How statements are moderated

Public rules for the uniteTOlove pilot conversation, 15–23 August 2026. → back to the conversation

The one thing to know first

Nothing you write appears until a person has read it and approved it.

This conversation runs in strict mode. When you add a statement, it goes into a queue. A human reads it, checks it against the rules below, and either publishes it — where everyone can then agree or disagree with it — or rejects it. No software decides this. Nothing is auto-published.

That means there is a wait. It is usually hours, and it can be up to a day. Your statement is not lost while it waits.

What gets published

Almost everything. The bar is low on purpose, because the statements people write themselves are the point of the whole exercise.

A statement is published if it is:

• A statement — something a reader can agree or disagree with.
• About Toronto — the city, its future, how it works, how it feels to live here.
• New — it says something the conversation does not already have.
• Aimed at ideas, not people.

That's it. You do not need to be polite about institutions. You do not need to be right — a statement everyone disagrees with is useful information, and disagreement is what this conversation is for. You do not need to write well. Angry statements get published. Statements that criticize the city, its government, its police, its landlords, its developers — or this project — get published.

We do not reject statements for being wrong, unpopular, or uncomfortable. If we did, the map this conversation produces would be a map of what the moderator thinks, and it would be worthless.

What gets rejected — the whole list

There are five reasons, and only five. Every rejection is one of these.

1. It's already in the conversation (the big one)

Why: if the same idea is in ten times, it splits everyone's votes ten ways and the results get distorted. Whichever version arrived first stays.

Example: the conversation already has "Bike lanes should be added even where they take away a lane of car traffic." A new statement reading "We need more bike lanes even if drivers lose a lane" is the same idea and gets rejected.

Not the same idea: "Bike lanes should be built on side streets instead of main roads." That is a different claim and gets published.

Expect this to be most of the rejections. In a comparable US city conversation in 2025, about three in ten submissions were rejected this way, and about half of all submissions were published.

2. It isn't a statement

Why: you can't agree or disagree with a question, a reply, or a greeting. There is nothing for anyone to vote on.

Examples that get rejected: "Why doesn't the city fix this?" · "I agree with the person who wrote about rent." · "Hi everyone, glad to be here." · a bare link with no claim attached.

How to fix it: turn it into a claim. "Why doesn't the city fix this?" becomes "The city is too slow to fix things residents report."

3. It attacks a person, or uses a slur

Why: this conversation criticizes decisions, institutions and policies with no limit at all. It does not attack people. A slur ends someone's participation, which is the opposite of the point.

Rejected: name-calling aimed at a person · slurs or demeaning language about any group, including groups you belong to · statements whose content is contempt for a category of people rather than a claim about the city.

Published: "The mayor's office has failed on housing." · "Police leadership is not accountable to residents." · "This project's whole approach is naive." Harsh criticism of people in their public roles is fair and stays.

The line: describe conduct, not character.

4. It identifies a private person

Why: a public conversation is not a place to put a neighbour's name, address, or licence plate. It is unfair to them and they cannot answer.

Rejected: a named private individual · a specific home address · a photo description or detail that identifies one household · anything naming a private person as an alleged wrongdoer.

Published: public officials by name and title, in their public roles · a street, an intersection, a neighbourhood, a building with many units.

How to fix it: drop the name and keep the point. "The landlord at [address] won't fix the heat" becomes "Landlords face no real consequence for leaving tenants without heat."

5. It isn't about Toronto

Why: this conversation is one city's, and off-topic statements pull the map apart.

Rejected: advertising · statements about national or foreign politics with no Toronto connection · unrelated spam.

Published if it lands here: a provincial or federal issue as it affects Toronto is on-topic — much of what the city needs sits with governments above it, and saying so is on-topic.

Who decides, and how it's recorded

• The moderator is a person, not a program. The bench for this pilot is two to three people for nine days. We say two to three rather than implying a rota that does not exist, and we do not publish their names.
• The queue is cleared once a day, in a short shift. If nobody clears it for 24 hours, statements wait. They are never auto-published to catch up, and they are never bulk-rejected to clear a backlog.
• Every decision is logged: the statement text, the decision, the reason (one of the five above), and the date. The log publishes with the pilot's results — including a count of rejections by reason, so anyone can see whether the "already in the conversation" class was used to quietly suppress a viewpoint.
• The moderator does not know who wrote what. The software shows statement text without author identity, so a decision cannot be made about a person.
• The moderator moderates their own statements by the same rules, and does not get a faster lane.

If you think a rejection was wrong

Write to unknownsoldier@unitetolove.ca with the statement, as close to word-for-word as you remember it.

What happens: a person who is not the original moderator re-reads it against the five rules and answers you. If it should have been published, it goes up, and the reversal is recorded in the published log as a reversal. If it should not have, you get told which of the five rules applied and why — not a form reply.

We expect to get some of these wrong. A small bench moderating a live conversation makes mistakes, and a published log with reversals in it is more honest than a clean one.

What we do with the results

Findings are reported as agreement rates on individual statements — how many people agreed with each one, including within groups that voted differently from each other. Statements that divide the city are published as prominently as statements that unite it.

We will not publish a claim that "Toronto has N camps." The clusters this kind of software draws are an illustration; run the same data again and they can come out differently, while the agreement rates cannot. That rule was set before the conversation opened, not after seeing what it said.

→ back to the conversation · where the starting statements came from