I Invited My Haters to a Live Q&A. 111 Wrote the Worst Things. Two Showed Up — Both in My Defense.

After the hate wave I wrote about two weeks ago, I did something several friends advised against: I invited my haters to a live Q&A to ask me questions directly. Under the invitation itself, the hate came back — indistinguishable from the original video. Not one of the authors of the threats and insults showed up to the stream. The people who did show up were the ones defending me. ZenFeed had hidden some of their earlier comments too — and I think that was the right call.

The idea

The October 1 video drew nearly 5,000 comments. 56% of them were hidden or deleted because of their high toxicity — including 186 criminal threats. Instead of shrugging it off as “that’s just the internet,” I proposed something concrete: a live Q&A where I’d answer questions — including from the people who’d written the worst things. I announced it on October 10 on Facebook and Instagram, with a link to a YouTube stream scheduled for Monday at 20:00. Right before the stream, I posted a reminder.

The announcement drew as much hate as the original

The Facebook post announcing the stream collected 68 comments with an average toxicity of 52.6/100. 57% of them were hidden or deleted because of their high toxicity. That’s practically the same as the original video on the same platform (62.7/100 and 61% hidden or deleted).

In the stream’s chat, average toxicity dropped to 8.7, and ZenFeed didn’t delete a single comment. The simplest explanation: the people who’d written the worst things knew I’d answer them live, so they wrote more politely. I checked this account by account. The explanation didn’t hold up.

I checked who showed up

I took every YouTube commenter from before the stream whose comment the system scored at 80 toxicity points or higher. In that band there’s no ordinary criticism left — only threats and insults. A few examples (profanity redacted):

“shame he didn’t run you over.” — 95 pts

“drive him somewhere out of the way and f***ing wreck him so his own family wouldn’t recognize him, he’d never climb on anyone’s car again” — 95 pts

“dude, you’re not the police. (…) the driver should’ve had the right to kick the st out of you, you f*ing a**hole” — 98 pts

“if you’d jumped on my hood like that, it would’ve been your last jump” — 95 pts

There were 111 such accounts. I checked how many of them wrote anything at all in the stream’s chat.

More broadly: of the 62 people who wrote in the chat, 8 had previously commented on the YouTube videos about this incident. Only 2 of them were among the 111 authors of the worst comments.

The chat wasn’t perfectly clean, either. 7 of 158 messages crossed 70 points — mostly one person’s tangents about Ukrainians and “Polish drivers” — but none of them were about me, and none crossed 80.

The defenders showed up. Their comments had disappeared too

Four people who had written sharply (70–88 points) under the videos about this incident also came to the chat. None of them attacked me — they went after the driver, his defenders, or institutions they felt had failed:

• Under the original video: “the cyclist told him off, and rightly so. (…) in a case of brazenly driving into a pedestrian, the vehicle should be forfeited to the state immediately” (70 pts) and “all you need is the murder weapon ;)” (85 pts). In the chat: “We’ve got exceptionally brazen drivers — they need civilizing.” (10 pts).
• Under the original video, about the driver: “What do you have to have in your head? EMPTINESS :)” (75 pts), and about his defenders: “the car guys proudly rallied to help the road pirate” (75 pts). In the chat: “good luck, champ” (0 pts).
• Under the original video: ”(…) they should confiscate the car (…) take away the license and lock him up for a year” (75 pts). In the chat: “I believe you.” (0 pts).
• Under the second video about the incident: “More like ‘they were told off’ — standard practice at that pathological institution.” (88 pts). In the chat: 👍👍👍👍👍👍👍 (0 pts).

Three of those comments — scored 75, 85 and 88 points — were hidden automatically by ZenFeed. Under my own videos, the system hid comments from people who were standing up for me.

This isn’t an anecdote from one page. Researchers who analyzed more than a hundred hate communities on Reddit saw the same thing: comments defending the people under attack were less toxic than the hate speech itself, but almost twice as toxic as the rest of the discussion (Hickey et al., 2026).

The system doesn’t know whose side you’re on. Good

A toxicity score is about language, not intent. The classifier can’t tell “confiscate the car” written in the author’s defense from the same sentence written against him. It sees an aggressive comment — and that’s it.

You could call that a flaw. We think it’s the right design, for three reasons.

Aggression escalates no matter whose side it’s on. A call to confiscate a driver’s car or jail him is, for the other side of the argument, an invitation to retaliate. Under a post with over a million views, an exchange like that quickly turns into what I saw on October 1.

Moderation that tells “ours” from “theirs” stops being moderation. It becomes censorship of one side. A page owner shouldn’t be the judge in their own case, and a system that lets brutal language through just because it supports the owner would make them exactly that judge.

Anyone who wants to defend me can do it without aggression. The best proof came from those same four people: in the stream’s chat they said the same things — calmly, without threats against the driver — and lost nothing in the process.

Haters avoid contact

This is a single case, not a study on a representative sample. Still, the pattern is hard to miss. Under a post where you can write anything and vanish, 111 people wrote threats and insults. On a stream where the other side was a person who could answer within the same minute, not one of them showed up.

And it wasn’t even a confrontation. Nobody had to show their face or sign their name — all it took was writing in a chat I was reading live. Minimal contact, still through a screen. That was enough for them to disappear.

This fits an experiment by Lapidot-Lefler and Barak (2012). They tested what drives aggression in online conversations most: anonymity, invisibility, or lack of eye contact. Lack of eye contact came out on top — ahead of anonymity.

That doesn’t mean the internet turns people into aggressors. Across eight studies in several countries, Bor and Petersen (2022) showed that people who are hostile online are about as hostile offline — online, they’re just more visible. My data doesn’t contradict that: the haters didn’t soften up on the stream. They simply didn’t come.

I think that’s good news for anyone hit by a wave of hate: the people who write the worst things are rarely willing to repeat them to someone who’s watching. They don’t soften on contact — they just avoid it.

Sources

1. Bor, A., & Petersen, M. B. (2022). The psychology of online political hostility: A comprehensive, cross-national test of the mismatch hypothesis. American Political Science Review, 116(1), 1–18.
2. Hickey, D., Schmitz, M., Fessler, D. M. T., Smaldino, P. E., Lerman, K., Murić, G., & Burghardt, K. (2026). Assessing how hate, counterspeech, and toxicity affect hate group newcomers. Proceedings of the International AAAI Conference on Web and Social Media, 20(1), 1007–1026.
3. Lapidot-Lefler, N., & Barak, A. (2012). Effects of anonymity, invisibility, and lack of eye-contact on toxic online disinhibition. Computers in Human Behavior, 28(2), 434–443.