Permission Structures for Political Murder

Botification, Part III

In this post, I summarize the findings from the second study reported in the paper shown below, and which is more or less in press.

In two prior posts, I first presented an overview of the working theory (below is the link, its not just an image).

And then presented the results from the first study (below is also a link, not just an image):

The Theory, in Brief (but you should read the first article shown above)

The central idea is that social media use lends itself to botification, a term we invented and which refers to a process by which people’s social media interactions, and, we argue, their underlying psychology while using social media, become more bot-like (bots are programs designed to simulate real people online, often to promote simplistic narratives, propaganda and to sow discord). We proposed that botification represents a cognitive state in which individuals rely heavily on mental shortcuts and heuristic-driven behaviors, mirroring the reactive and automated processes of a computer program or algorithmic bots. This includes sometimes enthusiastically or even gleefully promoting the erosion of trust in civic and democratic practices and institutions.

Study One: Civic Disalignment, in Brief (but you should read the second article shown above)

Our survey of over 1900 Americans found that:

1. People who support democracy in general (e.g., supporting the right of every citizen to vote) often did not support it in specific instances (e.g., by supporting not allowing extremists to vote).

2. Social media use correlated with lower support for democracy in specific instances.

Study Two: Digital Permission Structures for Political Murder

Key Findings: Social media use and a slew of attitudinal and psychological variables correlated with endorsement of murder to achieve political ends.

The Gates of Hell, Rodin, via Wikimedia

Background: Digital Permissive Structures for Political Violence

We explored how manifestations of botification can play some role in what we term Digital Permission Structures for Political Violence: online mechanisms, ranging from viral memes to subcultural mythologies, that can normalize, romanticize, and incentivize political violence. These structures could enable users not merely to disengage from civic responsibility but to view endorsement of violence as a morally coherent and socially rewarded form of activism. After Luigi Mangione was arrested for the murder of UnitedHealthcare CEO, Brian Thompson, there was an explosion of support for Mangione on social media. These digital dynamics might produce a dangerous feedback loop in which glorification, endorsement and encouragement of political violence converge to erode moral and democratic norms and, perhaps, embolden real-world violence.

The tweets shown below appeared in my previous post as an example of civic disalignment; however, it is also an example of a digital permission structure for political violence. In the first tweet, there is a call to “Luigi Mangione” Elon Musk — i.e., to murder him. The second tweet presents his “justification.”

Figure 3 (from the paper) presents additionnal concrete examples found on social media of the glorification, endorsement, and encouragement of political violence. However, although these are vivid examples, they do not address the breadth or generality of such digital permission structures for political violence; doing so was one of the main purposes of Study Two.

The NCRI reports cited can be found here.

Study Two addressed the following research questions:

1. How many people endorse some justification for the murder of UnitedHealthcare CEO, Brian Thompson (support for political violence)?

2. Does social media use predict justification for this murder? A positive correlation is consistent with the digital permission structure hypothesis.

3. Is social media use correlated with

• authoritarianism?

• anxiety?

• zero sum thinking?

  • For these three questions positive correlations are consistent with botification.

• internal locus of control? For this question, a negative correlation is consistent with botification.

4. Does authoritarianism correlate with belief that the murder was justified? Positive correlations are consistent with botification.

The Survey

The Study Two survey examined how respondents perceived and morally evaluated the Mangione incident in particular and more generalized versions of it. For example, in addition to asking how justified Mangione was, another question asked about how justified it is to kill a CEO of a health insurance company that denied someone life-saving coverage (see Supplement for all questions). This study aimed to understand the extent to which social media use and psychological traits predict permissive attitudes toward violence, particularly in the context of anti-institutional narratives circulating online.

1,068 U.S. adults were recruited through Amazon Prime Panels. The sample was matched to U.S. Census benchmarks for gender, age, race/ethnicity, political affiliation, and region. All participants completed an online survey in the two weeks following Mangione’s arrest in December 2024, when public discussion and social media activity surrounding the case were at their peak.

Measures

All measures are reported in full in the supplement. The survey included questions on:

1. Time spent on social media use.

2. Right-Wing Authoritarianism and Left-Wing Authoritarianism.

3. Psychological traits:

   ○ Anxiety

   ○ Locus of control

   ○ Zero-sum thinking

   ● Familiarity with the murder of UCH CEO Thompson.

   ● Moral permissiveness regarding murder. This was the sum of four questions assessing respondents’ beliefs about justifications for this murder.

Results

How Many People Considered the Murder of the UnitedHealthcare CEO Justified?

This analysis focused on responses to the first permissive murder question, which asked “How justified or not justified was the killing of the UnitedHealthcare CEO?” Those who had indicated that they had not heard about the murder were excluded (n=117), leaving 951 respondents who reported having heard at least something about it. Among those, we grouped responses based on those who said it was completely unjustified versus everyone else (i.e., those who reported believing there was at least some justification for the murder). We refer to this latter group below as those who reported believing the murder was “at least somewhat justified.”

43% indicated that they believed the murder was at least somewhat justified (Figure S1 presents the full distribution). Thus, a substantial proportion of our respondents failed to categorically reject an extreme form of political violence – murder – directed at a civilian who was neither a combatant nor government official and who had not been accused of breaking any law.

Botification and Digital Permission Structures for Political Murder

The next analysis addressed the digital permission structure question: Was social media use correlated with endorsing justification for this type of political murder? Table 3, below, presents all correlations. For this analysis, we included all participants and used the sum of the four questions comprising the Permissive Murder Scale (see Supplement. It was (r = .31, p < .001). The more time respondents spent on social media, the more strongly they endorsed justification for this type of murder.

Although not specifically predicted by the Botification Model, two additional correlations provide information about the profile of those more strongly endorsing the murder. Age was negatively correlated both with Permissive Murder (r = -.45, p < .001) and SMU (r=-.43, p<.001). It was disproportionately younger people on social media who more strongly justified the murder.

The next analyses addressed the research questions about aspects of botification. Social media use was positively correlated with general anxiety (r = .29, p < .001), zero-sum thinking (r = .28, p < .001), and left-wing authoritarianism (r = .35, p < .001), and negatively correlated with locus of control (r = -.20, p < .001). Respondents who spent more time on social media platforms had elevated anxiety, lower internal locus of control, more zero sum thinking and higher levels of left-wing authoritarianism (see Table 3), all components of the botification hypothesis. Right-wing authoritarianism was trivially correlated with social media use (r=.04).

The correlations of the psychological variables with Permissive Murder were mostly consistent with botification. Higher scores on the Permissive Murder scale were strongly positively associated with general anxiety (r = .35, p < .001), zero-sum thinking (r = .44, p < .001), and left-wing authoritarianism (r = .50, p < .001) and negatively correlated with internal locus of control (r = -.32, p < .001).

The one striking exception was for right-wing authoritarianism (RWA), which was negatively correlated with Permissive Murder (r=-.20, p<.001). People high in RWA scored lower on the Permissive Murder scale than did those low in RWA. Explaining this unexpected relationship will have to await future research.

Limitations and Things Unknown

Several substantial limitations temper the strength of the conclusions we can draw. The studies presented are entirely correlational and cross-sectional, meaning causal inferences are not justified. Perhaps social media usage contributes to botification, civic disalignment, and permissiveness toward political violence. Perhaps individuals with pre-existing authoritarian tendencies, generalized anxiety, or anti-institutional attitudes are disproportionately drawn to certain social media environments. Indeed, our Botification Model proposes that these social and psychological phenomena exist in a recursive feedback loop, so that it would not be surprising to discover that causality runs in multiple directions. Regardless, without longitudinal or experimental data, causal claims are speculative at best.

Second, we assessed beliefs and attitudes rather than behaviors. We are reasonably confident that precious few of the nearly 3000 respondents in our two surveys are likely to commit political murder or participate in overturning elections. Nonetheless, their expressed beliefs and attitudes in this survey reflect surprisingly high levels of cultural acceptance of political violence and anti-democratic behavior. Whether they personally engage in such acts may not be as relevant as whether they are contributing to a sociopolitical culture in which such acts are more likely to emerge. This possibility is consistent with pyramid models of political extremism and violence, which suggests that actual political violence, though usually perpetrated by only a few, is built upon a much larger base of ideologically-aligned supporters. The applicability of this perspective to the modern U.S. is an important question for future research.

Third, our aggregated measure of social media usage collapses across platforms with dramatically different affordances, user bases, and algorithms (e.g., TikTok, Twitter/X, Reddit, Instagram). Without disaggregating these effects, it is impossible to determine whether the observed patterns are driven by unique features of particular platforms or by general online engagement.

There are more limitations identified in the actual paper, but to prevent this post from running on and on and from getting too wonky, that will have to do for now.

Conclusion

Rather than trusting lawful justice to address societal grievances, many of our respondents appeared to be transferring moral legitimacy to ideologically aligned actors, even when those actors endorse political violence. In this context, trust is no longer grounded in shared civic norms or institutional legitimacy. Instead, it is increasingly contingent on ideological alignment, potentially creating dangerous conditions under which violence becomes endorsed and celebrated. This appears to be a path toward declining trust in democratic institutions.

---

Commenting

Before commenting, please review my commenting guidelines. They will prevent your comments from being deleted. Here are the core ideas:

1. Don’t attack or insult the author or other commenters.

2. Stay relevant to the post.

3. Keep it short.

4. Do not dominate a comment thread.

5. Do not mindread, its a loser’s game.

6. Don’t tell me how to run Unsafe Science or what to post. (Guest essays are welcome and inquiries about doing one should be submitted by email).