’s criticism (e.g., modern humans lack standing to blame humans who inhabited very different moral communities centuries ago). Ghezae et al. (2025) examined folk intuitions regarding the standing to blame and their results supported many of King’s theoretical suggestions. Presumably, recognition that one lacks standing to blame can lead one to refrain from expressing blame.

Interestingly, Bruneau and et al.’s (2018, 2020) collective blame hypocrisy intervention may have prompted participants to reflect on their standing to blame. In their 2018 study, White Americans were asked to what extent they blamed themselves and White people as a group for acts of White supremacist terrorism committed by individuals such as Dylan Roof, Anders Breivik, and Wade Page. Naturally, White Americans did not view themselves as responsible for the acts of these extremists. This exercise was found to decrease the level of blame White Americans assigned to individual Muslims and to Muslims as a whole for Islamist terrorist attacks. By having participants reflect on their potentially hypocritical stance in this context, they arguably prompted participants to reflect on their standing to blame (e.g., Who am I to blame all Muslims when people from my group have also engaged in horrific violence?).

Interventions that Target Offense Severity

From our review of the literature, we find no existing interventions that clearly target perceptions of offense severity. We find it surprising that this factor has not been studied in the context of partisan animosity, given that everyday political rhetoric seems highly focused on exaggerating perceptions of the amount of harm caused by outparty members (e.g., the outparty is destroying our nation, eviscerating our constitution, and allowing us to be overrun by immigrant invaders). We strongly encourage future research to test interventions to reduce partisan animosity via reducing perceptions of offense severity. In the proposed interventions section of this article (below), we will offer ideas for such interventions.

Interventions that Target Intentionality and Reasons

Some interventions lower animosity by changing perceptions of the intentions or reasons behind outparty policy stances. For example, Stanley et al. (2020) assigned participants to read about one of three issues (requiring standardized testing in primary schools, mandatory use of police-body cameras, animal testing for scientific purposes) and asked them whether they supported or opposed the issue stance. Then, they randomly assigned participants to one of three conditions: opposing reasons, supporting reasons, or control. In the opposing reasons condition, participants were provided with compelling reasons to believe what their opponents believe. For example, if a participant was against animal testing, they would learn that animal testing has been crucial in discovering many life-saving treatments for debilitating diseases. In the supporting reasons condition, participants were provided with reasons that supported their own stance (e.g., Alternative methods eliminate the need for animal testing). In the control condition, participants were not given any reasons. Results revealed that participants in the opposing reasons condition developed more positive views of the morality and intelligence of their opponents. Seeing that opponents adopt beliefs based on compelling reasons reduces blame for their possession of such “wrong” beliefs.

Similarly, Kubin et al. (2021) had liberals and conservatives read about how political opponents adopted their political beliefs based either on personal experiences of harm or on factual information. In one experiment, they had participants read about individuals who held opposing views to them on policies related to either coal, guns, or tax regulations. Some participants learned that the opponent held her opinion because of personal experiences of harm (e.g., coal regulations led her to lose her job) or because of factual information (i.e., information she learned while reading about coal regulations). The results revealed that when partisans were exposed to opponents who adopted political stances based on personal experiences of harm, they judged those opponents as more worthy of respect and more rational than opponents who relied only on facts. In short, partisans blame their opponents less when they come to see them as rational agents who are trying to avoid personal harm.

As another example, Waytz et al. (2014) randomly assigned Democrats and Republicans to one of two conditions: own-party or other-party. In the own-party condition, participants rated their party’s reasons for engaging in conflict with the opposition party, asking them how motivated their party is by love for their own party versus hate for the other party. In the other-party condition, participants rated the opposing party's reasons. Results showed that Democrats and Republicans believed that their own party was primarily motivated by ingroup love, whereas the outparty was motivated by outgroup hate. In a follow-up experiment, Waytz and colleagues randomly assigned participants to an incentive or control condition. In both conditions, participants were told that they would soon rate outgroup motives and that their responses would be compared to the actual responses of people in the opposing party. Those in the incentive condition were told that they could earn a bonus payment of $12 for accurately gauging the opposing party’s motivations. Results demonstrated that those in the incentive condition became more likely to rate the outparty as motivated by ingroup love. This change in perceived motives for outgroup actions was, in turn, associated with more favorable views of the outparty’s morality (i.e., reduced blame).

In short, a diverse set of interventions appears to reduce partisan blame by revealing that outparty members adopt their morally disagreeable beliefs and take their morally wrong actions for understandable, compelling reasons.

Interventions that Target Freedom of Action

To the best of our knowledge, the only exploration of the impact of freedom of action perceptions on partisan animosity is Severson’s (2020) unpublished manuscript. Severson randomly assigned participants to either a biological treatment group or a control group. In the biological treatment group, participants were exposed to an article summarizing research on the neurobiological and genetic underpinnings of political beliefs, while those in the control group read about sloths. The findings revealed that participants in the treatment group were less inclined to believe that political ideology was solely a matter of free choice, leaning more toward seeing a role of biological determinism. The results also showed that participants who strongly believed that ideology was a matter of free choice tended to hold more negative feelings toward political outgroups. Finally, there was an indirect effect of the intervention on negative feelings toward the outparty: Democrats and Republicans who came to see an impact of biological determinism on political ideology showed an associated tendency toward reduced blame of the outparty. This depolarizing effect was strongest among highly partisan individuals.

Interventions that Target Control of Self-formation

Research has shown that diminishment of perceived control of self-formation tempers partisan animosity. In one experiment, Gill et al. (2023) manipulated whether partisans learned the formative history of a political opponent (i.e., how she developed her political beliefs under the influence of family, community, and religion) or learned non-explanatory, neutral information about her (i.e., she’s from Denver, hates the cold winters). Then, participants read a harsh tweet she sent in which she criticized the participants’ political beliefs. Participants had the opportunity to tweet a reply, and this reply was coded for hostility. Results showed that when liberals and conservatives learned about the political upbringing of the harsh tweeter, their replies to her became less hostile. This effect was mediated by viewing the tweeter as less in control of developing her beliefs (i.e., reduced control of self-formation), and, therefore, less blameworthy. In follow-up experiments, Gill and colleagues replaced the individual narrative about the harsh tweeter and instead gave participants a historicist reminder—a message about how, as a general fact, formative experiences shape everyone’s political beliefs. They discovered that this general historicist reminder reduced the harshness of tweet responses among liberals but not conservatives, and this effect (for liberals) was, again, mediated by reductions in perceived control of self-formation and blame.

In a more recent paper, Alam and Gill (2024) extended this work to the context of changing attitudes toward the outparty as a whole (i.e., Democrats, Republicans). In their experiments, they randomly assigned partisans to a historicist thinking intervention or control. The intervention consisted of three elements: (a) a narrative detailing the development of a particular political opponent’s worldview (as in Gill et al., 2023), coupled with (b) a message emphasizing how individual life experiences influence everyone’s political beliefs (as in Gill et al., 2023), and (c) a suggestion that outparty members can change through future formative experiences. They found that for both Democrats and Republicans, the intervention reduced animosity toward the outparty as measured by the feeling thermometer. In follow-up experiments that sought to identify the particular blame emotions affected by their intervention, they found that for Democrats the intervention increased compassion for Republicans, while for Republicans, it reduced disgust, disapproval, and anger toward Democrats. For Democrats, the effects of the intervention were mediated by decreased perceptions of Republican control of self-formation across all experiments. For Republicans, however, the effects of the intervention on reduced animosity, while present, were not mediated by reductions in perceptions of control of self-formation.

Interventions That Target Dispositional Attributions

Other interventions have arguably targeted dispositional attributions. They do so by providing information to correct people’s negatively skewed views of outparty character. Such interventions are inspired by the fact that political partisans often hold exaggerated views of the immoral character of the opposing party based on inaccurate perceptions regarding the prevalence of extreme and objectionable beliefs among outparty members. For example, Democrats and Republicans overestimate the extent to which outparty members hold anti-democratic attitudes and are willing to use violence for political purposes (Mernyk et al., 2022; Pasek et al., 2022). Similarly, partisans overestimate how much outparty members dehumanize and dislike them (Landry et al., 2023; Moore-Berg et al., 2020). Together, these misperceptions give rise to negative dispositional attributions regarding the outparty: They are violent, callous, mean, and so on. In a truly vicious circle, these exaggerated perceptions of outparty wickedness predict reciprocal inparty animosity toward the outparty (e.g., Landry et al., 2023; Mernyk et al., 2022; Moore-Berg et al., 2020).

Many interventions aim to improve beliefs regarding outparty dispositions by correcting these underlying exaggerations and misperceptions. For example, Landry et al.’s (2023) intervention corrected exaggerated perceptions of meta-dehumanization by exposing participants to information on the actual (relatively low) extent to which the outparty dehumanizes the inparty. This correction reduced hostility toward outpartisans, an effect that remained evident one week later. In Voelkel et al. (2024) mega-study of 25 interventions to reduce partisan animosity, they included five interventions that corrected different types of misperceptions. These misperceptions included beliefs about the willingness of outpartisans to break democratic norms and to sacrifice lives to the pandemic for political gains. All five interventions significantly reduced partisan animosity, arguably by reducing negative dispositional attributions regarding the outparty’s supposed unfairness, heartlessness, Machiavellianism, and so on.

Moreover, the intervention with the largest effect size tested by Voelkel et al. (d = −0.53; 2024) was a video that featured pairs of opposing partisans discussing their political differences. The video featured three pairs of partisans, a leftist feminist and a rightwing antifeminist; an environmental activist and a climate change denier; and a transgender woman and a man who believed gender is binary. Prior to learning about their partner’s political views, the pairs did a variety of “ice breaker” and “team building” tasks with their partners. After these shared activities, the pairs viewed recordings of earlier interviews where their partner expressed their opposing political opinions. After this, they were offered the choice to discuss their differences over beer or to leave. All three pairs chose to stay and cordially discuss. Although it was not directly measured, and many variables were presumably manipulated by the video, we would wager that this intervention was effective at reducing partisan animosity, in part, because it reduced negative dispositional attributions regarding the outparty. By highlighting the presence of likeable outpartisans—people who were friendly and cooperative during the “ice breaker” tasks, people who were open-minded and willing to cheerfully chat over beer even after learning of deep political differences with their partners—the intervention challenged perceptions of outpartisans as dispositionally nasty and closed-minded.

Researchers have also examined a variety of interventions that teach individuals how to communicate constructively with outpartisans. We suggest that these interventions, too, are effective because they remove blame-aggravating dispositional attributions regarding the outparty. That is, those who learn to communicate positively—respectfully, open-mindedly—are seen as having positive dispositions, thereby reducing outpartisans’ animosity toward the communicator and her inparty. For example, Minson et al. (2024) trained liberals and conservatives to craft political persuasion messages for outpartisans that expressed “conversational receptiveness”, or a willingness to thoughtfully engage with opposing views. Those trained in this communication style received more positive evaluations from an outparty member and, more broadly, improved that outparty member’s view of the communicator’s inparty. Similarly, Puryear and Gray (2024) instructed partisans to use “balanced pragmatism”—considering multiple perspectives and seeking pragmatic solutions—when writing about their views on immigration. Compared to communicators instructed to focus on being logical and cohesive, those pursuing balanced pragmatism were judged by outparty members as more respectful and as more desirable as an interaction partner. This effect was mediated by perceiving these balanced pragmatists as being more rational and moral. Finally, Collins et al. (2022) found that partisans tend to underestimate how willing outpartisans are to learn about opposing views. When participants were informed that their ideological counterparts were indeed interested in understanding opposing viewpoints, they reported more favorable evaluations of outpartisans and expressed a greater willingness to engage with them in the future. By enabling partisans to demonstrate positive dispositions via positive communication strategies, these interventions reduced partisan animosity (see Chen et al., 2010; Hussein & Tormala 2021, 2024; Kalla & Broockman, 2023; Minson & Chen, 2022; Yeomans et al., 2020; for similar findings).

Interventions that Target Identity-Based Biases in Blaming

Above, we described how social identity motives can motivate overblaming of the outparty. One way to overcome such identity-based biases is to alter partisans’ sense that they belong to two discrete, mutually exclusive groups (i.e., reduce the “sharp sense of us-vs-them” described above). Indeed, researchers have found that highlighting commonalities between the parties can reduce partisan animosity. Wojcieszak and Warner (2020) had Democrats and Republicans engage in vicarious or imagined positive contact with the outparty. They found that both forms of contact reduced partisan animosity and did so by increasing perceived commonality—that is, shared identity—between oneself and the political outgroup. Similarly, Ahler and Sood (2018) corrected inaccurate perceptions of outparty demographic composition (e.g., percentage of Democrats who are part of the LGBTQ community; percentage of Republicans who are extremely wealthy) and found that this reduced partisan animosity. Though not explicitly measured, it seems likely that their intervention—by reducing how demographically different the outparty seemed—increased perceived commonality between the two parties. Finally, Voelkel et al.’s (2024) intervention tournament found that interventions that highlighted commonalities between Democrats and Republicans, such as their shared exhaustion from excessive political conflict or their common national identity as Americans, tended to be particularly effective for reducing animosity.

In short, increasing perceived commonality between political parties—challenging the idea that they are discrete, mutually exclusive entities—arguably reduces the social identity-based motivation to heap blame on the outparty, and thereby reduces partisan animosity.

Interventions That Target Beliefs About the Efficacy of Blame Expression

Some successful interventions appear to target beliefs about the efficacy of blame expression. Santos et al. (2022) tested the impact of increasing participants’ belief in the utility of having empathy for political opponents. In an experiment, they randomly assigned participants to one of two conditions: A high-utility condition where they read about how cross-partisan empathy can increase an individual’s political persuasiveness (e.g., treat your opponent with respect, and then they are more likely to compromise), or a low-utility condition where they read about how cross-partisan empathy can decrease an individual’s political persuasiveness (e.g., you will appear weak if you don’t stand up for your beliefs). The results showed an impact on blame expression: Participants in the high-empathy-utility condition utilized more conciliatory language when communicating with their political opponents, which, in turn, led to their opponents liking them more and judging their messages to be more persuasive. In short, when people are reminded that hostile blame expression is inefficacious—empathic communication is superior—their cross-partisan communication becomes kinder and partisan animosity is thereby reduced.

Interventions That Target Normative Beliefs About Blame Expression

Lastly, other successful interventions to reduce partisan animosity have targeted beliefs about the normative appropriateness of blame expression. For instance, Munger (2021) conducted a field experiment on social media, employing Twitter (now X) accounts to sanction individuals for publicly expressing harsh blame toward outpartisans. Initially, Munger created Twitter accounts in 2015 to simulate Democrat and Republican users. Subsequently, he scraped Twitter for hostile tweets from non-elite partisans directed at Donald Trump or Hillary Clinton prior to the 2016 U.S. presidential election. These hostile partisans were then randomly assigned sanction messages from in-party accounts created by Munger (e.g., Republicans who harshly tweeted at Hillary Clinton were sanctioned by one of Munger’s Republican accounts). Both Democrat and Republican users were assigned one of three sanctions: Care-based moral argument (“You shouldn’t use language like that . . . our opponents are real people with real feelings”), authority-based moral argument (“You shouldn’t use language like that. . .[we] need to behave according to proper rules of political civility”), or a non-moral public sanction (“Everyone can see that you tweeted this.”). Another group of participants received no sanctions. Results revealed that the sanctions with a moral basis (i.e., care- or authority-based) reduced political incivility across the first week post-treatment. That is, participants sent fewer uncivil tweets during the week after being sanctioned by morally grounded messages. We view Munger’s successful messages as signaling to partisans the normative inappropriateness of nasty blame expression, thereby reducing their willingness to publicly express partisan hostility.