Counterspeech Science How Replies Reduce Online Hate
Mapping a Million Posts to Find What Silences Hate
A research group has done something that had never been attempted at scale: it took more than one million social-media posts written over four years and asked, with mathematical precision, which replies actually reduce hostility — and which ones pour fuel on the fire. [1] The setting was Twitter, now known as X, between January 2015 and December 2018. The posts were not random. They were replies to roughly 130,000 messages published by German news organizations, journalists, bloggers and politicians. That window opened with what is often called the “migrant crisis” in Germany, when a policy change let more than one million refugees and asylum seekers apply to stay from inside the country rather than from the first European Union state they had entered. In other words, the researchers did not simulate a quarrel in a laboratory. They watched a real society argue with itself, in public, in real time, and they kept the receipts.
A co-leader of this work is Mirta Galesic, a psychologist who co-leads the Collective Minds research group at the Complexity Science Hub in Vienna. [1] Her team studies how groups of people change their beliefs, make decisions and organize themselves — and hate speech, together with the replies that push back against it, sits squarely inside that question. The replies have a name: counterspeech. When someone posts something hostile, other users can answer. The research question is not whether answering feels good. The question is which kind of answer works.
To get at that, the group moved beyond counting insults. In follow-up studies, they used large language models to measure how much hate a conversation actually contained. They then sorted the many ways posters responded to one another into categories and identified which strategies were most effective at countering hate.
The scale of the problem gives those measurements their urgency. Hate speech — discourse that attacks a person or a group on the basis of inherent characteristics or identity — is rising worldwide. The triggers are varied: elections, migrant crises, even the men’s football World Cup. One number captures how fast the tide can move. Posts hostile to people of South Asian origin in the United States rose by 240 percent on X between January 2024 and June 2026. [1] A single platform, a single target group, a span of two and a half years — the figure comes from the Center for the Study of Organized Hate. That is not a mood. That is a curve.
Uncertainty Is the Fuel, Not the Fire
Why does hate speech spread so readily? Galesic points to uncertainty as the raw material. Rising inequality, shifts in power, armed conflicts and institutions that no longer work as well as they once did all produce it. When that uncertainty accumulates, it is easily channelled into hatred toward particular groups of people.
Polarization makes the channel wider. In some countries, Galesic notes, polarization is quite prominent — especially what researchers call “affective polarization,” the tendency of individuals to dislike people who hold different political opinions from their own. The word “affective” is doing real work here. It signals that the divide is not about policy details. It is about feeling. People do not merely disagree with the other side. They dislike the other side. That is a different problem, and a harder one.
Technology has made the problem easier to manufacture. Galesic stresses how easy it now is to lead people into thinking the worst about others — especially by using AI tools to create fake photos and videos. Content generated by artificial-intelligence models and bot accounts, she says, is exacerbating divisive rhetoric. This is where the stakes rise above mere rudeness. A false image can be produced in seconds and believed in less. The old defense — seeing is believing — has been quietly retired.
From this, Galesic draws a conclusion that is as much civic as scientific: it is extremely important to study hate speech, and everyone should be able to recognize their own susceptibility to being manipulated. The first line of defense is not a platform policy or a government regulation. It is a person who knows they can be fooled.
There is a social dynamic underneath all of this that explains why online spaces turn so harsh so quickly. Humans are very social beings, Galesic observes. We align ourselves to our environment. The more something is discussed, the more extreme people’s opinions tend to become. That is why, she says, just showing up and saying ‘I disagree’ matters so much. Her practical advice is deliberately modest: if you see hate speech online, take it almost as your duty to disagree — you just say ‘I disagree’, without feeling pressured to have an elaborate argument. Then ask friends to join and say that they also disagree. And if you meet someone whose opinion differs from yours, do not cut them off; try to learn more about them and find what you share, because you will get through to them more easily afterwards.