AI & Society
Hate messages without toxic language strengthen audience identification with the author and increase support for harsh anti-immigrant policies
In a 2026 study published in Social Sciences, a peer-reviewed journal, 339 people viewed first-person hate threads on X that targeted immigrants. Hate messages that lacked toxic language created a stronger sense of oneness with the writer. That stronger connection predicted a greater intention to share the message, more negative attitudes toward immigrants, and more support for harsh policies against irregular immigration. Among participants with conservative political views, the non-toxic hate messages also increased narrative transportation, the feeling of being absorbed in a story, and that feeling further raised their intention to share.
What did the study test?
Researchers at Universidad de Salamanca and Universidad de Valladolid ran an online experiment with a 2 × 2 between-subjects design. Participants saw first-person hate messages on X that targeted immigrants. The study varied two things: whether the hate message used toxic language and how popular the message appeared to be. The researchers defined hate speech as content that promotes hatred based on personal or group characteristics.
What did the researchers find?
Hate messages without toxic language produced greater identity fusion with the author, a strong sense of oneness with the person behind the message. That sense of oneness predicted a stronger intention to share the message, more negative attitudes toward immigrants, and stronger support for harsh policies against irregular immigration. Among participants with conservative political views, the non-toxic hate messages increased narrative transportation, and that feeling made those participants even more likely to say they would share the message.
Why does it matter?
Subtle hate can bind an audience to a hostile view without using obviously abusive language. The authors argue that anti-hate campaigns should address both openly toxic and subtler hate content.
Did the study measure actual sharing?
No. The experiment measured participants’ stated intention to share the message, not real sharing on X.
Were all participants affected the same way?
No. The extra effect of non-toxic messages on narrative transportation appeared only among participants with conservative political views. The abstract does not say whether the increases in identity fusion, negative attitudes, and support for harsh policies appeared across the sample.
Limits and what's next
This was one online experiment with 339 people and a 2 × 2 between-subjects design. The participants viewed first-person hate threads on X and reported their intentions rather than their real-world behavior. The abstract does not say how the researchers recruited the sample or how closely the sample represents social media users. It also does not report an independent replication. The abstract does not state further limitations or next steps beyond the authors’ call for anti-hate campaigns to address both overt and subtle hate content.
Source: Juan-José Igartua and Carlos A. Ballesteros-Herencia, “Hate Speech on Social Media: Unpacking How Toxic Language Fuels Anti-Immigrant Hostility,” Social Sciences, 2026.