A series of recent events has raised concerns concerning the increasing levels of political violence within the United States. These episodes include the assassination of political activist Charlie Kirk on September 10, 2025; The Assassination of a Democratic Minnesota state representative and her husband in June 2025; and two attempts to assassinate Donald Trump through the 2024 presidential campaign.
Some polls have shown that giant numbers of Americans are willing to support it Use of violence for political purposesor they consider that political violence can sometimes be justified.
My research is in Political science and data evaluation. I even have been conducting surveys for nearly 25 years. Over the last three years, I even have been exploring recent techniques that use artificial intelligence to conduct and analyze interviews.
My own recent surveys, using AI to ask people why they offer the answers they do, show that the surprisingly high levels of support for answering these questions are likely on account of confusion about what these questions are asking, quite than actual support for political violence.
Stephen Maturen/Getty Images
A communication error
Why do several surveys get incorrect answers to this vital query? I consider the cause is an issue called response error. This implies that respondents don’t interpret a matter the way in which the researcher thinks.
As a result, people's answers don’t truly reflect what the researcher sees within the answers.
For example, if one asks whether someone would support using violence to attain a political goal, the query arises as to what the respondent understands by “use of violence” on this context. It could possibly be interpreted as violence, nevertheless it may be interpreted as using legal means to “force” someone to do something.
Such response errors are a cause for concern for pollsters since survey research began. They can influence even seemingly easy questions.
What did you mean by that?
To avoid this problem I used one AI interview system developed by Cloud Researcha widely known survey research company, asked respondents the identical questions on political violence as in previous surveys. Then I asked them what they thought after they answered these questions. This process known as cognitive interviewing.
I then looked through and categorized these interviews using AI. Two transient reports that summarize this process as applied to each surveys are: available on-line. These analyzes haven’t been peer reviewed and the outcomes must be considered very preliminary.
Nevertheless, the outcomes clearly show that respondents interpret these questions very otherwise.
Nuances are vital
For example, in my poll, about 33% of Democrats agreed with the statement that “using force to remove President Trump from office is justified.” However, when asked why they agreed, greater than 57% gave answers like this: “I assumed not physically, but more within the sense that he – the president – may need to be 'fired' or forced out of office due to rules or laws.” Still others imagined future scenarios by which a president illegally seized power through a coup.
Taking these different interpretations of the query under consideration, the AI coded only about 8% of Democrats as supporting using force in the shape of violence under current conditions.
Even here there was considerable ambiguity – for instance, such a response was common: “The phrase 'use of force' was a bit too broad for me. For example, I couldn't justify killing Trump, but less extreme uses of force were justified in my eyes.”
Likewise, 29% of Republicans agreed that “using the military to stop protests against President Trump’s agenda is justified.” However, just about all respondents who agreed with this statement imagined that the National Guard would intervene nonviolently to stop violent protests and unrest. Only about 2.6% of Republicans made comments supporting using the military against nonviolent protests.
Almost everyone who agreed that using the military was justified expressed thoughts like this: “I see the military coming and acting like a police force to stop or prevent the demonstrations from becoming violent. Peaceful protesters should be allowed to exercise their right to freedom of expression.”
Michael M. Santiago/Getty Images
When is political violence justified?
Even questions that explicitly ask about political violence allow for broad interpretation. Take, for instance, this query: “Do you think it’s ever justified for residents to make use of violence to attain political goals?”
The lack of a selected scenario or location on this query invites respondents to interact in all styles of philosophical and historical speculation.
In my survey, almost 15% of respondents said violence can sometimes be justified. When asked concerning the examples they considered, respondents cited the American Revolution, the anti-Nazi French resistance, and plenty of other incidents as reasons for his or her answers. Only about 3% of respondents said they were currently considering motion within the United States.
Furthermore, just about all respondents said that violence must be the last resort when all other peaceful and legal methods fail.
One respondent illustrated each problems with one sentence: “The (American) colonists first tried petitions and negotiations, but when these efforts failed, they resorted to armed conflict to realize independence.”
A call for understanding
Even these numbers likely overestimate Americans' support for political violence. I read the interviews, checked the AI system's labeling, and concluded that it tended to overestimate support for violence.
Other aspects may distort reports of public support for political violence. Many surveys are conducted primarily online. One study estimated that between 4 and seven% of respondents to online surveys “fallacious respondents“, which select arbitrary answers. Another study reported that such respondents dramatically increase positive responses to questions on political violence.
Respondents may additionally be willing to anonymously express attitudes online that they might never say or do in real life. Studies suggest that “Online disinhibition effects” or “Survey trolling“may affect survey results.
In summary, my preliminary research suggests that response bias is a big problem in political violence surveys.
Americans are almost unanimous of their condemnation of the recent political violence they’ve witnessed. Recent poll results show that the alternative is due more to confusion about what the questions are asking than actual support for political violence.
