Hostile tweets linked to upbringing and legal troubles, study finds

A new study published in the Proceedings of the National Academy of Sciences sheds light on why some people seem more aggressive on social media. By linking official government records to the online activity of Danish Twitter users, the researchers found that individuals with numerous criminal verdicts, more time spent in foster care, better performance in primary school, and higher childhood socioeconomic status tended to be more hostile in their social media interactions. A good portion of this behavior appears to be connected to the fact that these traits also predict whether a person actively discusses politics online, which itself is tied to higher levels of hostility.

The researchers undertook this investigation because aggression on social media has grown into a prominent concern. While early hopes for online platforms suggested they might create a friendlier, more connected world, many users have encountered the opposite. They see arguments, intimidation, and insults directed at strangers in ways that can create anxiety and discourage future online engagement.

Past work has highlighted features of social media, such as relative anonymity or engagement-driven platform designs, that might inflame tensions. However, studies have also shown that the angriest interactions often come from a small fraction of users, hinting that stable, individual traits may be shaping who ends up launching hostile attacks online. The authors of the new study aimed to move beyond anecdotal observations or self-reported survey data by using broader, real-world administrative records and direct measurements of online behavior.

To carry out the study, the research team combined two kinds of data. First, they drew a large random sample of Danish adults, obtained their unique names, and gathered information from official national registries covering everything from criminal verdicts to childhood environments. Next, they searched for matching Twitter accounts by using each adult’s name to find a single, unique match on the platform. Only individuals who posted at least one tweet during the study period were included.

In total, over 1.3 million Danish-language tweets from 4,931 users were analyzed. To gauge hostility, the researchers used a computer-based approach that rates each tweet by measuring how similar its wording is to common expressions of hatred. Experts had previously validated this method, confirming it could reliably capture the tone of the tweets. Finally, they created a separate measure of how often a user posted about politics by applying a similar technique keyed to political themes.

The investigators then connected people’s levels of online hostility with childhood and adult characteristics drawn from the official records. One of these attributes was the number of criminal verdicts a person had received. The researchers reasoned that repeated violations of the law might point to a long-standing inclination toward aggressive or antisocial behavior. They also looked at whether individuals experienced potential disruptions in childhood, including divorce in the family and multiple changes of address before the age of eighteen.

Another factor was the amount of time, if any, spent in foster care. Since other research suggests that a harsher environment in early life can encourage aggression, the team used two indicators of background conditions: parents’ financial and occupational status, and how well participants performed in primary school. Finally, they examined the roles of age and sex, since past work has long noted that men and younger adults can show higher tendencies toward aggression.

Results revealed that people with more total criminal verdicts were more hostile on Twitter, and those who spent a greater portion of childhood in foster care tended to be more hostile as well. On the other hand, individuals with particularly strong childhood academic performance and higher socioeconomic family backgrounds also posted more hostile tweets. At first glance, one might assume that coming from a more resourceful family would reduce anger or aggression online.

Yet the data suggest that those who grew up with better resources and skills were also more likely to talk about politics on Twitter, and political discussions were associated with more heated and confrontational posts. Men tended to show higher hostility compared to women, which is consistent with traditional findings in both psychological and criminological research. Younger people showed a slight tendency toward more hostile content, but this was less evident once political engagement was taken into account.

This study stands out for bringing together direct observations of online behavior and real-world, individual-level traits spanning decades. It points to a complex interplay between life history, personal dispositions, and the broader social media environment. One of the most interesting discoveries is that individuals with a greater inclination toward political debate appear more prone to writing hostile tweets, potentially because contentious topics encourage stronger language.

And while factors such as criminal background align with higher aggression in digital settings, other traits often linked to social advantage—such as higher childhood socioeconomic status—also correlate with more hostile online behavior in large part due to increased political engagement. This challenges simple stereotypes of who is most likely to start heated exchanges.

“The key contribution of the findings is instead the demonstration that individual differences in the propensity to engage in online hostility is connected to broader offline dispositions including dispositions related to aggressiveness and, in particular, political engagement,” the researchers wrote.

The study does have limitations that are worth noting. The sample focuses on Danish Twitter users, raising questions about whether the same trends would appear in countries with different cultures or social structures. The measure of hostility relies on language patterns identified by a computer algorithm, which captures many shades of hateful or angry speech but may miss more subtle forms of rudeness, harassment, or intimidation.

Also, while the connection to official data avoids some forms of bias that come with self-reported surveys, there are still other factors that likely shape hostile behavior that administrative data cannot capture, such as personality traits not flagged by criminal records or unique experiences that do not show up in government statistics. Future research might examine which features of social media platforms amplify the voices of aggressive users, as well as how everyday users perceive these confrontational individuals.

The study, “The offline roots of online hostility: Adult and childhood administrative records correlate with individual-level hostility on Twitter,” was authored by Stig Hebbelstrup Rye Rasmussen, Alexander Bor, and Michael Bang Petersen.