Moderation & safety
Trolling
Posting deliberately provocative or disruptive content to bait emotional reactions, derail a conversation, or exhaust a community, rather than to contribute to it in good faith.
Trolling is the act of posting deliberately provocative or disruptive content in order to bait emotional reactions, derail a discussion, or wear down other participants. The defining trait is intent: a troll is not arguing in good faith but seeking a response for its own sake. That distinguishes it from a strongly held opinion or an unpopular but sincere argument.
The main types
Trolling takes several recognisable forms in a comment or debate space:
- Provocation. Posting inflammatory statements designed to offend or outrage, with no real interest in the topic.
- Flame war. Escalating personal attacks back and forth until a thread collapses into hostility and the original subject disappears.
- Derailing. Steering a conversation off its topic on purpose, often with bait, whataboutism, or repetition, so the discussion never settles.
Why it harms communities
A handful of trolls can shape the perceived tone of an entire space. Good-faith readers leave when threads feel hostile or pointless, which lowers participation and shifts the remaining audience toward the most combative voices. For a newsroom, that erodes the editorial value of the comment section and increases the moderation load, because each troll generates disproportionate reactions and reports.
How to counter it
No single lever solves trolling. Publishers combine several:
- Hybrid moderation. AI flags likely toxic or provocative content, and a human team arbitrates the ambiguous cases instead of relying on either alone.
- Argument quality scoring. Surfacing well-reasoned contributions and ranking low-effort provocation down reduces the reach a troll can reach for.
- Graduated bans. Escalating sanctions, rather than an immediate permanent ban, give repeat offenders a clear signal while keeping the door open for genuine participants.
How Logora helps
Logora applies toxicity detection to incoming contributions and routes the uncertain ones to a human queue. Moderators can issue configurable bans of different lengths: 1 day, 1 week, 1 month, or permanent, so a publisher can match the sanction to the severity and the offender’s history.
Related concepts
- Toxicity detection, the automated signal behind flagging
- Hate speech detection
- Content moderation, the wider practice
- Shadow banning, a softer alternative to a hard ban